gpml-matlab-v3.0-2010-07-23/000755 000765 000024 00000000000 14222662506 015056 5ustar00carlstaff000000 000000 gpml-matlab-v3.0-2010-07-23/gp.m000644 000765 000024 00000016400 11422313220 015624 0ustar00carlstaff000000 000000 function [varargout] = gp(hyp, inf, mean, cov, lik, x, y, xs, ys) % Gaussian Process inference and prediction. The gp function provides a % flexible framework for Bayesian inference and prediction with Gaussian % processes for scalar targets, i.e. both regression and binary % classification. The prior is Gaussian process, defined through specification % of its mean and covariance function. The likelihood function is also % specified. Both the prior and the likelihood may have hyperparameters % associated with them. % % Two modes are possible: training or prediction: if no test cases are % supplied, then the negative log marginal likelihood and its partial % derivatives w.r.t. the hyperparameters is computed; this mode is used to fit the % hyperparameters. If test cases are given, then the test set predictive % probabilities are returned. Usage: % % training: [nlZ dnlZ ] = gp(hyp, inf, mean, cov, lik, x, y); % prediction: [ymu ys2 fmu fs2 ] = gp(hyp, inf, mean, cov, lik, x, y, xs); % or: [ymu ys2 fmu fs2 lp] = gp(hyp, inf, mean, cov, lik, x, y, xs, ys); % % where: % % hyp column vector of hyperparameters % inf function specifying the inference method % cov prior covariance function (see below) % mean prior mean function % lik likelihood function % x n by D matrix of training inputs % y column vector of length n of training targets % xs ns by D matrix of test inputs % ys column vector of length nn of test targets % % nlZ returned value of the negative log marginal likelihood % dnlZ column vector of partial derivatives of the negative % log marginal likelihood w.r.t. each hyperparameter % ymu column vector (of length ns) of predictive output means % ys2 column vector (of length ns) of predictive output variances % fmu column vector (of length ns) of predictive latent means % fs2 column vector (of length ns) of predictive latent variances % lp column vector (of length ns) of log predictive probabilities % % post struct representation of the (approximate) posterior % 3rd output in training mode and 6th output in prediction mode % % See also covFunctions.m, infMethods.m, likFunctions.m, meanFunctions.m. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23 if nargin<7 || nargin>9 disp('Usage: [nlZ dnlZ ] = gp(hyp, inf, mean, cov, lik, x, y);') disp(' or: [ymu ys2 fmu fs2 ] = gp(hyp, inf, mean, cov, lik, x, y, xs);') disp(' or: [ymu ys2 fmu fs2 lp] = gp(hyp, inf, mean, cov, lik, x, y, xs, ys);') return end if isempty(inf), inf = @infExact; else % set default inf if iscell(inf), inf = inf{1}; end % cell input is allowed if ischar(inf), inf = str2func(inf); end % convert into function handle end if isempty(mean), mean = {@meanZero}; end % set default mean if ischar(mean) || isa(mean, 'function_handle'), mean = {mean}; end % make cell if isempty(cov), error('Covariance function cannot be empty'); end % no default if ischar(cov) || isa(cov, 'function_handle'), cov = {cov}; end % make cell if isempty(lik), lik = @likGauss; else % set default lik if iscell(lik), lik = lik{1}; end % cell input is allowed if ischar(lik), lik = str2func(lik); end % convert into function handle end [n D] = size(x); if ~isfield(hyp,'mean'), hyp.mean = []; end % check the hyp specification if eval(feval(mean{:})) ~= numel(hyp.mean) error('Number of mean function hyperparameters disagree with mean function') end if ~isfield(hyp,'cov'), hyp.cov = []; end if eval(feval(cov{:})) ~= numel(hyp.cov) error('Number of cov function hyperparameters disagree with cov function') end if ~isfield(hyp,'lik'), hyp.lik = []; end if eval(feval(lik)) ~= numel(hyp.lik) error('Number of lik function hyperparameters disagree with lik function') end try % call the inference method % issue a warning if a classification likelihood is used in conjunction with % labels different from +1 and -1 if strcmp(func2str(lik),'likErf') || strcmp(func2str(lik),'likLogistic') uy = unique(y); if any( uy~=+1 & uy~=-1 ) warning('You attempt classification using labels different from {+1,-1}\n') end end if nargin>7 % compute marginal likelihood and its derivatives only if needed post = inf(hyp, mean, cov, lik, x, y); else if nargout==1 [post nlZ] = inf(hyp, mean, cov, lik, x, y); dnlZ = {}; else [post nlZ dnlZ] = inf(hyp, mean, cov, lik, x, y); end end catch if nargin > 7, error('Inference method failed'); else warning('Inference method failed .. attempting to continue') % when training dnlZ = struct('cov',0*hyp.cov, 'mean',0*hyp.mean, 'lik',0*hyp.lik); varargout = {NaN, dnlZ}; return % continue with a warning end end if nargin==7 % if no test cases are provided varargout = {nlZ, dnlZ, post}; % report -log marg lik, derivatives and post else alpha = post.alpha; L = post.L; sW = post.sW; if issparse(alpha) % handle things for sparse representations nz = alpha ~= 0; % determine nonzero indices if issparse(L), L = full(L(nz,nz)); end % convert L and sW if necessary if issparse(sW), sW = full(sW(nz)); end else nz = true(n,1); end % non-sparse representation if numel(L)==0 % in case L is not provided, we compute it K = feval(cov{:}, hyp.cov, x(nz,:)); L = chol(eye(sum(nz))+sW*sW'.*K); end Ltril = all(all(tril(L,-1)==0)); % is L an upper triangular matrix? ns = size(xs,1); % number of data points nperbatch = 1000; % number of data points per mini batch nact = 0; % number of already processed test data points ymu = zeros(ns,1); ys2 = ymu; fmu = ymu; fs2 = ymu; lp = ymu; % allocate mem while nact use Cholesky parameters (alpha,sW,L) V = L'\(repmat(sW,1,length(id)).*Ks); fs2(id) = kss - sum(V.*V,1)'; % predictive variances else % L is not triangular => use alternative parametrisation fs2(id) = kss + sum(Ks.*(L*Ks),1)'; % predictive variances end fs2(id) = max(fs2(id),0); % remove numerical noise i.e. negative variances if nargin<9 [lp(id) ymu(id) ys2(id)] = lik(hyp.lik, [], fmu(id), fs2(id)); else [lp(id) ymu(id) ys2(id)] = lik(hyp.lik, ys(id), fmu(id), fs2(id)); end nact = id(end); % set counter to index of last processed data point end if nargin<9 varargout = {ymu, ys2, fmu, fs2, [], post}; % assign output arguments else varargout = {ymu, ys2, fmu, fs2, lp, post}; end end gpml-matlab-v3.0-2010-07-23/mean/000755 000765 000024 00000000000 13330425170 015767 5ustar00carlstaff000000 000000 gpml-matlab-v3.0-2010-07-23/util/000755 000765 000024 00000000000 13330425170 016024 5ustar00carlstaff000000 000000 gpml-matlab-v3.0-2010-07-23/likFunctions.m000644 000765 000024 00000007213 11422313220 017670 0ustar00carlstaff000000 000000 % likelihood functions are provided to be used by the gp.m function: % % likErf (Error function, classification, probit regression) % likLogistic (Logistic, classification, logit regression) % % likGauss (Gaussian, regression) % likLaplace (Laplacian or double exponential, regression) % likSech2 (Sech-square, regression) % likT (Student's t, regression) % % The likelihood functions have three possible modes, the mode being selected % as follows (where "lik" stands for any likelihood function in "lik/lik*.m".): % % 1) With one or no input arguments: [REPORT NUMBER OF HYPERPARAMETERS] % % s = lik OR s = lik(hyp) % % The likelihood function returns a string telling how many hyperparameters it % expects, using the convention that "D" is the dimension of the input space. % For example, calling "likLogistic" returns the string '0'. % % % 2) With three or four input arguments: [PREDICTION MODE] % % lp = lik(hyp, y, mu) OR [lp, ymu, ys2] = lik(hyp, y, mu, s2) % % This allows to evaluate the predictive distribution. Let p(y_*|f_*) be the % likelihood of a test point and N(f_*|mu,s2) an approximation to the posterior % marginal p(f_*|x_*,x,y) as returned by an inference method. The predictive % distribution p(y_*|x_*,x,y) is approximated by. % q(y_*) = \int N(f_*|mu,s2) p(y_*|f_*) df_* % % lp = log( q(y) ) for a particular value of y, if s2 is [] or 0, this % corresponds to log( p(y|mu) ) % ymu and ys2 the mean and variance of the predictive marginal q(y) % note that these two numbers do not depend on a particular % value of y % All vectors have the same size. % % % 3) With five or six input arguments, the fifth being a string [INFERENCE MODE] % % [varargout] = lik(hyp, y, mu, s2, inf) OR % [varargout] = lik(hyp, y, mu, s2, inf, i) % % There are three cases for inf, namely a) infLaplace, b) infEP and c) infVB. % The last input i, refers to derivatives w.r.t. the ith hyperparameter. % % a1) [sum(lp),dlp,d2lp,d3lp] = lik(hyp, y, f, [], 'infLaplace') % lp, dlp, d2lp and d3lp correspond to derivatives of the log likelihood % log(p(y|f)) w.r.t. to the latent location f. % lp = log( p(y|f) ) % dlp = d log( p(y|f) ) / df % d2lp = d^2 log( p(y|f) ) / df^2 % d3lp = d^3 log( p(y|f) ) / df^3 % % a2) [lp_dhyp,d2lp_dhyp] = lik(hyp, y, f, [], 'infLaplace', i) % returns derivatives w.r.t. to the ith hyperparameter % lp_dhyp = d log( p(y|f) ) / (df dhyp_i) % d2lp_dhyp = d^3 log( p(y|f) ) / (df^2 dhyp_i) % % % b1) [lZ,dlZ,d2lZ] = lik(hyp, y, mu, s2, 'infEP') % let Z = \int p(y|f) N(f|mu,s2) df then % lZ = log(Z) % dlZ = d log(Z) / dmu % d2lZ = d^2 log(Z) / dmu^2 % % b2) [dlZhyp] = lik(hyp, y, mu, s2, 'infEP', i) % returns derivatives w.r.t. to the ith hyperparameter % dlZhyp = d log(Z) / dhyp_i % % % c1) [h,b,dh,db,d2h,d2b] = lik(hyp, y, [], ga, 'infVB') % ga is the variance of a Gaussian lower bound to the likelihood p(y|f). % p(y|f) \ge exp( b*s - s.^2/(2*ga) - h(ga)/2 ) \propto N(f|b*ga,ga) % The function returns the linear part b and the "scaling function" h(ga) and % derivatives dh = d h/dga, db = d b/dga, d2h = d^2 h/dga and d2b = d^2 b/dga. % % c2) [dhhyp] = lik(hyp, y, [], ga, 'infVB', i) % dhhyp = dh / dhyp_i, the derivative w.r.t. the ith hyperparameter % % Cumulative likelihoods are designed for binary classification. Therefore, they % only look at the sign of the targets y; zero values are treated as +1. % % See the help for the individual likelihood for the computations specific to % each likelihood function. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23 gpml-matlab-v3.0-2010-07-23/covFunctions.m000644 000765 000024 00000010515 11422313220 017677 0ustar00carlstaff000000 000000 % covariance functions to be use by Gaussian process functions. There are two % different kinds of covariance functions: simple and composite: % % simple covariance functions: % covConst - covariance for constant functions % covLIN - linear covariance function % covLINard - linear covariance function with ARD % covLINone - linear covariance function with bias % covMaterniso - Matern covariance function with nu=1/2, 3/2 or 5/2 % covNNone - neural network covariance function % covNoise - independent covariance function (i.e. white noise) % covPeriodic - smooth periodic covariance function (1d) with unit period % covPoly - polynomial covariance function % covPPiso - piecewise polynomial covariance function (compact support) % covRQard - rational quadratic covariance function with ARD % covRQiso - isotropic rational quadratic covariance function % covSEard - squared exponential covariance function with ARD % covSEiso - isotropic squared exponential covariance function % covSEisoU - as above but without latent scale % % composite (meta) covariance functions (see explanation at the bottom): % covScale - scaled version of a covariance function % covProd - products of covariance functions % covSum - sums of covariance functions % covADD - additive covariance function % covMask - mask some dimensions of the data % % Naming convention: all covariance functions are named "cov/cov*.m". A trailing % "iso" means isotropic, "ard" means Automatic Relevance Determination, and % "one" means that the distance measure is parameterized by a single parameter. % % The covariance functions are written according to a special convention where % the exact behaviour depends on the number of input and output arguments % passed to the function. If you want to add new covariance functions, you % should follow this convention if you want them to work with the function gp. % There are four different ways of calling the covariance functions: % % 1) With no input arguments: % % s = covNAME % % The covariance function returns a string s telling how many hyperparameters it % expects, using the convention that "D" is the dimension of the input space. % For example, calling "covRQard" returns the string '(D+2)'. % % 2) With two input arguments: % % K = covNAME(hyp, x) % % The function computes and returns the covariance matrix where hyp are % the hyperparameters and x is an n by D matrix of cases, where % D is the dimension of the input space. The returned covariance matrix is of % size n by n. % % 3) With three input arguments and two output arguments: % % [kss, Ks] = covNAME(hyp, x, xs) % % The function computes test set covariances; kss is a vector of self covariances % for the test cases in xs (of length ns) and Ks is an (n by ns) matrix of cross % covariances between training cases x and test cases xs. % % 4) With three input arguments and a single output: % % dKi = covNAME(hyp, x, i) % % The function computes and returns the n by n matrix of partial derivatives % of the training set covariance matrix with respect to hyp(i), i.e. with % respect to the hyperparameter number i. % % Covariance functions can be specified in two ways: either as a string % containing the name of the covariance function or using a cell array. For % example: % % cov = 'covRQard'; % cov = {'covRQard'}; % cov = {@covRQard}; % % are supported. Only the second and third form using the cell array can be used % for specifying composite covariance functions, made up of several % contributions. For example: % % cov = {'covScale', {'covRQiso'}}; % cov = {'covSum', {'covRQiso','covSEard','covNoise'}}; % cov = {'covProd',{'covRQiso','covSEard','covNoise'}}; % cov = {'covMask',{mask,'covSEiso'}} % q=1; cov = {'covPPiso',q}; % d=3; cov = {'covPoly',d}; % cov = {'covADD',{[1,2],'covSEiso'}}; % % specifies a covariance function which is the sum of three contributions. To % find out how many hyperparameters this covariance function requires, we do: % % feval(cov{:}) % % which returns the string '3+(D+1)+1' (i.e. the 'covRQiso' contribution uses % 3 parameters, the 'covSEard' uses D+1 and 'covNoise' a single parameter). % % See also doc/usageCov.m. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23 gpml-matlab-v3.0-2010-07-23/README000644 000765 000024 00000021604 11422313230 015723 0ustar00carlstaff000000 000000 GAUSSIAN PROCESS REGRESSION AND CLASSIFICATION Toolbox version 3.0 for GNU Octave 3.2.x and Matlab 7.x Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23. 0) HOW TO READ ============== If you want to get started without further delay, then read section 1) below and jump right to the examples in doc/index.html. 1) ABOUT THESE PROGRAMS ======================= This collection of matlab programs implement and demonstrates some of the algorithms described in a) the book by Rasmussen and Williams: "Gaussian Processes for Machine Learning", the MIT Press 2006 and in b) the article by Nickisch and Rasmussen: "Approximations for Binary Gaussian Process Classification", JMLR 2010. There are 6 code subdirectories: cov, doc, inf, lik, mean and util. cov/ contains covariance functions cov*.m => see covFunctions.m doc/ contains an index.html file providing documentation. This information is also available from http://www.GaussianProcess.org/gpml/code. Usage of mean, cov, classification and regression is demonstrated in usage*.m. Further details can be found in the developer documentation manual.pdf. inf/ contains the inference algorithms inf*.m => see infMethods.m lik/ contains the likelihood functions lik*.m => see likFunctions.m mean/ contains the mean functions mean*.m => see meanFunctions.m util/ contains optimisation routines, backward compatibility programs and small auxiliary scripts Before running the demos or any other part of the code, you should execute startup.m to add all necessary directories to your path. In Octave, you can also use the .octaverc file. 2) ABOUT MEX FILES ================== Some of the programs make use of the MEX facility in matlab for more efficient implementation. However, if you don't know about how to compile MEX files, you do not need to worry about this -- the code runs anyway. If you do compile the MEX files, this is automatically detected, and the program will run more efficiently. If you do not have a compiler installed on your system, you might want to consult [2]. Two components of the toolbox can be accelerated by MEX. First, there are some auxiliary functions (currently sq_dist.m and solve_chol.m) having a MEX equivalent. These can be compiled by either executing util/make.m from the Matlab/Octave command line. In addition to that, we provide a Makefile for Matlab in util/ that you can run by simply typing make in your shell after having supplied your Matlab path and your operating system. We recommend the make.m script because it works both under Matlab and Octave. Second, in order to use the L-BFGS minimiser for hyperparameter optimisation as an equivalent to minimize.m, you have to compile Peter Carbonetto's "Matlab interface for L-BFGS-B" [1]. The challenge here is the Fortran 77 code. We provide a Makefile suitable for Linux 32/64 bit and Mac whenever you have g77 properly installed. Compilation is done by editing util/lbfgsb/Makefile. In any case, you need to provide $MATLAB_HOME which can be found by the commands 'locate matlab' or 'find / -name "matlab"'. You can choose between two compilation modes: a) using the mex utility by Matlab [default] provide $MEX, then type 'make mex' b) without mex utility by Matlab provide $MEX_SUFFIX and $MATLAB_LIB, then type 'make nomex' In Ubuntu 10.04 LTS, the libg2c library needed for both a) and b) is not included per default. If 'ls /usr/lib/libg2c.*' does not list anything this is the case on your machine. You then whant to install the packages gcc-3.4-base and libg2c0 e.g. from http://packages.ubuntu.com/jaunty/gcc-3.4-base and http://packages.ubuntu.com/jaunty/libg2c0. After installation, you have to create a symbolic link by 'cd /usr/lib' and 'ln -s libg2c.so.0 libg2c.so'. [1] http://www.cs.ubc.ca/~pcarbo/lbfgsb-for-matlab.html [2] http://www.mathworks.com/support/compilers/R2010a 3) CURRENT VERSION ================== The current version of the programs is 3.0. Previous versions of the code are avaiable at http://www.gaussianprocess.org/gpml/code/oldcode.html. 4) DIFFERENCES TO PREVIOUS VERSIONS =================================== NEW in version 3.0, 2010-07-23 ------------------------------ A major code reorganisation effort did take place in the current release. First, classification and regression are now done by a single file gp.m which is completely generic in the likelihood. The previous regression program gpr.m corresponds to gp.m with Gaussian likelihood. Several other likelihoods for robust regresssion were added. Further, the code now supports mean functions with a similar specification mechanism as already used by the covariance functions. Previous implementations correspond to using meanZero.m. We merged the covariance functions covMatern3iso.m and covMatern5.iso into a single covMaterniso.m and added a covariance function for additive functions as well as the possibility to use only certain components of the data. Finally, we included covPPiso.m a piecewise polynomial covariance function with compact support. New likelihood functions in lik/: likGauss.m, likLaplace.m, likLogistic.m and likT.m New mean functions in mean/: meanConst.m, meanLinear.m, meanPow.m, meanProd.m, meanSum.m and meanZero.m New covariance functions in cov/: covADD.m, covMask.m, covPPiso.m The gprSRPP.m function which previously provided "Subset of Regressors" and the "Projected Process" approximation has now been removed. NEW in version 2.1, 2007-07-25 ------------------------------ covConst.m: fixed a bug which caused an error in the derivative of the log marginal likelihood for certain combinations of covariance functions and approximation methods. (Thanks to Antonio Eleuteri for reporting the problem) gauher.m: added the function "gauher.m" which was mistakenly missing from the previous release. This caused an error for certain combinations of approximation method and likelihood function. logistic.m: modified the approximation of moments calculation to use a mixture of cumulative Gaussian, rather than Gauss-Hermite quadrature, as the former turns out to be more accurate. NEW in version 2.0, 2007-06-25 ------------------------------ Some code restructuring has taken place for the classification code to make it more modular, to facilitate addition of new likelihood functions and approximations methods. Now, all classification is done using the binaryGP function, which (among other things) takes an approximation method and a likelihood function as an arguments. Thus, binaryGP replaces both binaryEPGP and binaryLapaceGP, although wrapper functions are still provided for backward compatibility. This gives added flexibility: now EP can also be used wth the logistic likelihood function (implemented using Gauss-Hermite quadrature). approxEP.m: New file, containing the Expectation Propagation approximation method, which was previously contained in binaryEPGP.m approxLA.m: New file, containing Laplaces approximation method, which was previously contained in binaryLaplace.m approximations.m: New file, help for the approximation methods. binaryEPGP.m: This file has been replaced by a wrapper (for backward compatibility) which calls the more general binaryGP function. binaryGP.m: New general function to do binary classification. binaryLaplaceGP.m: This file has been replaced by a wrapper (for backward compatibility) which calls the more general binaryGP function. covMatern3iso.m, covMatern5iso.m, covNNone.m, covRQard.m, covRQiso.m, cosSEard, covSEiso: now check more carefully, that persistent variables have the correct sizes, and some variable names have been modified. cumGauss.m: New file, containing code for the cumulative Gaussian likelihood function likelihoods.m: New file, help for likelihood functions logistic.m: New file, logistic likelihood NEW in version 1.3, 2006-09-08 ------------------------------ covRQard.m: bugfix: replaced x with x' and z with z' in line 36 covRQiso.m: bugfix: replaced x with x' and z with z' in line 28 minimize.m: correction: replaced "error()" with "error('')", and made a few cosmetic changes binaryEPGP.m: added the line "lml = -n*log(2);" in line 77. This change should be largely inconsequential, but occationally may save things when the covariance matrix is exceptionally badly conditioned. NEW in version 1.2, 2006-05-10 ------------------------------ added the "erfint" function to "binaryLaplaceGP.m". The erfint function was missing by mistake, preventing the use of the "logistic" likelihood. NEW in version 1.1, 2006-04-12 ------------------------------ added files: "covProd.m" and "covPeriodic.m" changes: "covSEiso.m" was changed slightly to avoid the use of persistent variables NEW in version 1.0, 2006-03-29 ------------------------------ initial version shipped with the book gpml-matlab-v3.0-2010-07-23/inf/000755 000765 000024 00000000000 13330425166 015630 5ustar00carlstaff000000 000000 gpml-matlab-v3.0-2010-07-23/infMethods.m000644 000765 000024 00000003476 11422313220 017327 0ustar00carlstaff000000 000000 % Inference methods: Compute the (approximate) posterior for a Gaussian process. % Methods currently implemented include: % % infExact Exact inference (only possible with Gaussian likelihood) % infLaplace Laplace's Approximation % infEP Expectation Propagation % infVB Variational Bayes % % The interface to the approximation methods is the following: % % function [post nlZ dnlZ] = inf..(hyp, cov, lik, x, y) % % where: % % hyp is a struct of hyperparameters % cov is the name of the covariance function (see covFunctions.m) % lik is the name of the likelihood function (see likFunctions.m) % x is a n by D matrix of training inputs % y is a (column) vector (of size n) of targets % % nlZ is the returned value of the negative log marginal likelihood % dnlZ is a (column) vector of partial derivatives of the negative % log marginal likelihood w.r.t. each hyperparameter % post struct representation of the (approximate) posterior containing % alpha is a (sparse or full column vector) containing inv(K)*m, where K % is the prior covariance matrix and m the approx posterior mean % sW is a (sparse or full column) vector containing diagonal of sqrt(W) % the approximate posterior covariance matrix is inv(inv(K)+W) % L is a (sparse or full) matrix, L = chol(sW*K*sW+eye(n)) % % Usually, the approximate posterior to be returned admits the form % N(m=K*alpha, V=inv(inv(K)+W)), where alpha is a vector and W is diagonal; % if not, then L contains instead -inv(K+inv(W)), and sW is unused. % % For more information on the individual approximation methods and their % implementations, see the separate inf??.m files. See also gp.m % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23 gpml-matlab-v3.0-2010-07-23/meanFunctions.m000644 000765 000024 00000003112 11422313220 020023 0ustar00carlstaff000000 000000 % mean functions to be use by Gaussian process functions. There are two % different kinds of mean functions: simple and composite: % % simple mean functions: % % meanZero - zero mean function % meanOne - one mean function % meanConst - constant mean function % meanLinear - linear mean function % % composite covariance functions (see explanation at the bottom): % % meanScale - scaled version of a mean function % meanPow - power of a mean function % meanProd - products of mean functions % meanSum - sums of mean functions % meanMask - mask some dimensions of the data % % Naming convention: all mean functions are named "mean/mean*.m". % % % 1) With no or only a single input argument: % % s = meanNAME or s = meanNAME(hyp) % % The mean function returns a string s telling how many hyperparameters hyp it % expects, using the convention that "D" is the dimension of the input space. % For example, calling "meanLinear" returns the string 'D'. % % 2) With two input arguments: % % m = meanNAME(hyp, x) % % The function computes and returns the mean vector where hyp are the % hyperparameters and x is an n by D matrix of cases, where D is the dimension % of the input space. The returned mean vector is of size n by 1. % % 3) With three input arguments: % % dm = meanNAME(hyp, x, i) % % The function computes and returns the n by 1 vector of partial derivatives % of the mean vector w.r.t. hyp(i) i.e. hyperparameter number i. % % See also doc/usageMean.m. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23 gpml-matlab-v3.0-2010-07-23/lik/000755 000765 000024 00000000000 13330425166 015633 5ustar00carlstaff000000 000000 gpml-matlab-v3.0-2010-07-23/doc/000755 000765 000024 00000000000 13330425166 015621 5ustar00carlstaff000000 000000 gpml-matlab-v3.0-2010-07-23/Copyright000644 000765 000024 00000003375 11422240326 016750 0ustar00carlstaff000000 000000 GAUSSIAN PROCESS REGRESSION AND CLASSIFICATION Toolbox version 3.0 for GNU Octave 3.2.x and Matlab 7.x Copyright (c) 2005-2010 Carl Edward Rasmussen & Hannes Nickisch. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. THIS SOFTWARE IS PROVIDED BY CARL EDWARD RASMUSSEN & HANNES NICKISCH ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL CARL EDWARD RASMUSSEN & HANNES NICKISCH OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. The views and conclusions contained in the software and documentation are those of the authors and should not be interpreted as representing official policies, either expressed or implied, of Carl Edward Rasmussen & Hannes Nickisch. The code and associated documentation is available from http://gaussianprocess.org/gpml/code. gpml-matlab-v3.0-2010-07-23/cov/000755 000765 000024 00000000000 13330425166 015643 5ustar00carlstaff000000 000000 gpml-matlab-v3.0-2010-07-23/startup.m000644 000765 000024 00000001410 11422240326 016721 0ustar00carlstaff000000 000000 % startup script to make Octave/Matlab aware of the GPML package % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch 2010-07-05. disp(['executing gpml startup script...']); OCT = exist('OCTAVE_VERSION') ~= 0; % check if we run Matlab or Octave me = mfilename; % what is my filename mydir = which(me); mydir = mydir(1:end-2-numel(me)); % where am I located if OCT && numel(mydir)==2 if strcmp(mydir,'./'), mydir = [pwd,mydir(2:end)]; end end % OCTAVE 3.0.x relative, MATLAB and newer have absolute path addpath(mydir(1:end-1)) addpath([mydir,'cov']) addpath([mydir,'doc']) addpath([mydir,'inf']) addpath([mydir,'lik']) addpath([mydir,'mean']) addpath([mydir,'util']) clear me mydir gpml-matlab-v3.0-2010-07-23/cov/covLIN.m000644 000765 000024 00000001245 11422240324 017144 0ustar00carlstaff000000 000000 function [A, B] = covLIN(hyp, x, z) % Linear covariance function. The covariance function is parameterized as: % % k(x^p,x^q) = x^p'*x^q % % The are no hyperparameters: % % hyp = [ ] % % Note that there is no bias or scale term; use covConst to add these. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-01-21. % % See also COVFUNCTIONS.M. if nargin<2, A = '0'; return; end % report number of parameters if nargin==2 A = x*x'; elseif nargout==2 % compute test set covariances A = sum(z.*z,2); B = x*z'; else % compute derivative matrices A = []; end gpml-matlab-v3.0-2010-07-23/cov/covLINard.m000644 000765 000024 00000001727 11422240324 017640 0ustar00carlstaff000000 000000 function [A, B] = covLINard(hyp, x, z) % Linear covariance function with Automatic Relevance Determination (ARD). The % covariance function is parameterized as: % % k(x^p,x^q) = x^p'*inv(P)*x^q % % where the P matrix is diagonal with ARD parameters ell_1^2,...,ell_D^2, where % D is the dimension of the input space. The hyperparameters are: % % hyp = [ log(ell_1) % log(ell_2) % .. % log(ell_D) ] % % Note that there is no bias term; use covConst to add a bias. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-18. % % See also COVFUNCTIONS.M. if nargin<2, A = 'D'; return; end % report number of parameters ell = exp(hyp); x = x*diag(1./ell); if nargin==2 A = x*x'; elseif nargout==2 % compute test set covariances z = z*diag(1./ell); A = sum(z.*z,2); B = x*z'; else % compute derivative matrices A = -2*x(:,z)*x(:,z)'; end gpml-matlab-v3.0-2010-07-23/cov/covPoly.m000644 000765 000024 00000002225 11422240324 017444 0ustar00carlstaff000000 000000 function [A, B] = covPoly(d, hyp, x, z) % Polynomial covariance function. The covariance function is parameterized as: % % k(x^p,x^q) = sf^2 * ( c + (x^p)'*(x^q) )^d % % The hyperparameters are: % % hyp = [ log(c) % log(sf) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-01-12. % % See also COVFUNCTIONS.M. if nargin<3, A = '2'; return; end % report number of parameters c = exp(hyp(1)); % inhomogeneous offset sf2 = exp(2*hyp(2)); % signal variance if d~=max(1,fix(d)), error('only nonzero integers allowed for d'), end % degree if nargin==3 A = sf2*( c + x*x' ).^d; elseif nargout==2 % compute test set covariances A = sf2*( c + sum(z.^2,2) ).^d; B = sf2*( c + x*z' ).^d; else % compute derivative matrix if z==1 % first parameter A = c*d*sf2*( c + x*x' ).^(d-1); else % second parameter A = 2*sf2*( c + x*x' ).^d; end endgpml-matlab-v3.0-2010-07-23/cov/covLINone.m000644 000765 000024 00000001652 11422240324 017650 0ustar00carlstaff000000 000000 function [A, B] = covLINone(hyp, x, z) % Linear covariance function with a single hyperparameter. The covariance % function is parameterized as: % % k(x^p,x^q) = (x^p'*x^q + 1)/t2; % % where the P matrix is t2 times the unit matrix. The second term plays the % role of the bias. The hyperparameter is: % % hyp = [ log(sqrt(t2)) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-18. % % See also COVFUNCTIONS.M. if nargin<2, A = '1'; return; end % report number of parameters it2 = exp(-2*hyp); % t2 inverse if nargin==2 % compute covariance A = it2*(1+x*x'); elseif nargout==2 % compute test set covariances A = it2*(1+sum(z.*z,2)); B = it2*(1+x*z'); else % compute derivative matrix A = -2*it2*(1+x*x'); end gpml-matlab-v3.0-2010-07-23/cov/covSEisoU.m000644 000765 000024 00000001711 11422240324 017667 0ustar00carlstaff000000 000000 function [A, B] = covSEisoU(hyp, x, z) % Squared Exponential covariance function with isotropic distance measure with % unit magnitude. The covariance function is parameterized as: % % k(x^p,x^q) = exp(-(x^p - x^q)'*inv(P)*(x^p - x^q)/2) % % where the P matrix is ell^2 times the unit matrix and sf2 is the signal % variance. The hyperparameters are: % % hyp = [ log(ell) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-18. % % See also COVFUNCTIONS.M. if nargin<2, A = '1'; return; end % report number of parameters ell = exp(hyp(1)); % characteristic length scale if nargin == 2 A = exp(-sq_dist(x'/ell)/2); elseif nargout == 2 % compute test set covariances A = ones(size(z,1),1); B = exp(-sq_dist(x'/ell,z'/ell)/2); else % compute derivative matrix A = exp(-sq_dist(x'/ell)/2).*sq_dist(x'/ell); endgpml-matlab-v3.0-2010-07-23/cov/covConst.m000644 000765 000024 00000001417 11422240324 017611 0ustar00carlstaff000000 000000 function [A, B] = covConst(hyp, x, z) % covariance function for a constant function. The covariance function is % parameterized as: % % k(x^p,x^q) = s2; % % The scalar hyperparameter is: % % hyp = [ log(sqrt(s2)) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-01-21. % % See also COVFUNCTIONS.M. if nargin<2, A = '1'; return; end % report number of parameters s2 = exp(2*hyp); % s2 n = size(x,1); if nargin==2 A = s2*ones(n,n); elseif nargout==2 && nargin==3 % compute test set covariances ns = size(z,1); A = s2*ones(ns,1); B = s2*ones(n,ns); else % compute derivative matrix A = 2*s2*ones(n,n); end gpml-matlab-v3.0-2010-07-23/cov/covADD.m000644 000765 000024 00000007021 11422240324 017110 0ustar00carlstaff000000 000000 function [A, B] = covADD(cov, hyp, x, z) % Additive covariance function using a 1d base covariance function % cov(x^p,x^q;hyp) with individual hyperparameters hyp. % % k(x^p,x^q) = \sum_{r \in R} sf_r \sum_{|I|=r} % \prod_{i \in I} cov(x^p_i,x^q_i;hyp_i) % % hyp = [ hyp_1 % hyp_2 % ... % hyp_D % log(sf_R(1)) % ... % log(sf_R(end)) ] % % where hyp_d are the parameters of the 1d covariance function which are shared % over the different values of R(1) to R(end). % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-06-20. % % See also COVFUNCTIONS.M. R = cov{1}; nh = eval(feval(cov{2})); % number of hypers per individual covariance nr = numel(R); % number of different degrees of interaction if nargin<3 % report number of hyper parameters A = ['D*', int2str(nh), '+', int2str(nr)]; return end [n,D] = size(x); % dimensionality sf2 = exp( 2*hyp(D*nh+(1:nr)) ); % signal variances of individual degrees if nargin==3 % evaluate covariance matrix K = Kd(cov{2},hyp,x); % evaluate dimensionwise covariances K EE = elsympol(K,max(R)); % Rth elementary symmetric polynomials A = 0; for i=1:nr, A = A + sf2(i)*EE(:,:,R(i)+1); end % sf2 weighted sum elseif nargout==2 % compute test set covariances [kss,Ks] = Kd(cov{2},hyp,x,z); % evaluate dimensionwise covariances kss, Ks E = elsympol(kss,max(R)); % Rth elementary symmetric polynomials A = 0; for i=1:nr, A = A + sf2(i)*E(:,:,R(i)+1); end % sf2 weighted sum E = elsympol(Ks,max(R)); B = 0; for i=1:nr, B = B + sf2(i)*E(:,:,R(i)+1); end % sf2 weighted sum else % compute derivative matrix K = Kd(cov{2},hyp,x); % evaluate dimensionwise covariances K if z <= D*nh % individual covariance function parameters j = fix(1+(z-1)/nh); % j is the dimension of the hyperparameter dKj = feval(cov{2},hyp(nh*(j-1)+(1:nh)),x(:,j),z-(j-1)*nh); % other dK zero % the final derivative is a sum of multilinear terms, so if only one term % depends on the hyperparameter under consideration, we can factorise it % out and compute the sum with one degree less E = elsympol(K(:,:,[1:j-1,j+1:D]),max(R)-1); % R-1th elementary sym polyn A = 0; for i=1:nr, A = A + sf2(i)*E(:,:,R(i)); end % sf2 weighted sum A = dKj.*A; else EE = elsympol(K,max(R)); % Rth elementary symmetric polynomials j = z-D*nh; A = 2*sf2(j)*EE(:,:,R(j)+1); % rest of the sf2 weighted sum end end % evaluate dimensionwise covariances K function [K, Ks] = Kd(cov,hyp,x,z) [n,D] = size(x); % dimensionality nh = eval(feval(cov)); % number of hypers per individual covariance if nargin==3 % allocate memory K = zeros(n,n,D); else ns = size(z,1); % dimensionality K = zeros(ns,1,D); Ks = zeros(n,ns,D); end for d=1:D hyp_d = hyp(nh*(d-1)+(1:nh)); % hyperparamter of dimension d if nargin==3 K(:,:,d) = feval(cov,hyp_d,x(:,d)); else [K(:,:,d),Ks(:,:,d)] = feval(cov,hyp_d,x(:,d),z(:,d)); end end gpml-matlab-v3.0-2010-07-23/cov/covPeriodic.m000644 000765 000024 00000002272 11422240324 020261 0ustar00carlstaff000000 000000 function [A, B] = covPeriodic(hyp, x, z) % covariance function for a smooth periodic 1d functions, with unit period. The % covariance function is: % % k(x^p, x^q) = sf2 * exp( -2*sin^2( lam*(x_p-x_q)/(2*pi) )/ell^2 ) % % where the hyperparameters are: % % hyp = [ log(ell) % log(lam) % log(sqrt(sf2)) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-02-23. % % See also COVFUNCTIONS.M. if nargin<2, A = '3'; return; end % report number of parameters n = size(x,1); ell = exp(hyp(1)); lam = exp(hyp(2)); sf2 = exp(2*hyp(3)); if nargin==2 A = sf2*exp(-2*(sin( lam*(repmat(x,1,n )-repmat(x',n,1))/(2*pi) )/ell).^2); elseif nargout==2 % compute test set covariances ns = size(z,1); A = sf2*ones(ns,1); B = sf2*exp(-2*(sin( lam*(repmat(x,1,ns)-repmat(z',n,1))/(2*pi) )/ell).^2); else % compute derivative matrices r = lam*(repmat(x,1,n)-repmat(x',n,1))/(2*pi); if z==1 r = (sin(r)/ell).^2; A = 4*sf2*exp(-2*r).*r; elseif z==2 sre = sin(r)/ell; A = -4/ell*sf2*exp(-2*sre.^2).*sre.*cos(r).*r; else A = 2*sf2*exp(-2*(sin(r)/ell).^2); end end gpml-matlab-v3.0-2010-07-23/cov/covNNone.m000644 000765 000024 00000003322 11422240324 017535 0ustar00carlstaff000000 000000 function [A, B] = covNNone(hyp, x, z) % Neural network covariance function with a single parameter for the distance % measure. The covariance function is parameterized as: % % k(x^p,x^q) = sf2 * asin(x^p'*P*x^q / sqrt[(1+x^p'*P*x^p)*(1+x^q'*P*x^q)]) % % where the x^p and x^q vectors on the right hand side have an added extra bias % entry with unit value. P is ell^-2 times the unit matrix and sf2 controls the % signal variance. The hyperparameters are: % % hyp = [ log(ell) % log(sqrt(sf2) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-18. % % See also COVFUNCTIONS.M. if nargin<2, A = '2'; return; end % report number of parameters n = size(x,1); ell = exp(hyp(1)); em2 = ell^(-2); sf2 = exp(2*hyp(2)); x = x/ell; if nargin==2 % compute covariance Q = x*x'; K = (em2+Q)./(sqrt(1+em2+diag(Q))*sqrt(1+em2+diag(Q)')); A = sf2*asin(K); elseif nargout==2 % compute test set covariances z = z/ell; A = sf2*asin((em2+sum(z.*z,2))./(1+em2+sum(z.*z,2))); B = sf2*asin((em2+x*z')./sqrt((1+em2+sum(x.*x,2))*(1+em2+sum(z.*z,2)'))); else % compute derivative matrix Q = x*x'; K = (em2+Q)./(sqrt(1+em2+diag(Q))*sqrt(1+em2+diag(Q)')); if z==1 % first parameter v = (em2+sum(x.*x,2))./(1+em2+diag(Q)); A = -2*sf2*((em2+Q)./(sqrt(1+em2+diag(Q))*sqrt(1+em2+diag(Q)'))- ... K.*(repmat(v,1,n)+repmat(v',n,1))/2)./sqrt(1-K.^2); else % second parameter A = 2*sf2*asin(K); end end gpml-matlab-v3.0-2010-07-23/cov/covRQard.m000644 000765 000024 00000003146 11422240324 017535 0ustar00carlstaff000000 000000 function [A, B] = covRQard(hyp, x, z) % Rational Quadratic covariance function with Automatic Relevance Determination % (ARD) distance measure. The covariance function is parameterized as: % % k(x^p,x^q) = sf2 * [1 + (x^p - x^q)'*inv(P)*(x^p - x^q)/(2*alpha)]^(-alpha) % % where the P matrix is diagonal with ARD parameters ell_1^2,...,ell_D^2, where % D is the dimension of the input space, sf2 is the signal variance and alpha % is the shape parameter for the RQ covariance. The hyperparameters are: % % hyp = [ log(ell_1) % log(ell_2) % .. % log(ell_D) % log(sqrt(sf2)) % log(alpha) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-18. % % See also COVFUNCTIONS.M. if nargin<2, A = '(D+2)'; return; end % report number of parameters [n,D] = size(x); ell = exp(hyp(1:D)); sf2 = exp(2*hyp(D+1)); alpha = exp(hyp(D+2)); if nargin == 2 K = (1+0.5*sq_dist(diag(1./ell)*x')/alpha); A = sf2*(K.^(-alpha)); elseif nargout == 2 % compute test set covariances A = sf2*ones(size(z,1),1); B = sf2*((1+0.5*sq_dist(diag(1./ell)*x',diag(1./ell)*z')/alpha).^(-alpha)); else % compute derivative matrix K = (1+0.5*sq_dist(diag(1./ell)*x')/alpha); if z <= D % length scale parameters A = sf2*K.^(-alpha-1).*sq_dist(x(:,z)'/ell(z)); elseif z == D+1 % magnitude parameter A = 2*sf2*(K.^(-alpha)); else A = sf2*K.^(-alpha).*(0.5*sq_dist(diag(1./ell)*x')./K - alpha*log(K)); end endgpml-matlab-v3.0-2010-07-23/cov/covSum.m000644 000765 000024 00000004445 11422240324 017273 0ustar00carlstaff000000 000000 function [A B] = covSum(cov, hyp, x, z) % covSum - compose a covariance function as the sum of other covariance % functions. This function doesn't actually compute very much on its own, it % merely does some bookkeeping, and calls other covariance functions to do the % actual work. % % Copyright (c) by Carl Edward Rasmussen & Hannes Nickisch 2009-12-17. % % See also COVFUNCTIONS.M. for i = 1:numel(cov) % iterate over covariance functions f = cov(i); if iscell(f{:}), f = f{:}; end % expand cell array if necessary j(i) = cellstr(feval(f{:})); % collect number hypers end if nargin<3 % report number of parameters A = char(j(1)); for i=2:length(cov), A = [A, '+', char(j(i))]; end; return end [n,D] = size(x); v = []; % v vector indicates to which covariance parameters belong for i = 1:length(cov), v = [v repmat(i, 1, eval(char(j(i))))]; end switch nargin case 3 % compute covariance matrix A = zeros(n); % allocate space for covariance matrix for i = 1:length(cov) % iteration over summand functions f = cov(i); if iscell(f{:}), f = f{:}; end % expand cell array if necessary A = A + feval(f{:}, hyp(v==i), x); % accumulate covariances end case 4 % compute derivative matrix or test set covariances if nargout == 2 % compute test set cavariances A = zeros(size(z,1),1); B = zeros(size(x,1),size(z,1)); % allocate space for i = 1:length(cov) f = cov(i); if iscell(f{:}) f = f{:}; end % expand cell array if necessary [AA BB] = feval(f{:}, hyp(v==i), x, z); % compute test covariances A = A + AA; B = B + BB; % and accumulate end else % compute derivative matrices i = v(z); % which covariance function j = sum(v(1:z)==i); % which parameter in that covariance f = cov(i); if iscell(f{:}), f = f{:}; end % dereference cell array if necessary A = feval(f{:}, hyp(v==i), x, j); % compute derivative end endgpml-matlab-v3.0-2010-07-23/cov/covPPiso.m000644 000765 000024 00000004006 11422240324 017552 0ustar00carlstaff000000 000000 function [A, B] = covPPiso(v, hyp, x, z) % Piecewise polynomial covariance function with compact support, v = 0,1,2,3. % The covariance functions are 2v times contin. diff'ble and the corresponding % processes are hence v times mean-square diffble. The covariance function is: % % k(x^p,x^q) = s2f * (1-r)_+.^j * f(r,j) % % where r is the distance sqrt((x^p-x^q)'*inv(P)*(x^p-x^q)), P is ell^2 times % the unit matrix and sf2 is the signal variance. The hyperparameters are: % % hyp = [ log(ell) % log(sqrt(sf2)) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-01-12. % % See also COVFUNCTIONS.M. if nargin<3, A = '2'; return; end % report number of parameters ell = exp(hyp(1)); sf2 = exp(2*hyp(2)); if all(v~=[0,1,2,3]), error('only 0,1,2 and 3 allowed for v'), end % degree j = floor(size(x,2)/2)+v+1; % exponent x = x/ell; switch v case 0, f = @(r,j) 1; df = @(r,j) 0; case 1, f = @(r,j) 1 + (j+1)*r; df = @(r,j) (j+1); case 2, f = @(r,j) 1 + (j+2)*r + ( j^2+ 4*j+ 3)/ 3*r.^2; df = @(r,j) (j+2) + 2*( j^2+ 4*j+ 3)/ 3*r; case 3, f = @(r,j) 1 + (j+3)*r + (6*j^2+36*j+45)/15*r.^2 ... + (j^3+9*j^2+23*j+15)/15*r.^3; df = @(r,j) (j+3) + 2*(6*j^2+36*j+45)/15*r ... + (j^3+9*j^2+23*j+15)/ 5*r.^2; end k = @(r,j,v,f) max(1-r,0).^(j+v).*f(r,j); dk = @(r,j,v,f) max(1-r,0).^(j+v-1).*r.*( (j+v)*f(r,j) - max(1-r,0).*df(r,j) ); if nargin==3 % compute covariance matrix A = sf2*k( sqrt(sq_dist(x')), j, v, f ); elseif nargout==2 % compute test set covariances A = sf2*ones(size(z,1),1); B = sf2*k( sqrt(sq_dist(x',z'/ell)), j, v, f ); else % compute derivative matrices if z==1 A = sf2*dk( sqrt(sq_dist(x')), j, v, f ); else A = 2*sf2* k( sqrt(sq_dist(x')), j, v, f ); end endgpml-matlab-v3.0-2010-07-23/cov/covMask.m000644 000765 000024 00000003023 11422240324 017411 0ustar00carlstaff000000 000000 function [A, B] = covMask(cov, hyp, x, z) % Apply a covariance function to a subset of the dimensions only. The subset can % either be specified by a 0/1 mask by a boolean mask or by an index set. % % This function doesn't actually compute very much on its own, it merely does % some bookkeeping, and calls another covariance function to do the actual work. % % The function was suggested by Iain Murray, 2010-02-18 and is based on an % earlier implementation of his dating back to 2009-06-16. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-06-20. % % See also COVFUNCTIONS.M. mask = cov{1}(:); % either a binary mask or an index set cov = cov(2); % covariance function to be masked nh_string = feval(cov{:}); % number of hyperparameters of the full covariance if max(mask)<2 && length(mask)>1, mask = boolean(mask); end % convert 1/0->T/F if islogical(mask), D = sum(mask); else D = length(mask); end % masked dimension if nargin<3, A = num2str(eval(nh_string)); return; end % number of parameters if eval(nh_string)~=length(hyp) % check hyperparameters error('number of hyperparameters does not match size of masked data') end if nargin==3 A = feval(cov{:}, hyp, x(:,mask)); elseif nargout==2 % compute test set covariances [A, B] = feval(cov{:}, hyp, x(:,mask), z(:,mask)); else % compute derivative matrix A = feval(cov{:}, hyp, x(:,mask), z); endgpml-matlab-v3.0-2010-07-23/cov/covScale.m000644 000765 000024 00000002056 11422240324 017552 0ustar00carlstaff000000 000000 function [A, B] = covScale(cov, hyp, x, z) % meanScale - compose a mean function as a scaled version of another one. % % k(x^p,x^q) = sf^2 * k_0(x^p,x^q) % % The hyperparameter is: % % hyp = [ log(sf) ] % % This function doesn't actually compute very much on its own, it merely does % some bookkeeping, and calls other mean function to do the actual work. % % Copyright (c) by Carl Edward Rasmussen & Hannes Nickisch 2010-07-15. % % See also MEANFUNCTIONS.M. if nargin<3 % report number of parameters A = [feval(cov{:}),'+1']; return end [n,D] = size(x); sf2 = exp(2*hyp(1)); % signal variance if nargin==3 A = sf2*feval(cov{:},hyp(2:end),x); elseif nargout==2 % compute test set covariances [A, B] = feval(cov{:},hyp(2:end),x,z); A = sf2*A; B = sf2*B; else % compute derivative vector if z==1 A = 2*sf2*feval(cov{:},hyp(2:end),x); else A = sf2*feval(cov{:},hyp(2:end),x,z-1); end endgpml-matlab-v3.0-2010-07-23/cov/covRQiso.m000644 000765 000024 00000002722 11422240324 017560 0ustar00carlstaff000000 000000 function [A, B] = covRQiso(hyp, x, z) % Rational Quadratic covariance function with isotropic distance measure. The % covariance function is parameterized as: % % k(x^p,x^q) = sf2 * [1 + (x^p - x^q)'*inv(P)*(x^p - x^q)/(2*alpha)]^(-alpha) % % where the P matrix is ell^2 times the unit matrix, sf2 is the signal % variance and alpha is the shape parameter for the RQ covariance. The % hyperparameters are: % % hyp = [ log(ell) % log(sqrt(sf2)) % log(alpha) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-18. % % See also COVFUNCTIONS.M. if nargin<2, A = '3'; return; end % report number of parameters n = size(x,1); ell = exp(hyp(1)); sf2 = exp(2*hyp(2)); alpha = exp(hyp(3)); if nargin == 2 % compute covariance matrix K = (1+0.5*sq_dist(x'/ell)/alpha); A = sf2*(K.^(-alpha)); elseif nargout == 2 % compute test set covariances A = sf2*ones(size(z,1),1); B = sf2*((1+0.5*sq_dist(x'/ell,z'/ell)/alpha).^(-alpha)); else % compute derivative matrices K = (1+0.5*sq_dist(x'/ell)/alpha); if z == 1 % length scale parameters A = sf2*K.^(-alpha-1).*sq_dist(x'/ell); elseif z == 2 % magnitude parameter A = 2*sf2*(K.^(-alpha)); else A = sf2*K.^(-alpha).*(0.5*sq_dist(x'/ell)./K - alpha*log(K)); end endgpml-matlab-v3.0-2010-07-23/cov/covNoise.m000644 000765 000024 00000002023 11422240324 017572 0ustar00carlstaff000000 000000 function [A, B] = covNoise(hyp, x, z) % Independent covariance function, ie "white noise", with specified variance. % The covariance function is specified as: % % k(x^p,x^q) = s2 * \delta(p,q) % % where s2 is the noise variance and \delta(p,q) is a Kronecker delta function % which is 1 iff p=q and zero otherwise. The hyperparameter is % % hyp = [ log(sqrt(s2)) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-17. % % See also COVFUNCTIONS.M. if nargin<2, A = '1'; return; end % report number of hyperparameters n = size(x,1); s2 = exp(2*hyp); % noise variance if nargin==2 % compute covariance matrix A = s2*eye(n); elseif nargout==2 % compute test set covariances A = s2*ones(size(z,1),1); B = zeros(n,size(z,1)); % zeros cross covariance by independence else % compute derivative matrix A = 2*s2*eye(n); end gpml-matlab-v3.0-2010-07-23/cov/covSEiso.m000644 000765 000024 00000002517 11422240324 017547 0ustar00carlstaff000000 000000 function [A, B] = covSEiso(hyp, x, z) % Squared Exponential covariance function with isotropic distance measure. The % covariance function is parameterized as: % % k(x^p,x^q) = sf^2 * exp(-(x^p - x^q)'*inv(P)*(x^p - x^q)/2) % % where the P matrix is ell^2 times the unit matrix and sf^2 is the signal % variance. The hyperparameters are: % % hyp = [ log(ell) % log(sf) ] % % For more help on design of covariance functions, try "help covFunctions". % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-17. % % See also COVFUNCTIONS.M. if nargin<2, A = '2'; return; end % report number of parameters ell = exp(hyp(1)); % characteristic length scale sf2 = exp(2*hyp(2)); % signal variance if nargin==2 A = sf2*exp(-sq_dist(x'/ell)/2); elseif nargout==2 % compute test set covariances A = sf2*ones(size(z,1),1); B = sf2*exp(-sq_dist(x'/ell,z'/ell)/2); else % compute derivative matrix if z==1 % first parameter A = sf2*exp(-sq_dist(x'/ell)/2).*sq_dist(x'/ell); else % second parameter A = 2*sf2*exp(-sq_dist(x'/ell)/2); end endgpml-matlab-v3.0-2010-07-23/cov/covProd.m000644 000765 000024 00000004736 11422240324 017436 0ustar00carlstaff000000 000000 function [A, B] = covProd(cov, hyp, x, z) % covProd - compose a covariance function as the product of other covariance % functions. This function doesn't actually compute very much on its own, it % merely does some bookkeeping, and calls other covariance functions to do the % actual work. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-18. % % See also COVFUNCTIONS.M. for i = 1:length(cov) % iterate over covariance functions f = cov(i); if iscell(f{:}), f = f{:}; end % dereference cell array if necessary j(i) = cellstr(feval(f{:})); end if nargin<3 % report number of parameters A = char(j(1)); for i=2:length(cov), A = [A, '+', char(j(i))]; end return end [n,D] = size(x); v = []; % v vector indicates to which covariance parameters belong for i = 1:length(cov), v = [v repmat(i, 1, eval(char(j(i))))]; end switch nargin case 3 % compute covariance matrix A = ones(n, n); % allocate space for covariance matrix for i = 1:length(cov) % iteration over factor functions f = cov(i); if iscell(f{:}), f = f{:}; end % dereference cell array if necessary A = A .* feval(f{:}, hyp(v==i), x); % multiply covariances end case 4 % compute derivative matrix or test set covariances if nargout == 2 % compute test set cavariances A = ones(size(z,1),1); B = ones(size(x,1),size(z,1)); % allocate space for i = 1:length(cov) f = cov(i); if iscell(f{:}), f = f{:}; end % dereference cell array if necessary [AA BB] = feval(f{:}, hyp(v==i), x, z); % compute test covariances A = A .* AA; B = B .* BB; % and accumulate end else % compute derivative matrices A = ones(n, n); ii = v(z); % which covariance function j = sum(v(1:z)==ii); % which parameter in that covariance for i = 1:length(cov) f = cov(i); if iscell(f{:}), f = f{:}; end % dereference cell array if necessary if i == ii A = A .* feval(f{:}, hyp(v==i), x, j); % multiply derivative else A = A .* feval(f{:}, hyp(v==i), x); % multiply covariance end end end endgpml-matlab-v3.0-2010-07-23/cov/covSEard.m000644 000765 000024 00000002747 11422240324 017530 0ustar00carlstaff000000 000000 function [A, B] = covSEard(hyp, x, z) % Squared Exponential covariance function with Automatic Relevance Detemination % (ARD) distance measure. The covariance function is parameterized as: % % k(x^p,x^q) = sf2 * exp(-(x^p - x^q)'*inv(P)*(x^p - x^q)/2) % % where the P matrix is diagonal with ARD parameters ell_1^2,...,ell_D^2, where % D is the dimension of the input space and sf2 is the signal variance. The % hyperparameters are: % % hyp = [ log(ell_1) % log(ell_2) % . % log(ell_D) % log(sqrt(sf2)) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2009-12-18. % % See also COVFUNCTIONS.M. if nargin<2, A = '(D+1)'; return; end % report number of parameters [n,D] = size(x); ell = exp(hyp(1:D)); % characteristic length scale sf2 = exp(2*hyp(D+1)); % signal variance if nargin == 2 K = sf2*exp(-sq_dist(diag(1./ell)*x')/2); A = K; elseif nargout == 2 % compute test set covariances A = sf2*ones(size(z,1),1); B = sf2*exp(-sq_dist(diag(1./ell)*x',diag(1./ell)*z')/2); else % compute derivative matrix K = sf2*exp(-sq_dist(diag(1./ell)*x')/2); if z <= D % length scale parameters A = K.*sq_dist(x(:,z)'/ell(z)); else % magnitude parameter A = 2*K; end endgpml-matlab-v3.0-2010-07-23/cov/covMaterniso.m000644 000765 000024 00000003136 11422240324 020464 0ustar00carlstaff000000 000000 function [A, B] = covMaterniso(d, hyp, x, z) % Matern covariance function with nu = d/2 and isotropic distance measure. For % d=1 the function is also known as the exponential covariance function or the % Ornstein-Uhlenbeck covariance in 1d. The covariance function is: % % k(x^p,x^q) = s2f * f( sqrt(d)*r ) * exp(-sqrt(d)*r) % % with f(t)=1 for d=1, f(t)=1+t for d=3 and f(t)=1+t+t²/3 for d=5. % Here r is the distance sqrt((x^p-x^q)'*inv(P)*(x^p-x^q)), P is ell times % the unit matrix and sf2 is the signal variance. The hyperparameters are: % % hyp = [ log(ell) % log(sqrt(sf2)) ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-01-12. % % See also COVFUNCTIONS.M. if nargin<3, A = '2'; return; end % report number of parameters ell = exp(hyp(1)); sf2 = exp(2*hyp(2)); if all(d~=[1,3,5]), error('only 1, 3 and 5 allowed for d'), end % degree x = sqrt(d)*x/ell; switch d case 1, f = @(t) 1; df = @(t) 1; case 3, f = @(t) 1 + t; df = @(t) t; case 5, f = @(t) 1 + t.*(1+t/3); df = @(t) t.*(1+t)/3; end k = @(t,f) f(t).*exp(-t); dk = @(t,f) df(t).*t.*exp(-t); if nargin==3 % compute covariance matrix A = sf2*k( sqrt(sq_dist(x')), f ); elseif nargout==2 % compute test set covariances A = sf2*ones(size(z,1),1); B = sf2*k( sqrt(sq_dist(x',sqrt(d)*z'/ell)), f ); else % compute derivative matrices if z==1 A = sf2*dk( sqrt(sq_dist(x')), f ); else A = 2*sf2* k( sqrt(sq_dist(x')), f ); end end gpml-matlab-v3.0-2010-07-23/doc/index.html000644 000765 000024 00000072344 11422301524 017617 0ustar00carlstaff000000 000000 Documentation for GPML Matlab Code

Documentation for GPML Matlab Code

What?

The code provided here originally demonstrated the main algorithms from Rasmussen and Williams: Gaussian Processes for Machine Learning. It has since grown to allow more likelihood functions, further inference methods and a flexible framework for specifying GPs. It does not currently contain inference methods for sparse approximations. Other GP packages can be found here.

The code is written by Carl Edward Rasmussen and Hannes Nickisch; it runs on both Octave 3.2.x and Matlab® 7.x. The code is based on previous versions written by Carl Edward Rasmussen and Chris Williams.

Download, Install and Documentation

All the code including demonstrations and html documentation can be downloaded in a tar or zip archive file.

Please read the copyright notice.

After unpacking the tar or zip file you will find 6 subdirectories: cov, doc, inf, lik, mean and util. It is not necessary to install anything to get started, just run the startup.m script to set your path.

Details about the directory contents and on how to compile mex files can be found in the README. The getting started guide is the remainder of the html file you are currently reading (also available at http://gaussianprocess.org/gpml/code/matlab/doc). A Developer's Guide containing technical documentation is found in manual.pdf, but for the casual user, the guide is below.

Theory

Gaussian Processes (GPs) can conveniently be used for Bayesian supervised learning, such as regression and classification. In its simplest form, GP inference can be implemented in a few lines of code. However, in practice, things typically get a little more complicated: you might want to use complicated covariance functions and mean functions, learn good values for hyperparameters, use non-Gaussian likelihood functions (rendering exact inference intractable), use approximate inference algorithms, or combinations of many or all of the above. This is what the GPML software package does.

Before going straight to the examples, just a brief note about the organization of the package. There are four types of objects which you need to know about:

Gaussian Process
A Gaussian Process is fully specified by a mean function and a covariance function. These functions are specified separately, and consist of a specification of a functional form as well as a set of parameters called hyperparameters, see below.
Mean functions
Several mean functions are available, all start with the four letters mean and reside in the mean directory. An overview is provided by the meanFunctions help function (type help meanFunctions to get help), and an example is the meanLinear function.
Covariance functions
There are many covariance functions available, all start with the three letters cov and reside in the cov directory. An overview is provided by the covFunctions help function (type help covFunctions to get help), and an example is the covSEard "Squared Exponential with Automatic Relevance Determination" covariance function.
For both mean functions and covariance functions, two types exist: simple and composite. Whereas simple types are specified by the function name (or function pointer), composite functions join together several components using cell arrays. Composite functions can be composed of other composite functions, allowing for very flexible and interesting structures. Examples are given below and in the usageMean and usageCov functions.
Hyperparameters
GPs are typically specified using mean and covariance functions which have free parameters called hyperparameters. Also likelihood functions may have such parameters. These are encoded in a struct with the fields mean, cov and lik (some of which may be empty). When specifying hyperparameters, it is important that the number of elements in each of these struct fields, precisely match the number of parameters expected by the mean function, the covariance function and the likelihood functions respectively (otherwise an error will result). Hyperparameters whose natural domain is positive are represented by their logarithms.
Likelihood Functions
The likelihood function specifies the probability of the observations given the latent function, i.e. the GP (and the hyperparameters). All likelihood functions begin with the three letters lik and reside in the lik directory. An overview is provided by the likFunctions help function (type help likFunctions to get help). Some examples are likGauss the Gaussian likelihood and likLogistic the logistic function used in classification (a.k.a. logistic regression).
Inference Methods
The inference methods specify how to compute with the model, i.e. how to infer the (approximate) posterior process, how to find hyperparameters, evaluate the log marginal likelihood and how to make predictions. Inference methods all begin with the three letters inf and reside in the inf directory. An overview is provided by the infMethods help file (type help infMethods to get help). Some examples are infExact for exact inference or infEP for the Expectation Propagation algorithm. Further usage examples are provided for both regression and classification. However, not all combinations of likelihood function and inference method are possible (e.g. you cannot do exact inference with a Laplace likelihood). An exhaustive compatibility matrix between likelihoods (rows) and inference methods (columns) is given in the table below:
Likelihood \ Inference
 		  
GPML name		 Exact
infExact		 EP
infEP		 Laplace
infLaplace		 Variational Bayes
infVB		 type, output domain
 		 alternate names
 
Gaussian likGauss $\checkmark$ $\checkmark$ $\checkmark$ $\checkmark$ regression, IR  
Sech-squared likSech2   $\checkmark$ $\checkmark$ $\checkmark$ regression, IR  
Laplacian likLaplace   $\checkmark$   $\checkmark$ regression, IR double exponential
Student's t likT     $\checkmark$ $\checkmark$ regression, IR  
Error function likErf   $\checkmark$ $\checkmark$ $\checkmark$ classification, ±1 probit regression
Logistic function likLogistic   $\checkmark$ $\checkmark$ $\checkmark$ classification, ±1 logistic regression
logit regression

All of the objects described above are written in a modular way, so you can add functionality if you feel constrained despite the considerable flexibility provided. Details about how to do this are provided in the developer documentation.

Practice

Using the GPML package is simple. There is only one single function to call: gp, it does posterior inference, learns hyperparameters, computes the marginal likelihood and makes predictions. Generally, the gp function takes the following arguments: a hyperparameter struct, an inference method, a mean function, a covariance function, a likelihood function, training inputs, training targets, and possibly test cases. The exact computations done by the function is controlled by the number of input and output arguments in the call. Here is part of the help message for the gp function (follow the link to see the whole thing):

  function [varargout] = gp(hyp, inf, mean, cov, lik, x, y, xs, ys)

  [ ... snip ...]

  Two modes are possible: training or prediction: if no test cases are
  supplied, then the negative log marginal likelihood and its partial
  derivatives wrt the hyperparameters is computed; this mode is used to fit the
  hyperparameters. If test cases are given, then the test set predictive
  probabilities are returned. Usage:

    training: [nlZ dnlZ          ] = gp(hyp, inf, mean, cov, lik, x, y);
  prediction: [ymu ys2 fmu fs2   ] = gp(hyp, inf, mean, cov, lik, x, y, xs);
          or: [ymu ys2 fmu fs2 lp] = gp(hyp, inf, mean, cov, lik, x, y, xs, ys);

  [ .. snip ...]

Here x and y are training inputs and outputs, and xs and ys are test set inputs and outputs, nlZ is the negative log marginal likelihood and dnlZ its partial derivatives wrt the hyperparameters (which are used for training the hyperparameters). The prediction outputs are ymu and ys2 for test output mean and covariance, and fmu and fs2 are the equivalent quenteties for the corresponding latent variables. Finally, lp are the test output log probabilities.

Instead of exhaustively explaining all the possibilities, we will give two illustrative examples to give you the idea; one for regression and one for classification. You can either follow the example here on this page, or using the two scripts demoRegression and demoClassification (using the scripts, you still need to follow the explanation on this page).

Regression

You can either follow the example here on this page, or use the script demoRegression.

This is a simple example, where we first generate n=20 data points from a GP, where the inputs are scalar (so that it is easy to plot what is going on). We then use various other GPs to make inferences about the underlying function.

First, generate some data from a Gaussian process (it is not essential to understand the details of this):

  clear all, close all
 
  meanfunc = {@meanSum, {@meanLinear, @meanConst}}; hyp.mean = [0.5; 1];
  covfunc = {@covMaterniso, 3}; ell = 1/4; sf = 1; hyp.cov = log([ell; sf]);
  likfunc = @likGauss; sn = 0.1; hyp.lik = log(sn);
 
  n = 20;
  x = gpml_randn(0.3, n, 1);
  K = feval(covfunc{:}, hyp.cov, x);
  mu = feval(meanfunc{:}, hyp.mean, x);
  y = chol(K)'*gpml_randn(0.15, n, 1) + mu + exp(hyp.lik)*gpml_randn(0.2, n, 1);

  plot(x, y, '+')

Above, we first specify the mean function meanfunc, covariance function covfunc of a GP and a likelihood function, likfunc. The corresponding hyperparameters are specified in the hyp structure:

The mean function is composite, adding (using meanSum function) a linear (meanLinear) and a constant (meanConst) to get an affine function. Note, how the different components are composed using cell arrays. The hyperparameters for the mean are given in hyp.mean and consists of a single (because the input will one dimensional, i.e. D=1) slope (set to 0.5) and an off-set (set to 1). The number and the order of these hyperparameters conform to the mean function specification. You can find out how many hyperparameters a mean (or covariance or likelihood function) expects by calling it without arguments, such as feval(@meanfunc{:}). For more information on mean functions see meanFunctions and the directory mean/.

The covariance function is of the Matérn form with isotropic distance measure covMaterniso. This covariance function is also composite, as it takes a constant (related to the smoothness of the GP), which in this case is set to 3. The covariance function takes two hyperparameters, a characteristic length-scale ell and the standard deviation of the signal sf. Note, that these positive parameters are represented in hyp.cov using their logarithms. For more information on covariance functions see covFunctions and cov/.

Finally, the likelihood function is specified to be Gaussian. The standard deviation of the noise sn is set to 0.1. Again, the representation in the hyp.lik is given in terms of its logarithm. For more information about likelihood functions, see likFunctions and lik/.

Then, we generate a dataset with n=20 examples. The inputs x are drawn from a unit Gaussian (using the gpml_randn utility, which generates unit Gaussian pseudo random numbers with a specified seed). We then evaluate the covariance matrix K and the mean vector m by calling the corresponding functions with the hyperparameters and the input locations x. Finally, the targets y are computed by drawing randomly from a Gaussian with the desired covariance and mean and adding Gaussian noise with standard deviation exp(hyp.lik). The above code is a bit special because we explicitly call the mean and covariance functions (in order to generate samples from a GP); ordinarily, we would only directly call the gp function.

f0.gif

Let's ask the model to compute the (joint) negative log probability (density) nlml (also called marginal likelihood or evidence) and to generalize from the training data to other (test) inputs z:

  nlml = gp(hyp, @infExact, meanfunc, covfunc, likfunc, x, y)

  z = linspace(-1.9, 1.9, 101)';
  [m s2] = gp(hyp, @infExact, meanfunc, covfunc, likfunc, x, y, z);

  f = [m+2*sqrt(s2); flipdim(m-2*sqrt(s2),1)]; 
  fill([z; flipdim(z,1)], f, [7 7 7]/8)
  hold on; plot(z, mu); plot(x, y, '+')

The gp function is called with a struct of hyperparameters hyp, and inference method, in this case @infExact for exact inference and the mean, covariance and likelihood functions, as well as the inputs and outputs of the training data. With no test inputs, gp returns the negative log probability of the training data, in this example nlml=11.97.

To compute the predictions at test locations we add the test inputs z as a final argument, and gp returns the mean m variance s2 at the test location. The program is using algorithm 2.1 from the GPML book. Plotting the mean function plus/minus two standard deviations (corresponding to a 95% confidence interval):

f1.gif

Typically, we would not a priori know the values of the hyperparameters hyp, let alone the form of the mean, covariance or likelihood functions. So, let's pretend we didn't know any of this. We assume a particular structure and learn suitable hyperparameters:

  covfunc = @covSEiso; hyp2.cov = [0; 0]; hyp2.lik = log(0.1);

  hyp2 = minimize(hyp2, @gp, -100, @infExact, [], covfunc, likfunc, x, y);
  exp(hyp2.lik)
  nlml2 = gp(hyp2, @infExact, [], covfunc, likfunc, x, y)

  [m s2] = gp(hyp2, @infExact, [], covfunc, likfunc, x, y, z);
  f = [m+2*sqrt(s2); flipdim(m-2*sqrt(s2),1)];
  fill([z; flipdim(z,1)], f, [7 7 7]/8)
  hold on; plot(z, m); plot(x, y, '+')

First, we guess that a squared exponential covariance function covSEiso may be suitable. This covariance function takes two hyperparameters: a characteristic length-scale and a signal standard deviation (magnitude). These hyperparameters are non-negative and represented by their logarithms; thus, initializing hyp2.cov to zero, correspond to unit characteristic length-scale and unit signal standard deviation. The likelihood hyperparameter in hyp2.lik is also initialized. We assume that the mean function is zero, so we simply ignore it (and when in the following we call gp, we give an empty argument for the mean function).

In the following line, we optimize over the hyperparameters, by minimizing the negative log marginal likelihood w.r.t. the hyperparameters. The third parameter in the call to minimize limits the number of function evaluations to a maximum of 100. The inferred noise standard deviation is exp(hyp2.lik)=0.15, somewhat larger than the one used to generate the data (0.1). The final negative log marginal likelihood is nlml2=14.13, showing that the joint probability (density) of the training data is about exp(14.13-11.97)=8.7 times smaller than for the setup actually generating the data. Finally, we plot the predictive distribution.

f2.gif

This plot shows clearly, that the model is indeed quite different from the generating process. This is due to the different specifications of both the mean and covariance functions. Below we'll try to do a better job, by allowing more flexibility in the specification.

Note that the confidence interval in this plot is the confidence for the distribution of the (noisy) data. If instead you want the confidence region for the underlying function, you should use the 3rd and 4th output arguments from gp as these refer to the latent process, rather than the data points.

  hyp.cov = [0; 0]; hyp.mean = [0; 0]; hyp.lik = log(0.1);
  hyp = minimize(hyp, @gp, -100, @infExact, meanfunc, covfunc, likfunc, x, y);
  [m s2] = gp(hyp, @infExact, meanfunc, covfunc, likfunc, x, y, z);
 
  f = [m+2*sqrt(s2); flipdim(m-2*sqrt(s2),1)];
  fill([z; flipdim(z,1)], f, [7 7 7]/8)
  hold on; plot(z, m); plot(x, y, '+');

Here, we have changed the specification by adding the affine mean function. All the hyperparameters are learnt by optimizing the marginal likelihood.

f3.gif

This shows that a much better fit is achieved when allowing a mean function (although the covariance function is still different from that of the generating process).

Exercise for the reader

Inference Methods
Try using Expectation Propagation instead of exact inference in the above, by exchanging @infExact with @infEP. You get exactly identical results, why?
Mean or Covariance
Try training a GP where the affine part of the function is captured by the covariance function instead of the mean function. That is, use a GP with no explicit mean function, but further additive contributions to the covariance. How would you expect the marginal likelihood to compare to the previous case?

Classification

You can either follow the example here on this page, or use the script demoClassification.

The difference between regression and classification isn't of fundamental nature. We can use a Gaussian process latent function in essentially the same way, it is just that the Gaussian likelihood function often used for regression is inappropriate for classification. And since exact inference is only possible for Gaussian likelihood, we also need an alternative, approximate, inference method.

Here, we will demonstrate binary classification, using two partially overlapping Gaussian sources of data in two dimensions. First we generate the data:

  clear all, close all
 
  n1 = 80; n2 = 40;                   % number of data points from each class
  S1 = eye(2); S2 = [1 0.95; 0.95 1];           % the two covariance matrices
  m1 = [0.75; 0]; m2 = [-0.75; 0];                            % the two means
 
  x1 = bsxfun(@plus, chol(S1)'*gpml_randn(0.2, 2, n1), m1);
  x2 = bsxfun(@plus, chol(S2)'*gpml_randn(0.3, 2, n2), m2);
 
  x = [x1 x2]'; y = [-ones(1,n1) ones(1,n2)]';
  plot(x1(1,:), x1(2,:), 'b+'); hold on;
  plot(x2(1,:), x2(2,:), 'r+');

120 data points are generated from two Gaussians with different means and covariances. One Gaussian is isotropic and contains 2/3 of the data (blue), the other is highly correlated and contains 1/3 of the points (red). Note, that the labels for the targets are ±1 (and not 0/1).

In the plot, we superimpose the data points with the posterior equi-probability contour lines for the probability of class two given complete information about the generating mechanism

  [t1 t2] = meshgrid(-4:0.1:4,-4:0.1:4);
  t = [t1(:) t2(:)]; n = length(t);                 % these are the test inputs
  tmm = bsxfun(@minus, t, m1');
  p1 = n1*exp(-sum(tmm*inv(S1).*tmm/2,2))/sqrt(det(S1));
  tmm = bsxfun(@minus, t, m2');
  p2 = n2*exp(-sum(tmm*inv(S2).*tmm/2,2))/sqrt(det(S2));
  contour(t1, t2, reshape(p2./(p1+p2), size(t1)), [0.1:0.1:0.9]);
f5.gif

We specify a Gaussian process model as follows: a constant mean function, with initial parameter set to 0, a squared exponential with automatic relevance determination (ARD) covariance function covSEard. This covariance function has one characteristic length-scale parameter for each dimension of the input space, and a signal magnitude parameter, for a total of 3 parameters (as the input dimension is D=2). ARD with separate length-scales for each input dimension is a very powerful tool to learn which inputs are important for predictions: if length-scales are short, inputs are very important, and when they grow very long (compared to the spread of the data), the corresponding inputs will be largely ignored. Both length-scales and the signal magnitude are initialized to 1 (and represented in the log space). Finally, the likelihood function likErf has the shape of the error-function (or cumulative Gaussian), which doesn't take any hyperparameters (so hyp.lik does not exist).

  meanfunc = @meanConst; hyp.mean = 0;
  covfunc = @covSEard; ell = 1.0; sf = 1.0; hyp.cov = log([ell ell sf]);
  likfunc = @likErf;

  hyp = minimize(hyp, @gp, -40, @infEP, meanfunc, covfunc, likfunc, x, y);
  [a b c d lp] = gp(hyp, @infEP, meanfunc, covfunc, likfunc, x, y, t, ones(n, 1));

  plot(x1(1,:), x1(2,:), 'b+'); hold on; plot(x2(1,:), x2(2,:), 'r+')
  contour(t1, t2, reshape(exp(lp), size(t1)), [0.1:0.1:0.9]);

We train the hyperparameters using minimize, to minimize the negative log marginal likelihood. We allow for 40 function evaluations, and specify that inference should be done with the Expectation Propagation (EP) inference method @infEP, and pass the usual parameters. Training is done using algorithm 3.5 and 5.2 from the gpml book. When computing test probabilities, we call gp with additional test inputs, and as the last argument a vector of targets for which the log probabilities lp should be computed. The fist four output arguments of the function are mean and variance for the targets and corresponding latent variables respectively. The test set predictions are computed using algorithm 3.6 from the gpml book. The contour plot for the predictive distribution is shown below. Note, that the predictive probability is fairly close to the probabilities of the generating process in regions of high data density. Note also, that as you move away from the data, the probability approaches 1/3, the overall class probability.

f6.gif

Examining the two ARD characteristic length-scale parameters after learning, you will find that they are fairly similar, reflecting the fact that for this data set, both inputs important.

Exercise for the reader

Inference Methods
Use the Laplace Approximation for inference @infLaplace, and compare the approximate marginal likelihood for the two methods. Which approximation is best?
Covariance Function
Try using the squared exponential with isotropic distance measure covSEiso instead of ARD distance measure covSEard. Which is best?

Acknowledgements

Innumerable colleagues have helped to improve this software. Some of these are: John Cunningham, Máté Lengyel, Joris Mooij, Iain Murray, Ed Snelson and Chris Williams.

Last modified: July 23rd 2010 gpml-matlab-v3.0-2010-07-23/doc/usageRegression.m000644 000765 000024 00000005201 11422307416 021137 0ustar00carlstaff000000 000000 % demonstrate usage of regression % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch 2010-07-21. clear all, close all %% SAY WHICH CODE WE WISH TO EXERCISE id = [1,1]; % use Exact % id = [1,2; 3,2; 4,2]; % compare Laplace id = [1,3; 2,3; 3,3]; % compare EP % id = [1,4; 2,4; 3,4; 4,4]; % compare VB seed = 197; randn('seed',seed), rand('seed',seed) ntr = 50; nte = 1e4; % number of training and test points xtr = 10*sort(rand(ntr,1)); % sample dataset f = @(x) sin(x)+sqrt(x); % "true" underlying function sn = 0.2; ytr = f(xtr) + randn(ntr,1)*sn; % add Gaussian noise i = randperm(ntr); nout = 3; % add outliers ytr(i(1:nout)) = 5; xte = linspace(0,10,1e4)'; % support, we test our function on cov = {@covSEiso}; sf = 1; ell = 0.4; % setup the GP hyp0.cov = log([ell;sf]); mean = {@meanSum,{@meanLinear,@meanConst}}; a = 1/5; b = 1; % m(x) = a*x+b hyp0.mean = [a;b]; lik_list = {'likGauss','likLaplace','likSech2','likT'}; % possible likelihoods inf_list = {'infExact','infLaplace','infEP','infVB'}; % allowable inference algs Ncg = 50; % number of conjugate gradient steps sdscale = 0.5; % how many sd wide should the error bars become? col = {'k',[.8,0,0],[0,.5,0],'b',[0,.75,.75],[.7,0,.5]}; % colors ymu{1} = f(xte); ys2{1} = sn^2; nlZ(1) = -Inf; for i=1:size(id,1) lik = lik_list{id(i,1)}; % setup the likelihood if strcmp(lik,'likT') nu = 4; hyp0.lik = log([nu-1;sqrt((nu-2)/nu)*sn]); else hyp0.lik = log(sn); end inf = inf_list{id(i,2)}; fprintf('OPT: %s/%s\n',lik_list{id(i,1)},inf_list{id(i,2)}) if Ncg==0 hyp = hyp0; else hyp = minimize(hyp0,'gp', -Ncg, inf, mean, cov, lik, xtr, ytr); % opt hypers end [ymu{i+1}, ys2{i+1}] = gp(hyp, inf, mean, cov, lik, xtr, ytr, xte); % predict [nlZ(i+1)] = gp(hyp, inf, mean, cov, lik, xtr, ytr); end figure, hold on for i=1:size(id,1)+1 plot(xte,ymu{i},'Color',col{i},'LineWidth',2) if i==1 leg = {'function'}; else leg{end+1} = sprintf('%s/%s -lZ=%1.2f',... lik_list{id(i-1,1)},inf_list{id(i-1,2)},nlZ(i)); end end for i=1:size(id,1)+1 ysd = sdscale*sqrt(ys2{i}); fill([xte;flipud(xte)],[ymu{i}+ysd;flipud(ymu{i}-ysd)],... col{i},'EdgeColor',col{i},'FaceAlpha',0.1,'EdgeAlpha',0.3); end for i=1:size(id,1)+1, plot(xte,ymu{i},'Color',col{i},'LineWidth',2), end plot(xtr,ytr,'k+'), plot(xtr,ytr,'ko'), legend(leg) gpml-matlab-v3.0-2010-07-23/doc/gpml_randn.m000644 000765 000024 00000002125 11422240326 020111 0ustar00carlstaff000000 000000 function x = gpml_randn(seed,varargin) % Generate pseudo-random numbers in a quick and dirty way. % The function makes sure, we obtain the same random numbers using Octave and % Matlab for the demo scripts. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch 2010-07-07. if nargin==2 sz = varargin{1}; else sz = zeros(1,nargin-1); for i=1:nargin-1 sz(i) = varargin{i}; end end n = prod(sz); N = ceil(n/2)*2; % minimal uniform random number generator for uniform deviates from [0,1] % by Park and Miller a = 7^5; m = 2^31-1; % using Schrage's algorithm q = fix(m/a); r = mod(m,a); % m = a*q+r u = zeros(N+1,1); u(1) = fix(seed*2^31); for i=2:N+1 % Schrage's algorithm for mod(a*u(i),m) u(i) = a*mod(u(i-1),q) - r*fix(u(i-1)/q); if u(i)<0, u(i) = u(i)+m; end end u = u(2:N+1)/2^31; % Box-Muller transform: Numerical Recipies, 2nd Edition, $7.2.8 % http://en.wikipedia.org/wiki/Box-Muller_transform w = sqrt(- 2*log(u(1:N/2))); % split into two groups x = [w.*cos(2*pi*u(N/2+1:N)); w.*sin(2*pi*u(N/2+1:N))]; x = reshape(x(1:n),sz); gpml-matlab-v3.0-2010-07-23/doc/usageCov.m000644 000765 000024 00000004062 11422313220 017540 0ustar00carlstaff000000 000000 % demonstrate usage of covariance functions % % See also covFunctions.m. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23 clear all, close all n = 5; D = 3; x = randn(n,D); xs = randn(3,D); % create a data set % set up simple covariance functions cn = {'covNoise'}; sn = .1; hypn = log(sn); % one hyperparameter cc = {@covConst}; sf = 2; hypc = log(sf); % function handles OK cl = {@covLIN}; hypl = []; % linear is parameter-free cla = {'covLINard'}; L = rand(D,1); hypla = log(L); % linear (ARD) clo = {@covLINone}; ell = .9; hyplo = log(ell); % linear with bias cp = {@covPoly,3}; c = 2; hypp = log([c;sf]); % third order poly cga = {@covSEard}; hypga = log([L;sf]); % Gaussian with ARD cgi = {'covSEiso'}; hypgi = log([ell;sf]); % isotropic Gaussian cgu = {'covSEisoU'}; hypgu = log(ell); % isotropic Gauss no scale cra = {'covRQard'}; al = 2; hypra = log([L;sf;al]); % ration. quad. cri = {@covRQiso}; hypri = log([ell;sf;al]); % isotropic cm = {'covMaterniso',3}; hypm = log([ell;sf]); % Matern class q=3 cnn = {'covNNone'}; hypnn = log([L;sf]); % neural network cpe = {'covPeriodic'}; om = 2; hyppe = log([ell;om;sf]); % periodic ccc = {'covPPiso',2}; hypcc = hypm; % compact support poly degree 2 % set up composite covariance functions csc = {'covScale',{cgu}}; hypsc = [log(3); hypgu]; % scale by 9 csu = {'covSum',{cn,cc,cl}}; hypsu = [hypn; hypc; hypl]; % sum cpr = {@covProd,{cc,ccc}}; hyppr = [hypc; hypcc]; % product mask = [0,1,0]; % binary mask excluding all but the 2nd component cma = {'covMask',{mask,cgi{:}}}; hypma = hypgi; % additive based on SEiso using unary and pairwise interactions cad = {'covADD',{[1,2],'covSEiso'}}; % 0) specify covariance function cov = cma; hyp = hypma; % 1) query the number of parameters feval(cov{:}) % 2) evaluate the function on x feval(cov{:},hyp,x) % 3) evaluate the function on x and xs to get cross-terms [kss,Ks] = feval(cov{:},hyp,x,xs) % 4) compute the derivatives w.r.t. to hyperparameter i i = 1; feval(cov{:},hyp,x,i) gpml-matlab-v3.0-2010-07-23/doc/README000644 000765 000024 00000021604 11422313200 016465 0ustar00carlstaff000000 000000 GAUSSIAN PROCESS REGRESSION AND CLASSIFICATION Toolbox version 3.0 for GNU Octave 3.2.x and Matlab 7.x Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23. 0) HOW TO READ ============== If you want to get started without further delay, then read section 1) below and jump right to the examples in doc/index.html. 1) ABOUT THESE PROGRAMS ======================= This collection of matlab programs implement and demonstrates some of the algorithms described in a) the book by Rasmussen and Williams: "Gaussian Processes for Machine Learning", the MIT Press 2006 and in b) the article by Nickisch and Rasmussen: "Approximations for Binary Gaussian Process Classification", JMLR 2010. There are 6 code subdirectories: cov, doc, inf, lik, mean and util. cov/ contains covariance functions cov*.m => see covFunctions.m doc/ contains an index.html file providing documentation. This information is also available from http://www.GaussianProcess.org/gpml/code. Usage of mean, cov, classification and regression is demonstrated in usage*.m. Further details can be found in the developer documentation manual.pdf. inf/ contains the inference algorithms inf*.m => see infMethods.m lik/ contains the likelihood functions lik*.m => see likFunctions.m mean/ contains the mean functions mean*.m => see meanFunctions.m util/ contains optimisation routines, backward compatibility programs and small auxiliary scripts Before running the demos or any other part of the code, you should execute startup.m to add all necessary directories to your path. In Octave, you can also use the .octaverc file. 2) ABOUT MEX FILES ================== Some of the programs make use of the MEX facility in matlab for more efficient implementation. However, if you don't know about how to compile MEX files, you do not need to worry about this -- the code runs anyway. If you do compile the MEX files, this is automatically detected, and the program will run more efficiently. If you do not have a compiler installed on your system, you might want to consult [2]. Two components of the toolbox can be accelerated by MEX. First, there are some auxiliary functions (currently sq_dist.m and solve_chol.m) having a MEX equivalent. These can be compiled by either executing util/make.m from the Matlab/Octave command line. In addition to that, we provide a Makefile for Matlab in util/ that you can run by simply typing make in your shell after having supplied your Matlab path and your operating system. We recommend the make.m script because it works both under Matlab and Octave. Second, in order to use the L-BFGS minimiser for hyperparameter optimisation as an equivalent to minimize.m, you have to compile Peter Carbonetto's "Matlab interface for L-BFGS-B" [1]. The challenge here is the Fortran 77 code. We provide a Makefile suitable for Linux 32/64 bit and Mac whenever you have g77 properly installed. Compilation is done by editing util/lbfgsb/Makefile. In any case, you need to provide $MATLAB_HOME which can be found by the commands 'locate matlab' or 'find / -name "matlab"'. You can choose between two compilation modes: a) using the mex utility by Matlab [default] provide $MEX, then type 'make mex' b) without mex utility by Matlab provide $MEX_SUFFIX and $MATLAB_LIB, then type 'make nomex' In Ubuntu 10.04 LTS, the libg2c library needed for both a) and b) is not included per default. If 'ls /usr/lib/libg2c.*' does not list anything this is the case on your machine. You then whant to install the packages gcc-3.4-base and libg2c0 e.g. from http://packages.ubuntu.com/jaunty/gcc-3.4-base and http://packages.ubuntu.com/jaunty/libg2c0. After installation, you have to create a symbolic link by 'cd /usr/lib' and 'ln -s libg2c.so.0 libg2c.so'. [1] http://www.cs.ubc.ca/~pcarbo/lbfgsb-for-matlab.html [2] http://www.mathworks.com/support/compilers/R2010a 3) CURRENT VERSION ================== The current version of the programs is 3.0. Previous versions of the code are avaiable at http://www.gaussianprocess.org/gpml/code/oldcode.html. 4) DIFFERENCES TO PREVIOUS VERSIONS =================================== NEW in version 3.0, 2010-07-23 ------------------------------ A major code reorganisation effort did take place in the current release. First, classification and regression are now done by a single file gp.m which is completely generic in the likelihood. The previous regression program gpr.m corresponds to gp.m with Gaussian likelihood. Several other likelihoods for robust regresssion were added. Further, the code now supports mean functions with a similar specification mechanism as already used by the covariance functions. Previous implementations correspond to using meanZero.m. We merged the covariance functions covMatern3iso.m and covMatern5.iso into a single covMaterniso.m and added a covariance function for additive functions as well as the possibility to use only certain components of the data. Finally, we included covPPiso.m a piecewise polynomial covariance function with compact support. New likelihood functions in lik/: likGauss.m, likLaplace.m, likLogistic.m and likT.m New mean functions in mean/: meanConst.m, meanLinear.m, meanPow.m, meanProd.m, meanSum.m and meanZero.m New covariance functions in cov/: covADD.m, covMask.m, covPPiso.m The gprSRPP.m function which previously provided "Subset of Regressors" and the "Projected Process" approximation has now been removed. NEW in version 2.1, 2007-07-25 ------------------------------ covConst.m: fixed a bug which caused an error in the derivative of the log marginal likelihood for certain combinations of covariance functions and approximation methods. (Thanks to Antonio Eleuteri for reporting the problem) gauher.m: added the function "gauher.m" which was mistakenly missing from the previous release. This caused an error for certain combinations of approximation method and likelihood function. logistic.m: modified the approximation of moments calculation to use a mixture of cumulative Gaussian, rather than Gauss-Hermite quadrature, as the former turns out to be more accurate. NEW in version 2.0, 2007-06-25 ------------------------------ Some code restructuring has taken place for the classification code to make it more modular, to facilitate addition of new likelihood functions and approximations methods. Now, all classification is done using the binaryGP function, which (among other things) takes an approximation method and a likelihood function as an arguments. Thus, binaryGP replaces both binaryEPGP and binaryLapaceGP, although wrapper functions are still provided for backward compatibility. This gives added flexibility: now EP can also be used wth the logistic likelihood function (implemented using Gauss-Hermite quadrature). approxEP.m: New file, containing the Expectation Propagation approximation method, which was previously contained in binaryEPGP.m approxLA.m: New file, containing Laplaces approximation method, which was previously contained in binaryLaplace.m approximations.m: New file, help for the approximation methods. binaryEPGP.m: This file has been replaced by a wrapper (for backward compatibility) which calls the more general binaryGP function. binaryGP.m: New general function to do binary classification. binaryLaplaceGP.m: This file has been replaced by a wrapper (for backward compatibility) which calls the more general binaryGP function. covMatern3iso.m, covMatern5iso.m, covNNone.m, covRQard.m, covRQiso.m, cosSEard, covSEiso: now check more carefully, that persistent variables have the correct sizes, and some variable names have been modified. cumGauss.m: New file, containing code for the cumulative Gaussian likelihood function likelihoods.m: New file, help for likelihood functions logistic.m: New file, logistic likelihood NEW in version 1.3, 2006-09-08 ------------------------------ covRQard.m: bugfix: replaced x with x' and z with z' in line 36 covRQiso.m: bugfix: replaced x with x' and z with z' in line 28 minimize.m: correction: replaced "error()" with "error('')", and made a few cosmetic changes binaryEPGP.m: added the line "lml = -n*log(2);" in line 77. This change should be largely inconsequential, but occationally may save things when the covariance matrix is exceptionally badly conditioned. NEW in version 1.2, 2006-05-10 ------------------------------ added the "erfint" function to "binaryLaplaceGP.m". The erfint function was missing by mistake, preventing the use of the "logistic" likelihood. NEW in version 1.1, 2006-04-12 ------------------------------ added files: "covProd.m" and "covPeriodic.m" changes: "covSEiso.m" was changed slightly to avoid the use of persistent variables NEW in version 1.0, 2006-03-29 ------------------------------ initial version shipped with the book gpml-matlab-v3.0-2010-07-23/doc/checkmark.png000644 000765 000024 00000000306 11422240326 020247 0ustar00carlstaff000000 000000 ‰PNG  IHDR§óa*PLTE³³³¨¨¨œœœ„„„xxxlll```TTTHHH<<<$$$ ÍÍtRNS@æØfJIDAT™c````d€6 N`€8,Åà~€Àe€ˆ‹ RS!K”Áe0O€Ðâ åPÎë`ʳÁ&Õ]¹ ‡NIEND®B`‚gpml-matlab-v3.0-2010-07-23/doc/demoRegression.m000644 000765 000024 00000007617 11422240326 020770 0ustar00carlstaff000000 000000 disp(' '); disp('clear all, close all'); clear all, close all disp(' '); disp('meanfunc = {@meanSum, {@meanLinear, @meanConst}}; hyp.mean = [0.5; 1];'); meanfunc = {@meanSum, {@meanLinear, @meanConst}}; hyp.mean = [0.5; 1]; disp('covfunc = {@covMaterniso, 3}; ell = 1/4; sf = 1; hyp.cov = log([ell; sf]);'); covfunc = {@covMaterniso, 3}; ell = 1/4; sf = 1; hyp.cov = log([ell; sf]); disp('likfunc = @likGauss; sn = 0.1; hyp.lik = log(sn);'); likfunc = @likGauss; sn = 0.1; hyp.lik = log(sn); disp(' '); disp('n = 20;'); n = 20; disp('x = gpml_randn(0.3, n, 1);'); x = gpml_randn(0.3, n, 1); disp('K = feval(covfunc{:}, hyp.cov, x);'); K = feval(covfunc{:}, hyp.cov, x); disp('mu = feval(meanfunc{:}, hyp.mean, x);'); mu = feval(meanfunc{:}, hyp.mean, x); disp('y = chol(K)''*gpml_randn(0.15, n, 1) + mu + exp(hyp.lik)*gpml_randn(0.2, n, 1);'); y = chol(K)'*gpml_randn(0.15, n, 1) + mu + exp(hyp.lik)*gpml_randn(0.2, n, 1); figure(1) set(gca, 'FontSize', 24) disp(' '); disp('plot(x, y, ''+'')'); plot(x, y, '+', 'MarkerSize', 12) axis([-1.9 1.9 -0.9 3.9]) grid on xlabel('input, x') ylabel('output, y') %print -depsc f0.eps disp(' '); disp('Hit any key to continue...'); pause disp(' '); disp('nlml = gp(hyp, @infExact, meanfunc, covfunc, likfunc, x, y)'); nlml = gp(hyp, @infExact, meanfunc, covfunc, likfunc, x, y) disp(' ') disp('z = linspace(-1.9, 1.9, 101)'';'); z = linspace(-1.9, 1.9, 101)'; disp('[m s2] = gp(hyp, @infExact, meanfunc, covfunc, likfunc, x, y, z);'); [m s2] = gp(hyp, @infExact, meanfunc, covfunc, likfunc, x, y, z); figure(2) set(gca, 'FontSize', 24); disp(' '); disp('f = [m+2*sqrt(s2); flipdim(m-2*sqrt(s2),1)];'); f = [m+2*sqrt(s2); flipdim(m-2*sqrt(s2),1)]; disp('fill([z; flipdim(z,1)], f, [7 7 7]/8);'); fill([z; flipdim(z,1)], f, [7 7 7]/8); disp('hold on; plot(z, m); plot(x, y, ''+'')') hold on; plot(z, m, 'LineWidth', 2); plot(x, y, '+', 'MarkerSize', 12) axis([-1.9 1.9 -0.9 3.9]) grid on xlabel('input, x') ylabel('output, y') %print -depsc f1.eps disp(' '); disp('Hit any key to continue...'); pause disp(' '); disp('covfunc = @covSEiso; hyp2.cov = [0; 0]; hyp2.lik = log(0.1);'); covfunc = @covSEiso; hyp2.cov = [0; 0]; hyp2.lik = log(0.1); disp('hyp2 = minimize(hyp2, @gp, -100, @infExact, [], covfunc, likfunc, x, y)'); hyp2 = minimize(hyp2, @gp, -100, @infExact, [], covfunc, likfunc, x, y); disp(' '); disp('exp(hyp2.lik)'); exp(hyp2.lik) disp('nlml2 = gp(hyp2, @infExact, [], covfunc, likfunc, x, y)'); nlml2 = gp(hyp2, @infExact, [], covfunc, likfunc, x, y) disp('[m s2] = gp(hyp2, @infExact, [], covfunc, likfunc, x, y, z);'); [m s2] = gp(hyp2, @infExact, [], covfunc, likfunc, x, y, z); disp(' ') figure(3) set(gca, 'FontSize', 24) f = [m+2*sqrt(s2); flipdim(m-2*sqrt(s2),1)]; disp('fill([z; flipdim(z,1)], f, [7 7 7]/8)'); fill([z; flipdim(z,1)], f, [7 7 7]/8) disp('hold on; plot(z, m); plot(x, y, ''+'')'); hold on; plot(z, m, 'LineWidth', 2); plot(x, y, '+', 'MarkerSize', 12) grid on xlabel('input, x') ylabel('output, y') axis([-1.9 1.9 -0.9 3.9]) %print -depsc f2.eps disp(' '); disp('Hit any key to continue...'); pause disp(' '); disp('hyp.cov = [0; 0]; hyp.mean = [0; 0]; hyp.lik = log(0.1);'); hyp.cov = [0; 0]; hyp.mean = [0; 0]; hyp.lik = log(0.1); disp('hyp = minimize(hyp, @gp, -100, @infExact, meanfunc, covfunc, likfunc, x, y);'); hyp = minimize(hyp, @gp, -100, @infExact, meanfunc, covfunc, likfunc, x, y); disp('[m s2] = gp(hyp, @infExact, meanfunc, covfunc, likfunc, x, y, z);'); [m s2] = gp(hyp, @infExact, meanfunc, covfunc, likfunc, x, y, z); figure(4) set(gca, 'FontSize', 24) disp(' '); disp('f = [m+2*sqrt(s2); flipdim(m-2*sqrt(s2),1)];'); f = [m+2*sqrt(s2); flipdim(m-2*sqrt(s2),1)]; disp('fill([z; flipdim(z,1)], f, [7 7 7]/8)'); fill([z; flipdim(z,1)], f, [7 7 7]/8) disp('hold on; plot(z, m); plot(x, y, ''+'');'); hold on; plot(z, m, 'LineWidth', 2); plot(x, y, '+', 'MarkerSize', 12) grid on xlabel('input, x') ylabel('output, y') axis([-1.9 1.9 -0.9 3.9]) %print -depsc f3.eps disp(' '); gpml-matlab-v3.0-2010-07-23/doc/usageClassification.m000644 000765 000024 00000005057 11422307416 021763 0ustar00carlstaff000000 000000 % demonstrate usage of classification % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch 2010-06-21. clear all, close all %% SAY WHICH CODE WE WISH TO EXERCISE % id = [1,1]; % use Gauss/Exact id = [1,2; 2,2; 3,2]; % compare Laplace % id = [1,3; 2,3; 3,3]; % compare EP % id = [1,4; 3,4]; % compare VB seed = 943; randn('seed',seed), rand('seed',seed) ntr = 50; nte = 1e4; % number of training and test points xtr = 10*sort(rand(ntr,1)); % sample dataset p = @(x) 1./(1+exp(-5*sin(x))); % "true" underlying probability ytr = 2*(p(xtr)>rand(ntr,1))-1; % draw labels +1/-1 i = randperm(ntr); nout = 3; % add outliers ytr(i(1:nout)) = -ytr(i(1:nout)); xte = linspace(0,10,1e4)'; % support, we test our function on cov = {@covSEiso}; sf = 1; ell = 0.7; % setup the GP hyp0.cov = log([ell;sf]); mean = {@meanZero}; % m(x) = 0 hyp0.mean = []; lik_list = {'likGauss','likErf','likLogistic'}; % allowable likelihoods inf_list = {'infExact','infLaplace','infEP','infVB'}; % possible inference algs Ncg = 50; % number of conjugate gradient steps sdscale = 0.5; % how many sd wide should the error bars become? col = {'k',[.8,0,0],[0,.5,0],'b',[0,.75,.75],[.7,0,.5]}; % colors ymu{1} = 2*p(xte)-1; ys2{1} = 0; for i=1:size(id,1) lik = lik_list{id(i,1)}; % setup the likelihood if strcmp(lik,'likGauss') sn = .2; hyp0.lik = log(sn); else hyp0.lik = []; end inf = inf_list{id(i,2)}; fprintf('OPT: %s/%s\n',lik_list{id(i,1)},inf_list{id(i,2)}) hyp = minimize(hyp0,'gp', -Ncg, inf, mean, cov, lik, xtr, ytr); % opt hypers [ymu{i+1}, ys2{i+1}] = gp(hyp, inf, mean, cov, lik, xtr, ytr, xte); % predict [nlZ(i+1)] = gp(hyp, inf, mean, cov, lik, xtr, ytr); end figure, hold on for i=1:size(id,1)+1 plot(xte,ymu{i},'Color',col{i},'LineWidth',2) if i==1 leg = {'function'}; else leg{end+1} = sprintf('%s/%s -lZ=%1.2f',... lik_list{id(i-1,1)},inf_list{id(i-1,2)},nlZ(i)); end end for i=1:size(id,1)+1 ysd = sdscale*sqrt(ys2{i}); fill([xte;flipud(xte)],[ymu{i}+ysd;flipud(ymu{i}-ysd)],... col{i},'EdgeColor',col{i},'FaceAlpha',0.1,'EdgeAlpha',0.3); end for i=1:size(id,1)+1, plot(xte,ymu{i},'Color',col{i},'LineWidth',2), end plot(xtr,ytr,'k+'), plot(xtr,ytr,'ko'), legend(leg) gpml-matlab-v3.0-2010-07-23/doc/style.css000644 000765 000024 00000000115 11422240326 017461 0ustar00carlstaff000000 000000 body {font-family: sans-serif; font-size: 16px} table {font-size: inherit;} gpml-matlab-v3.0-2010-07-23/doc/usageMean.m000644 000765 000024 00000002406 11422313220 017671 0ustar00carlstaff000000 000000 % demonstrate usage of mean functions % % See also meanFunctions.m. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23 clear all, close all n = 5; D = 2; x = randn(n,D); % create a random data set % set up simple mean functions m0 = {'meanZero'}; hyp0 = []; % no hyperparameters are needed m1 = {'meanOne'}; hyp1 = []; % no hyperparameters are needed mc = {@meanConst}; hypc = 2; % also function handles are possible ml = {@meanLinear}; hypl = [2;3]; % m(x) = 2*x1 + 3*x2 % set up composite mean functions msc = {'meanScale',{m1}}; hypsc = [3; hyp1]; % scale by 3 msu = {'meanSum',{m0,mc,ml}}; hypsu = [hyp0; hypc; hypl]; % sum mpr = {@meanProd,{mc,ml}}; hyppr = [hypc; hypl]; % product mpo = {'meanPow',{3,msu}}; hyppo = hypsu; % third power mask = [0,1,0]; % binary mask excluding all but the 2nd component mma = {'meanMask',{mask,mpo{:}}}; hypma = hyppo; % 0) specify mean function % mean = m0; hyp = hyp0; % mean = msu; hyp = hypsu; % mean = mpr; hyp = hyppr; mean = mpo; hyp = hyppo; % 1) query the number of parameters feval(mean{:}) % 2) evaluate the function on x feval(mean{:},hyp,x) % 3) compute the derivatives w.r.t. to hyperparameter i i = 2; feval(mean{:},hyp,x,i) gpml-matlab-v3.0-2010-07-23/doc/manual.pdf000644 000765 000024 00002127216 11422313230 017570 0ustar00carlstaff000000 000000 %PDF-1.4 %ÐÔÅØ 5 0 obj << /S /GoTo /D (section.1) >> endobj 8 0 obj (Gaussian Process Training and Prediction) endobj 9 0 obj << /S /GoTo /D (section.2) >> endobj 12 0 obj (The gp Function) endobj 13 0 obj << /S /GoTo /D (section.3) >> endobj 16 0 obj (Inference Methods) endobj 17 0 obj << /S /GoTo /D (subsection.3.1) >> endobj 20 0 obj (Exact Inference ) endobj 21 0 obj << /S /GoTo /D (subsection.3.2) >> endobj 24 0 obj (Laplace's Approximation) endobj 25 0 obj << /S /GoTo /D (subsection.3.3) >> endobj 28 0 obj (Expectation Propagation) endobj 29 0 obj << /S /GoTo /D (subsection.3.4) >> endobj 32 0 obj (Variational Bayes) endobj 33 0 obj << /S /GoTo /D (section.4) >> endobj 36 0 obj (Likelihood Functions) endobj 37 0 obj << /S /GoTo /D (subsection.4.1) >> endobj 40 0 obj (Prediction) endobj 41 0 obj << /S /GoTo /D (subsection.4.2) >> endobj 44 0 obj (Interface) endobj 45 0 obj << /S /GoTo /D (subsection.4.3) >> endobj 48 0 obj (Implemented Likelihood Functions) endobj 49 0 obj << /S /GoTo /D (subsection.4.4) >> endobj 52 0 obj (Compatibility Between Likelihoods and Inference Methods) endobj 53 0 obj << /S /GoTo /D (subsection.4.5) >> endobj 56 0 obj (Gaussian Likelihood) endobj 57 0 obj << /S /GoTo /D (subsubsection.4.5.1) >> endobj 60 0 obj (Exact Inference) endobj 61 0 obj << /S /GoTo /D (subsubsection.4.5.2) >> endobj 64 0 obj (Laplace's Approximation) endobj 65 0 obj << /S /GoTo /D (subsubsection.4.5.3) >> endobj 68 0 obj (Expectation Propagation) endobj 69 0 obj << /S /GoTo /D (subsubsection.4.5.4) >> endobj 72 0 obj (Variational Bayes) endobj 73 0 obj << /S /GoTo /D (subsection.4.6) >> endobj 76 0 obj (Laplace Likelihood) endobj 77 0 obj << /S /GoTo /D (subsubsection.4.6.1) >> endobj 80 0 obj (Laplace's Approximation) endobj 81 0 obj << /S /GoTo /D (subsubsection.4.6.2) >> endobj 84 0 obj (Expectation Propagation) endobj 85 0 obj << /S /GoTo /D (subsubsection.4.6.3) >> endobj 88 0 obj (Variational Bayes) endobj 89 0 obj << /S /GoTo /D (subsection.4.7) >> endobj 92 0 obj (Student's t Likelihood) endobj 93 0 obj << /S /GoTo /D (subsubsection.4.7.1) >> endobj 96 0 obj (Laplace's Approximation) endobj 97 0 obj << /S /GoTo /D (subsection.4.8) >> endobj 100 0 obj (Cumulative Logistic Likelihood) endobj 101 0 obj << /S /GoTo /D (subsubsection.4.8.1) >> endobj 104 0 obj (Laplace's Approximation) endobj 105 0 obj << /S /GoTo /D (section.5) >> endobj 108 0 obj (Mean Functions ) endobj 109 0 obj << /S /GoTo /D (subsection.5.1) >> endobj 112 0 obj (Interface) endobj 113 0 obj << /S /GoTo /D (subsection.5.2) >> endobj 116 0 obj (Implemented Mean Functions) endobj 117 0 obj << /S /GoTo /D (subsection.5.3) >> endobj 120 0 obj (Usage of Implemented Mean Functions) endobj 121 0 obj << /S /GoTo /D (section.6) >> endobj 124 0 obj (Covariance Functions) endobj 125 0 obj << /S /GoTo /D (subsection.6.1) >> endobj 128 0 obj (Interface) endobj 129 0 obj << /S /GoTo /D (subsection.6.2) >> endobj 132 0 obj (Implemented Covariance Functions) endobj 133 0 obj << /S /GoTo /D (subsection.6.3) >> endobj 136 0 obj (Usage of Implemented Covariance Functions) endobj 137 0 obj << /S /GoTo /D [138 0 R /Fit ] >> endobj 143 0 obj << /Length 1363 /Filter /FlateDecode >> stream xÚuWKoÛ8¾÷WøTÈ@ÅR¤ž{k‹4Û"Ù Z{Øî–h›¨,"Äÿ~g8”-%ÎÅä<8š×7¤ùb»à‹ÛwüõóêÝǯY±H &’,]¬6 !+–çb‘W“¼Z¬šÅ¿Ñj§—ÿ­¾ü*gºR²4/À’Wº}¸¿#­¹E0(åYkµ¬Ê¨ïÛuÿ|QNX•e•ãD$,õXf¬*éÌ5´ËXetÓ<©¡¡ýe÷GkuGä{ZþT]§-íÿ2õocë~hQ&Œ‹ìr–ˆ`øû±=‘ª†'õ_å <­«r‘££IÛt§ ËdJæ>­­TíÞ:_LÎÇI‘²¢‚@!à ¸šôà|ôWnFÆз;"†;óÒpíµ  2¾°p¨ +ª,ªûG5@ËàhAÁ˜ „½ùO]÷]ÓŽÙUE!£1dP{2(xÏ@ؘ_áÜd˜¡Ô æ¸úpÍg{D4É$÷ÀU³-–5« ÜxxætB¢®ïØÜ<ƒgç‘ @ÅAm' œz7~ê1b¬vfÀˆÉn°ïƒ]ÆÐÛD¿ ™ð•Zó[·f×÷èRZ^RŒdÂɲt:šA Žë:´ÊÇ•ÞÙÍ(ô§˜ 3xøÌl}Ü[ðÒßHh°ßëLMÔhqêxÀsýbfŒ9JEôs‚@‘—þħ­~ÄáÔ¥sQD‹…ùë1•—4z….«ûýÁ–ˆF‡ÞÌÐ=~÷JÆA ú ­e"ZŸh…Û Ò0Ñhk¶í}iñVð VØheÃIÕ4#Þ€ò]ròfdŸÏ~½ÝA«&ØØ(3´§ÙÉG}½[Ö R¿œóSÔˆv·ëûa¤Æ¡n® ™á»˜5’”sØL›t¦GÿÚ„œx0À{è1o¤F®¢Åéz×™ZµDîUw¤}NÁMTh§"‘0yœ2þžFEø§IšåÑ· qOý1€uÐÄéz÷ö„?iG}µQ{h)5õdüÅ•R%_?+ÁŠ7Þä³Ç¢/‹x{eåÅììÊÏáBMÊñÄÑêàΘ۷nmt‡Ÿß®þØc jöÉ ýõ e'•>ë@¯µszO•‚3žäóD=©ð¾p=­[Ìn¬SƒÓ ›½Çég¯z]CòñÿF!‘·™ÌŽÜ¬ÞýÚ<›Æ endstream endobj 138 0 obj << /Type /Page /Contents 143 0 R /Resources 142 0 R /MediaBox [0 0 612 792] /Parent 151 0 R /Annots [ 139 0 R 140 0 R 141 0 R ] >> endobj 139 0 obj << /Type /Annot /Border[0 0 1]/H/I/C[0 1 1] /Rect [144.994 542.879 341.928 554.764] /Subtype/Link/A<> >> endobj 140 0 obj << /Type /Annot /Border[0 0 1]/H/I/C[0 1 1] /Rect [152.075 530.999 415.497 542.809] /Subtype/Link/A<> >> endobj 141 0 obj << /Type /Annot /Border[0 0 1]/H/I/C[0 1 1] /Rect [272.178 411.372 366.877 423.258] /Subtype/Link/A<> >> endobj 144 0 obj << /D [138 0 R /XYZ 71 791.866 null] >> endobj 145 0 obj << /D [138 0 R /XYZ 72 754.004 null] >> endobj 142 0 obj << /Font << /F57 146 0 R /F37 147 0 R /F59 148 0 R /F55 149 0 R /F61 150 0 R >> /ProcSet [ /PDF /Text ] >> endobj 187 0 obj << /Length 903 /Filter /FlateDecode >> stream xÚí™ÏoÚ0Çïý+r[8ÄóïØǵj«N­4il—j‡ IDB×ý÷s° “º¶@N(Ɖ퟿ïù=èM=è]ŸAû{><ûxŤ‡( ”co8ñBì…”H™7{÷þEš*)òÁágÓ %*ûB/Àp"MW4$¥þu´Ìó8J…ÿe‘ŽTž›‡á@bÅIœLMK”Œ]?5ŽGEœ–ïÄ©OÊ!í” 36c ”™Õ ÀúUßiÖ4ér­0t}®–‰´‹"¤\ÿGWåz|Áëó f7ÉD-T2RfwªxHǹù˜}Í~Lع„Õ¹”êcìˆf‰P‚Dп|ŠF…ùpe®Á `à•ò¥¿©±®íÙ»“ÌÀÞ´B×JÀ·Q6‹Fêà ¤¾µÐOY¶HŸâyd,‘ptÚ$eIâL5S£ÂB³Ç>‹¦¶Áðt)B"};0Rƒñû@p?ZÄ+nÑÌ <þ(mL’þÜïŒjÍÛš §VÂ¥“÷•„ßÆ¿Ô,~HS눜GÐÈ " ù7I·ûbu©Sîª3ãè€õúM÷·êTüF(‹ITúFÊØq²~-䤹•û›y6Ss*{hZ‘ ä(%v ²’~‘Î3-ç?ãY\ü±‚®ŠßJ%Ûäò­ˆxGl0.þïrÛVÍ:VÍ̪ëw€ªu„‚ô'n7PÌ@é`ª÷ uåWB¾ãFÀä …ûm,y—q®\†æØøãôÀ40â ¯-¾§Hž¬°5°Z£xÈNü*´…1¨*Þ&¯ÝП»‘‹ª_ ®ØÛpçMÞ­ ¾-´g›;·Ñ«à>°H¯‚1ÜKC£‚_‹åX_¶ªµx.‰bÖÇÔ%\´[+l{l¥4:aoµËùr¦é<:לN㼈G ÷=‚N]³Å‰ÞâêÔp{Ò–"@µåÕênÌ$mï”Ë-l2M«¢€® €˜ì(ºa¯os¹Û>ørÜžNdë îv:±i{ßGáóµŽFsv‘9p6û-¦VÓ‰Í íŲžC<Pí…«!ÜhÈEúX†§ëk­ðƒ´ÛÂ|ßÂïÅãåw¶=9ÅwŠG˾r|”µ,:Pýƒ\´Ðcˆ:šM„†¡ÔÂ!à®$\H.‡gó¾„ endstream endobj 186 0 obj << /Type /Page /Contents 187 0 R /Resources 185 0 R /MediaBox [0 0 612 792] /Parent 151 0 R /Annots [ 152 0 R 153 0 R 154 0 R 155 0 R 156 0 R 157 0 R 158 0 R 159 0 R 160 0 R 161 0 R 162 0 R 163 0 R 164 0 R 165 0 R 166 0 R 167 0 R 168 0 R 169 0 R 170 0 R 171 0 R 172 0 R 173 0 R 174 0 R 175 0 R 176 0 R 177 0 R 178 0 R 179 0 R 180 0 R 181 0 R 182 0 R 183 0 R 184 0 R ] >> endobj 152 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [71.004 714.988 284.785 727.579] /A << /S /GoTo /D (section.1) >> >> endobj 153 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [71.004 690.528 167.321 703.12] /A << /S /GoTo /D (section.2) >> >> endobj 154 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [71.004 668.395 177.232 678.662] /A << /S /GoTo /D (section.3) >> >> endobj 155 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 654.835 190.182 665.107] /A << /S /GoTo /D (subsection.3.1) >> >> endobj 156 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 638.908 233.043 651.558] /A << /S /GoTo /D (subsection.3.2) >> >> endobj 157 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 625.358 231.451 637.573] /A << /S /GoTo /D (subsection.3.3) >> >> endobj 158 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 611.809 196.563 624.46] /A << /S /GoTo /D (subsection.3.4) >> >> endobj 159 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [71.004 589.74 189.407 600.007] /A << /S /GoTo /D (section.4) >> >> endobj 160 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 576.18 162.342 586.452] /A << /S /GoTo /D (subsection.4.1) >> >> endobj 161 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 562.631 155.633 572.903] /A << /S /GoTo /D (subsection.4.2) >> >> endobj 162 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 546.703 278.709 559.354] /A << /S /GoTo /D (subsection.4.3) >> >> endobj 163 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 533.154 392.043 545.805] /A << /S /GoTo /D (subsection.4.4) >> >> endobj 164 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 521.983 211.465 532.256] /A << /S /GoTo /D (subsection.4.5) >> >> endobj 165 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [112.458 508.434 222.058 518.706] /A << /S /GoTo /D (subsubsection.4.5.1) >> >> endobj 166 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [112.458 492.507 267.952 505.157] /A << /S /GoTo /D (subsubsection.4.5.2) >> >> endobj 167 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [112.458 478.957 266.36 491.172] /A << /S /GoTo /D (subsubsection.4.5.3) >> >> endobj 168 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [112.458 465.408 231.472 478.059] /A << /S /GoTo /D (subsubsection.4.5.4) >> >> endobj 169 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 451.859 204.778 464.51] /A << /S /GoTo /D (subsection.4.6) >> >> endobj 170 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [112.458 438.31 267.952 450.96] /A << /S /GoTo /D (subsubsection.4.6.1) >> >> endobj 171 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [112.458 424.761 266.36 436.975] /A << /S /GoTo /D (subsubsection.4.6.2) >> >> endobj 172 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [112.458 411.211 231.472 423.862] /A << /S /GoTo /D (subsubsection.4.6.3) >> >> endobj 173 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 400.04 217.312 410.313] /A << /S /GoTo /D (subsection.4.7) >> >> endobj 174 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [112.458 384.113 267.952 396.764] /A << /S /GoTo /D (subsubsection.4.7.1) >> >> endobj 175 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 370.564 262.956 383.214] /A << /S /GoTo /D (subsection.4.8) >> >> endobj 176 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [112.458 357.015 267.952 369.665] /A << /S /GoTo /D (subsubsection.4.8.1) >> >> endobj 177 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [71.004 334.945 168.778 344.776] /A << /S /GoTo /D (section.5) >> >> endobj 178 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 321.385 155.633 331.658] /A << /S /GoTo /D (subsection.5.1) >> >> endobj 179 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 305.458 255.047 318.109] /A << /S /GoTo /D (subsection.5.2) >> >> endobj 180 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 291.909 298.683 304.559] /A << /S /GoTo /D (subsection.5.3) >> >> endobj 181 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [71.004 269.84 191.807 279.67] /A << /S /GoTo /D (section.6) >> >> endobj 182 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 256.279 155.633 266.552] /A << /S /GoTo /D (subsection.6.1) >> >> endobj 183 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 240.352 281.109 253.003] /A << /S /GoTo /D (subsection.6.2) >> >> endobj 184 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [87.367 226.803 324.745 239.454] /A << /S /GoTo /D (subsection.6.3) >> >> endobj 188 0 obj << /D [186 0 R /XYZ 125.795 791.866 null] >> endobj 189 0 obj << /D [186 0 R /XYZ 72 731.826 null] >> endobj 185 0 obj << /Font << /F59 148 0 R /F18 190 0 R /F57 146 0 R >> /ProcSet [ /PDF /Text ] >> endobj 199 0 obj << /Length 3186 /Filter /FlateDecode >> stream xÚ½Ërã¸ñ¾_¡[䊅€ @•ÚC^»™ì#{pe™( ¶™¥H-IÙ3ùút£¤!gR•‹ÕšF¿0ßÀøõx±çØÊ-˲-ÃO‚’÷-="c™Î‰Ô»î¶ëŽàfzìAè+ ¬r–‹`¤@3â%ðáðd ¹¥‘~kú `GÞ<ô§óe2› €÷\ñú|úÍ©ž 3Z˜µ€Ç¾%›- Æ•\è›ÄUiJ¸Ð™‡zjž,šÞ¶ýMŸêá¡!CQÛübÚæ±Çcá˜Bà7ÓH3ç|À£µ'KÕ­>ccNGcôq`îñã¨Ôv—R9®A¦fÑæ•Þ>a¡Ì¬\áçÔMKàx6‡æ=çâP;IÃ$ž¾a†9ßþä>3µ§C^Šáì© [eJ±ôÚÏmL‚I²MÒÊÕ¬u§>åB¸èçXOÍ„!«T±yâÂÉ™'}f£¹ßR‡qäl ´T¼™Á w4ÏËSN°ö‘ «Áüzip/™qäçfÆI°¨§æhÃ.Z•"Á><Rä 1…ºm¦4ebéõ÷­›6¦¡Nq¹"þÛ¸ì"ÄÂhu«Ò¦Ñó¹mõ¾5¯G @fbÆN¦²‚öÿ+Zìl˜¯¾‡ôÏç< ¦S’ÑG$ìdãƒ7 )™Œ‰vèÛÖešrN  55Æ¡ÔJr—0Рؼn‡ªXyu!aôQpç#¢p:©ØGpa¶\ h¤ˆ"[*<>²‘W$Ä{!DnÜoM8N" “xn& ¦Þ Œ1"˜–Âûºö¨@A\zÙ§j a÷rX·)BH¨’@JI2X“$Vª*P‡<òKVC«²lfÎYmM}pR°V¿>Ëh­‹<‡O%zŸ8wRJgœ0•N Ä·¦5'( FG‹Ê‹]VJ&±8ëB”^ì48]ln¨~°ŠAÔ†ßa[0 á<]xOàEÌŸIDXÎœCAºÄp!Ź 0‡ìMÛ?c·1úÞÅâO ƒÇÚš®§_Ý`}ÔF«É—1.l™«Q‚÷.‚™ÓÙG·~ðĺùEPûœ¸å4óý*&~ âj˸bŠ¯C˜\86Ž£òFæ³V(.I$Ö™'‚(TEKK’ùKÊÝÒ¥|k‚ºdB¬ÂE!%£ÊJ¨| FtiKlã\á›C( -•]dB ³A«Ø~C_D ·‹B9÷¹Ó}Åè—U¹5Õ¹ Z¥.X8š³ñ%—e­òv5XTÇÂ8†žÃÚ¡½PzƵ8Sà¸3ØhÖÃG"Ib¨‚Û%ëç‹.ÇšzÀÎô–†¡îÀüûüÁxx›e–Ø¥-íR _ש85ἊºTX™À‚5¡à…‚C`üîÊcÍïÑÜ×—ÖÅßf\Uÿ1C¿¬¿ <üS\!}F˜Õ/ìÐrYoi×[°J'P”ƒ¤WNQãdlÔ¸4yÒ¶Ê{e Fö[rV¬#ì;k|z`}SÅ’¢ÒÞÞ}˨ôÚÞuÒÞEì½pö-%ì]ÇöžÊò e)œ(ˆÆqT¥êR“6,ž0BNÀt>€pì8~áR¸š¶|'àp v+‹ŠýÇ‘_ÄÍÐ" N›Þbíeˆ|q’‹Êw·ÔÆMÏnƒüÖ·ÌÖhÒÞ3œz=aØSQƒàôš] ãÚ‘™o%e{å%\.E`¾îHùst…è.Œ_4€LÛë%š´rå¸V@óeM •Î|ÅxïpäÕ§ Æg [bß'‰å¬yd :×ۿ߃SÞbã fªÝºÄkÒ¡?!ä"ó "œ},XÛ¤à%“³gœSžQ0™‡ÞãUŠÓÍ×a Zb¨Y>ÖdUxW+—låLTÊ#ÿ˜"W25ûëk\ÉW%¿ÊÕ¿“"¯ ';¥´U±ªÊW6´æ£Táæá»tø†b!ÄF%Ós¹n£$1çÕ²Zž¯®)—ÅÏØ€OTWÄ¥œm3S€`bÜ –û5m)Å|Á2ëDUŸ§“bÀ¬“?§vÊ2ðÛjiMJ]³&ó嬩ÌÖÖ¤áo•åž$¿bQ˜Ý.ÂüÌJÎEê˜Óstú:™Ú2¨MôÊ“ËþÍl®¡Q«ÖwAØp—¥½›Æ_òHœpý‰ï!¶ÿ*«y5vx{3v¿¹Ö`¢ [Õ­(ÖûþÉXßaÿ¸œü·¢\B­ô–¥d\y㌫¡øš)yò`Eµòâ·x•õÓÝMvµ`€ÎAÕ«¥vÿWÒÀÙçŽççDú¶9¾„BµúDúÎ^-: aàµR òj±ðù›­/ ÀXîƯ…hĪÞ¢ Hd!›Û›U%¶—î2š£öó|×Ñ‚ÿo²[wÅI³ã¹Æðäþ‘Ê?LÂ:Ý^ {)–ÿR%ld§§U%Y¦Ä¢òOƒÈÑíò®Ïÿßýs”Y5Cøp8ßmÏo¹6…ÜÏÈ)éIè`”|[ÊTToA„î‰Ùñy´w-+Æêƒû÷£ù‰tyã… ge†]JiK)K9_àüåî«ÿõî½ endstream endobj 198 0 obj << /Type /Page /Contents 199 0 R /Resources 197 0 R /MediaBox [0 0 612 792] /Parent 151 0 R /Annots [ 191 0 R 192 0 R 193 0 R 194 0 R 195 0 R 196 0 R ] >> endobj 191 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [419.766 709.253 427.824 721.904] /A << /S /GoTo /D (section.2) >> >> endobj 192 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [457.335 597.134 465.393 609.435] /A << /S /GoTo /D (section.3) >> >> endobj 193 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [217.665 429.811 225.723 442.461] /A << /S /GoTo /D (section.4) >> >> endobj 194 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [507.705 380.197 515.763 392.847] /A << /S /GoTo /D (section.5) >> >> endobj 195 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [295.894 303.833 303.952 316.135] /A << /S /GoTo /D (section.6) >> >> endobj 196 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [218.184 272.311 226.242 284.961] /A << /S /GoTo /D (section.3) >> >> endobj 200 0 obj << /D [198 0 R /XYZ 71 791.866 null] >> endobj 6 0 obj << /D [198 0 R /XYZ 72 754.004 null] >> endobj 205 0 obj << /D [198 0 R /XYZ 99.944 210.638 null] >> endobj 197 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F18 190 0 R /F70 201 0 R /F74 202 0 R /F79 203 0 R /F80 204 0 R /F81 206 0 R /F83 207 0 R >> /ProcSet [ /PDF /Text ] >> endobj 219 0 obj << /Length 3794 /Filter /FlateDecode >> stream xÚí]YoãF~ϯÐË búf3A^‚Ý ²û ¼/IæA–(›ê€O´‹ýïÛ¬n’EIöPtI–2‚ŠæÑgÕWGWYï¡Çzï¿aá÷§»o¾ûY'=®"©ŒèÝMz±èÅJFLéÞݸ÷{_¼pÆXÿî1}÷ñîŸßýlcü¸æãÚ?,ü3Í"¹tõÈÏ›ÙhÍgáAW‹–ðâÁI"mto Y$™õ/Ts‹Ÿ<,Þ«¸Q^цú™IYó@ĶŸ­üïz»ÈFÃ<߆]mp2Ÿ•—šïUlVéÒŸ}žoò±?gËt´.ß,ÊŠ†U^þBKm”£|ÇM¤­ì ¸ûöª¡ïS,qŸ¤p“”}zôŽQ‘Pª™hZM]£0‘qSÍ^ª*‰T]SöúÂÝ¥âïS¯<û—ïOâJú#Ý»Œ»þð(Ñ’øÎH0÷þò¡çO>À˜hTFÑ[‘ÍäÝ@Åý;šþ ÎGp\Õ Îçá.®¦÷ûÀhÞÿ=\Œx/R¾Vú‰eÿ Šº£í/«ó8œ?TXhDQñN5®ŒïF²þÅOR¼#uáþáý?˜fïÚº HÞwô!EqË&ýo‹§uÑwc7&õ¤?…bÑIU4É»‡’¸|oïÍá½'\`72¸ñ©º!xÿOhQøg[üøWXqG˜¾c G»¡ö-Ô¸*;ÑžÂô5(@4Æ È­1ßZìÓuý€ŒllÃ`ÎÇ4_ öØáYMJê¾"öºpwóŠÖ[÷2‰#åZÝ襤ëe6ËÖÙ0Ïþ3,ú¸:Ð5S°kÕpMØ5«#alèš CGÙµIºLg£ô@§4«Ì vJ¹ÆÝ)£#Yu*Ì—>—,‚EàÉNÏÀŒÑ1€—lÑ€-°ø_é}𲪀 âØ1.qLû àôÃ{\ý»®,Á•°ý¿í¶µKQ4<ÓõNÅ)WÜœ¤8­@0ˆ6ÝqÇ5IÑF‚ä,vÇ¢h Õç$p€ù¤ wLi U…´qÒqY ²D^);Çpž†s\eK°/x¢ÁV¦Oì°–‘J*œ~Ú‘Âö‰ìÕ»/‘ Á_Èb‘ôÿ Â×KW/Ã2v\ÊÕ 7uq;1p[ wô3oÐÒ ŽûÿƒGxOʘ±¤J xj\Àù ¸q´(È=êÐ.žÃñ¦zЈÜ Ḅ#QÑ|GijÕK "¢ê .L`þl ù ŽOù[¶æuÿ<úxÙÄ€’(xÅœí²ß¡Æ6(à¡2ññF…¬„·—9âë´ ô%6e_R“\/˜P­õ¤Ñ|ºØ¬ƒÍ¹NWk¶X¦ãlôœVÈYìZU!©!T ¹rÍc¦©&ÇÏh¥Í€êÆ­å†ÕŒý® 6qmÒQFðü².PÙÎÄ)˜+>Ž²Ü©Òiùÿ¼®†õ×Ëa6{ç8~ö.æ}ñÈãv‘.Ãåpš®«ræÿ;ô?ïõ¿Óù8ÍÃðäç4Ï}‰å7(‘|˜"Û 礡•¬vzÒA¹ *ÂyVH9¸‡+í—¶¼Šð=•}¾®Ü„åòP†vQdH% Ðóâ’RixyñÜ"½Ù´º_Z<§ji/:"ZŸ:"Z¢ˆèS°­ˆÞSF[©‰¤ØäaÔ—C‹¤&LÆ䦓\ Û*CÓÖÒÂàœù÷”> zgä&¥^=CKãh]G‡<…gèTyâ(g¯ Wˆ½¬Ü›MÄ‹È)ÑÖ'R®+›$'<öŽÿ$Xp+*utQ~醮eåõüu0¼!œ_Li„n—àt2¿.wú&ýñòe®Ï½ì,O¦¦6»5k¬„4»5 çÍîÊX'Û'E¼¨7B‘ƒSäûÛ ùqh‡ÔKjü§ÃYÐሆqŠâÆ„^­©}ΛÆV¶QÛý‚tʶ€¥ð9¹²³,.D¼È.>4švðµDš‚O  ;ðÊÅÑ70Úä ¶±’­ 8>-bÉAG• aH€dÙÚÜçìÊL#º±³'#& áKÒó£¢ÐãwM<ð±uFFy1Š÷!š7¹îÞ£˜”ÅžÔ“*áÕQb'êéWk´:¾Döî±ÚWDåŠû7Úìïó! f»¦°'cgªT´TÙ¹úí¹ãËXo™&vSúæ%«Ò}‘=‚WÄ;š~£íqYµ³ñ·ÓÏŠ¶®žT#ø#•ÞÖ~ZñBÈèÔÎÆ÷YŸäRâm« K+rKÁÎiCX¨o©+ZœA…‰K$,EB%,“Ó Ë:EâMX¾‘°¬“J^ƒ°$«þa)èªÜž[>Ûb÷}K×¼7`©¤t¹°_©…ŽˆfòDñ=\_ŠP&u ùI/”auŸP}!`{w3i/\J“ÙSÒÞLZiP®Ä¯Â¤} )]§än'¥ õÐí©¥o/ÄÑ\H˜¨ß^¡`I°\©]¾É„`—15‚_ÃDaÀûˆ ¯O-N_§ÔaKz™þ×'‘ÙÀqz Ïø+öüžâ&„ßZ‡Ok*¢±VÄ¢ø„ƒÕA¨žCäÊ Yw•W³îJ»iªÝÕmT|ý6Æ$èÝìå¶DïþÝ´ h_GœðûkZ2hÒuÚ§Rçƾ¨'ÊÉ:í“ëÍ¥ï6Fx£´O¥4íßËD[æ3ÎÑ’Í#òò"vq-ëo²YR$±ºé7_ú°°>*×+8e7Vrþï3ó¿4w†‘kXô·±0_µÐ Òy´hN´ÎÙ&[ç;¦Ì5i™­Ò‰âußenûRÜÜË9šÙàlõ…WDªk1G¶¤{_¾º0PG7G„Úˆ«t¤ÄG8Ñæù ü°S#o÷þ÷Vc”·Û¢¬¼d¡`,ÐvÃ6³D1SïmðÙÜ×}]èFµ ï†n7t»¡Û£›«æ ÙK´ˆ=<ðíoÿù Ô3ׂz“›NwC½ê õò½¸óË-zé†wW‹wÚ/â0ÚWéðÑ”eã £õ6¥ýÏõ‘‚(Nï—Ù|†t»—-=AKUå~þìÙOuF=ò˜Mÿ us%,áÅ;B°HÅqo eÄÃÊHsaäwßü¤ó endstream endobj 218 0 obj << /Type /Page /Contents 219 0 R /Resources 217 0 R /MediaBox [0 0 612 792] /Parent 151 0 R /Annots [ 208 0 R 209 0 R 210 0 R 211 0 R 212 0 R 213 0 R 214 0 R 215 0 R 216 0 R ] >> endobj 208 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.935 692.124 51.396 700.563] /A << /S /GoTo /D (noweb.NW2i2LRH-4XnUPC-1) >> >> endobj 209 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [101.92 689.487 113.381 702.383] /A << /S /GoTo /D (noweb.NW2i2LRH-4XnUPC-1) >> >> endobj 210 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [159.469 665.813 171.432 677.917] /A << /S /GoTo /D (noweb.NW2i2LRH-KwRQ2-1) >> >> endobj 211 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [147.285 653.858 159.248 665.962] /A << /S /GoTo /D (noweb.NW2i2LRH-1uBDkH-1) >> >> endobj 212 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [127.888 641.902 139.349 654.007] /A << /S /GoTo /D (noweb.NW2i2LRH-2o3xLS-1) >> >> endobj 213 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [205.823 594.082 217.786 606.186] /A << /S /GoTo /D (noweb.NW2i2LRH-OTtxn-1) >> >> endobj 214 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 527.954 51.396 536.752] /A << /S /GoTo /D (noweb.NW2i2LRH-KwRQ2-1) >> >> endobj 215 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [155.276 525.316 167.239 538.381] /A << /S /GoTo /D (noweb.NW2i2LRH-KwRQ2-1) >> >> endobj 216 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [417.204 525.316 428.665 538.381] /A << /S /GoTo /D (noweb.NW2i2LRH-4XnUPC-1) >> >> endobj 220 0 obj << /D [218 0 R /XYZ 125.795 791.866 null] >> endobj 10 0 obj << /D [218 0 R /XYZ 72 754.004 null] >> endobj 222 0 obj << /D [218 0 R /XYZ 81.963 693.21 null] >> endobj 224 0 obj << /D [218 0 R /XYZ 72 690.483 null] >> endobj 225 0 obj << /D [218 0 R /XYZ 83.504 693.21 null] >> endobj 226 0 obj << /D [218 0 R /XYZ 83.504 681.255 null] >> endobj 227 0 obj << /D [218 0 R /XYZ 83.504 669.3 null] >> endobj 228 0 obj << /D [218 0 R /XYZ 83.504 657.345 null] >> endobj 229 0 obj << /D [218 0 R /XYZ 83.504 645.389 null] >> endobj 230 0 obj << /D [218 0 R /XYZ 83.504 633.434 null] >> endobj 231 0 obj << /D [218 0 R /XYZ 83.504 621.479 null] >> endobj 232 0 obj << /D [218 0 R /XYZ 83.504 609.524 null] >> endobj 233 0 obj << /D [218 0 R /XYZ 83.504 597.569 null] >> endobj 234 0 obj << /D [218 0 R /XYZ 81.963 529.039 null] >> endobj 235 0 obj << /D [218 0 R /XYZ 72 526.312 null] >> endobj 236 0 obj << /D [218 0 R /XYZ 83.504 529.039 null] >> endobj 237 0 obj << /D [218 0 R /XYZ 83.504 517.084 null] >> endobj 238 0 obj << /D [218 0 R /XYZ 83.504 505.129 null] >> endobj 239 0 obj << /D [218 0 R /XYZ 83.504 493.174 null] >> endobj 240 0 obj << /D [218 0 R /XYZ 83.504 481.219 null] >> endobj 241 0 obj << /D [218 0 R /XYZ 83.504 469.264 null] >> endobj 242 0 obj << /D [218 0 R /XYZ 83.504 457.308 null] >> endobj 243 0 obj << /D [218 0 R /XYZ 83.504 445.353 null] >> endobj 244 0 obj << /D [218 0 R /XYZ 83.504 433.398 null] >> endobj 245 0 obj << /D [218 0 R /XYZ 83.504 421.443 null] >> endobj 246 0 obj << /D [218 0 R /XYZ 83.504 409.488 null] >> endobj 247 0 obj << /D [218 0 R /XYZ 83.504 397.533 null] >> endobj 248 0 obj << /D [218 0 R /XYZ 83.504 385.577 null] >> endobj 249 0 obj << /D [218 0 R /XYZ 83.504 373.622 null] >> endobj 250 0 obj << /D [218 0 R /XYZ 83.504 361.667 null] >> endobj 251 0 obj << /D [218 0 R /XYZ 83.504 349.712 null] >> endobj 252 0 obj << /D [218 0 R /XYZ 83.504 337.757 null] >> endobj 253 0 obj << /D [218 0 R /XYZ 83.504 325.802 null] >> endobj 254 0 obj << /D [218 0 R /XYZ 83.504 313.846 null] >> endobj 255 0 obj << /D [218 0 R /XYZ 83.504 301.891 null] >> endobj 256 0 obj << /D [218 0 R /XYZ 83.504 289.936 null] >> endobj 257 0 obj << /D [218 0 R /XYZ 83.504 277.981 null] >> endobj 258 0 obj << /D [218 0 R /XYZ 83.504 266.026 null] >> endobj 259 0 obj << /D [218 0 R /XYZ 83.504 254.071 null] >> endobj 260 0 obj << /D [218 0 R /XYZ 83.504 242.115 null] >> endobj 261 0 obj << /D [218 0 R /XYZ 83.504 230.16 null] >> endobj 262 0 obj << /D [218 0 R /XYZ 83.504 218.205 null] >> endobj 263 0 obj << /D [218 0 R /XYZ 83.504 206.25 null] >> endobj 264 0 obj << /D [218 0 R /XYZ 83.504 194.295 null] >> endobj 265 0 obj << /D [218 0 R /XYZ 83.504 182.34 null] >> endobj 266 0 obj << /D [218 0 R /XYZ 83.504 170.384 null] >> endobj 267 0 obj << /D [218 0 R /XYZ 83.504 158.429 null] >> endobj 268 0 obj << /D [218 0 R /XYZ 83.504 146.474 null] >> endobj 269 0 obj << /D [218 0 R /XYZ 83.504 134.519 null] >> endobj 270 0 obj << /D [218 0 R /XYZ 83.504 122.564 null] >> endobj 271 0 obj << /D [218 0 R /XYZ 83.504 110.609 null] >> endobj 272 0 obj << /D [218 0 R /XYZ 83.504 98.653 null] >> endobj 273 0 obj << /D [218 0 R /XYZ 83.504 86.698 null] >> endobj 274 0 obj << /D [218 0 R /XYZ 83.504 74.743 null] >> endobj 217 0 obj << /Font << /F59 148 0 R /F87 221 0 R /F57 146 0 R /F18 190 0 R /F73 223 0 R /F61 150 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 307 0 obj << /Length 4056 /Filter /FlateDecode >> stream xÚí]Ýs㶿¿B/Ð3g†ø&ÓI›ôÒ´Ó>$×—\n:´YK¤†¤|ç4Éß^`’EÉ”ŸíÆã™Lø\ìþ»€¢Éõ$š|ó*²ŸHQýw3iRÿxõ—·¯>ÃÄ$ Žùäí|ÂEH)Jˆ²ÉÛÙä]@£‹÷oÿÞMB6 ѤÔÿšäwßèr˜Sâ!ÁXe€~7¢€s¾¼»Dqë‹KŠƒh´~øšq"(ÔÆTmš–@7@§öù¹­Põp è$@×@Ý'•“ÎÖ@S'-lÍâHGX0÷RÓ“vI’`Tz)”’àLjác•[0±F¸¡%ÐèG ÐP“¿*G×¢š„üŒ‚V hŸ>mktºš9yÊ­Š‡„CO2\Ú%}‰P˜°F4 g x£€Œ›kL1óÌÏĘán¤‘ë­'žªÉ Õ~V43¢aÍÜÜ2OÞ2·ék?B„ÃJj„÷ ¨ŸµÏÌòŒt{•@™ù(|4GÕ07 _½é8Š>GÅ^9ŠGV@•­PŠa’…£SG€vb*öU½¸—»N–€Ø¯<¹äñ%±]ß[ŭל×%Œƒ¥í ?¬ÑðžD·~éå@îógÒò§^ãT> Âq9ÕÊ^/¯}á®NÄÇ ÌiúOx"Î/œÅ9ê¿ódt籿Î/¹½i¹ .Hì L[ ½Ëã‚=m.X9VDê0y·(v‡#±b¼plïÃm—“ôÊËr%\ˆàA‘›Ïz!M"߬®¤ÞñQ0¤˜›gY¾ÞÔ&¹NËt%kYV¯ÐG±[%&8ä¸S9­bpò ªÂ›<7yñ¡2eXHÕŒÒÖi+Ìì—ÅZ–i-g¦eYÓì2Íò®3Í»°T¿©P:€© !£‰©p]ÊY6­3Ós¬Š™ /”àÁ[3,X(í¸”·riþ­êr3­7¥Ü–ºÉ>UEØvUæ3µŸób¹Tý³åÿy^7C[«æ•<«³t™U©n“V5ÝQðìD5HÛm—å¼(ÕâÇH)‰|.K™O¥ù7Íg:Áƒ™œf3û°RýO-«Ú¤*iÝ`Tæå´´oæRÎä̤ [@‘/ïlQ iò—²Ú,maz\œ/Ý:}±ªX”Vë3¹R-—³pµûðH&*( #Š-H¹ÚtÎÑ𨙆ŸÌ4 @òX´Ðèj%â>d„9 ¢[È(= ÛŠCLÅ6|Ûôq2<¯Ô`¬4KÖfŽ6Uz-‡Á¢À-dÓH¬3‹mH<¦‡‰Û=Èýup]SYY´ÒM'Óòz³’y=4ý(!éf—Í €ä„Î"¢åtª½.äôÆtpq§DH'û‰Â˜Äm7åžIå‡íœ]]„yˆbº­Œ¾›f‰¡öÑéb~`~Ì7í N‹²”Óúµùÿƒ-MN…}Ñ| 3¶ÉâS%J©$}>R©yPìmÄÙY®829G‹£±K0V³…ã%è¼½÷I%¥+¹Ø’JìêH©„@´¸¢‰D:1hmÅ® =èEE“ZTÉn§Æ Õh ¸éÒÐmX ˆŽ€:Ôâ 8ÿ ÿüló©HUQBº­óNE$ø”Œ— f(ˆÀáöFãÒÖ Ÿí@àu¿­<îlDƒZ †mÓíÛg[¯ÎûopÎT@S Úº%½íðe(3è‹QyÇ´‡DÁ»Ñçv¶—@ðÖ–ØéÈ|?ŨžÅª¶÷>Gôþ ‹a9±sæ°ƒ‡` ¬7]_Mœô,À³zÉ@.a‚zbîÀ‹Iä³°AÀú&`•ï¸85…û)ŒGãqû8“Âa#˜Ü<ʤ`|<‚ó_û\ÅwÇ-9 óû#¤øg}…£tÁq;»F£@ËÏC1z˜0$bfcÍÊ#æí‘)ƒ=îŽö›yS¤Ü\AÁØgÁs(rÅoüì³Å£ø%‰ÿ”>>Né‹Ó”¾_}ˬfÉ¡)óÇQrÏ_éûž”NéãñJßë¤<šÒWß=ÚHtà0éû|Ç'€9ì†+}Á/˜ã7Ž9ºP±#­õ§ B’O³ó€„àÇÝyxd,Nè‰ Dø!æØ~¬‡2÷{¬@G€Øï*¸{ŽÈî»°ãÜ;ºÑâŸÒ û©¥ :6NÔ–¾Õ@~´ÿ)A­gUZ·‘Vij£Coâ¤zó½„s/¼ þP馒óÍÒ<Éy ±ú‹Ût¹‘•ùbq-ê¡\­ë;ó¬õ„¾6±(à¾Ôy¦E~+Ëz(8‡ Ƹóæó¡ð…­?Q—:åƒÔ@·®Âevs_9ua{²É›Èm«â—²2±:Dõ*ÚŽwÙWrÔ¾’i>ز.rsTQÓâv0bI1Hçš­­Oy*—ËÆ-]¦w/Úz­s©# Òòî"ÁA¨ ¾mbL¬Ïù6-³ôJ Á8o3BqHé¹›g>¢_N ˆ@"dI2Ñ ˆA÷úžµ¯³+é9úžc{°*®€®Ö@ïú2XØ-,m¨dN°¥b"έœgZÝ]"Œ£,”(ì ñÄ_:qÃÝyªÎç6¤íÁAZýÑR4ÝØ>bäåô" ¦þNWUö\•?aÒœ€mÞÌã¼ Xå ¡ÀEsꩲ&<ÁLëÑÇ#F ܯ—I§—‡X¿;T 1ºÝI~£¾— ¨ahÞuKMÏXÜ24Ñ ­§n»ã-?›o\NQð_`@ô ‡¸—%öö¼Œ°à±/¾å6øØ]cK ¥Æ0‡8‹5Ð <ñ|¸ºòy–±YU†• HpVáìÜÖhs÷hzÒMgQïÊ+MÿtߦÎóÐ¥=?vGšîßl+dH¤…Ì>‘%Ä1‚ì•^ÅŽüY+§W7/øQœvŸqqêòÏkOe PÜö÷T¯k«¼((¯†`FíÊ•ÈÌVÒ:­~=,œvŒßÆë™9:s9€ß\ãwç<ÍýÛOÇnÄG÷úJm劭!²ÇÌO»Ü]„ 8ì­L½÷Þxô tâþ}a :1_ñ+P@Än ›'7¾~n¡†þûü 5¨í¼Ñ¨w–ë `W9¿f${ÅÛvî\¶>íýšQ|à׌>Q(éY[aÔ²ú­ÃðËCgb÷ ˆš£ÔòÒm{ñÙ-°ñÞí]p1‡|ürÑíî¶ Í ¿Boïµ£¸µ®]½ršy iÆú‹Fˈø¸Ã=C3â¥*ÜÌÖqJ÷)ý4îSS{¤hôÊÙu*ýyüâÁŸû:«Hz²Îåaå#»¢µc½ù“‰0¹ë ¸æGÛÖNüI鈶öXd:î[•ÏÎœþWm –¹‹ÛÛÙ¥¥§ãƒc Õ—RŸUJ8|ôÔÝ8)žêf¢ç.‹·Ç¹b‡uén¼PçT´ÅçÁ©‘ü7„Á7¸™„˜ú"€¾è6ú"Ç£/#¢ˆý¹¸}ÝÓ¨m¤¯y%jlãz=w‰ÜõX¸ uF)IrpOý¡Bé â¢I‡¸XÔG\-Ê*`÷'¯zˆ Ü@žwWóóš×ÎnJ«›ž âbIO,xk7}FX+~¬µÏÿ€qÝœ„q¬¿¾}õ?Õ t endstream endobj 306 0 obj << /Type /Page /Contents 307 0 R /Resources 305 0 R /MediaBox [0 0 612 792] /Parent 151 0 R /Annots [ 278 0 R 279 0 R 280 0 R 281 0 R 282 0 R 283 0 R 284 0 R 285 0 R 286 0 R 287 0 R 288 0 R 289 0 R 290 0 R 291 0 R 292 0 R 293 0 R 294 0 R 295 0 R 296 0 R 297 0 R 298 0 R 299 0 R 300 0 R 301 0 R 302 0 R 303 0 R 304 0 R ] >> endobj 278 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.478 679.783 165.939 691.887] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 279 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.935 665.014 51.396 673.454] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 280 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [149.811 662.377 161.272 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 281 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [326.967 662.377 338.93 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-KwRQ2-1) >> >> endobj 282 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [339.43 662.377 346.408 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-2yIl95-1) >> >> endobj 283 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [346.908 662.377 353.886 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-4ZZkMy-1) >> >> endobj 284 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [354.386 662.377 366.349 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-1oKncb-1) >> >> endobj 285 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [366.849 662.377 378.812 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-FUd2L-1) >> >> endobj 286 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [379.313 662.377 391.276 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-1KNHNk-1) >> >> endobj 287 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [391.776 662.377 403.739 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-4BH61R-1) >> >> endobj 288 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [404.239 662.377 416.202 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-4Mxm99-1) >> >> endobj 289 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [416.702 662.377 428.665 675.442] /A << /S /GoTo /D (noweb.NW2i2LRH-3dLroe-1) >> >> endobj 290 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 570.355 51.396 579.153] /A << /S /GoTo /D (noweb.NW2i2LRH-1uBDkH-1) >> >> endobj 291 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [141.934 567.717 153.898 580.782] /A << /S /GoTo /D (noweb.NW2i2LRH-1uBDkH-1) >> >> endobj 292 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [417.204 567.717 428.665 580.782] /A << /S /GoTo /D (noweb.NW2i2LRH-4XnUPC-1) >> >> endobj 293 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [161.113 555.999 172.072 568.103] /A << /S /GoTo /D (noweb.NW2i2LRH-CpvDB-1) >> >> endobj 294 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [193.781 544.043 205.744 555.994] /A << /S /GoTo /D (noweb.NW2i2LRH-1fgq0k-1) >> >> endobj 295 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [188.221 532.088 199.18 544.193] /A << /S /GoTo /D (noweb.NW2i2LRH-27nDss-1) >> >> endobj 296 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [40.437 492.432 51.396 500.872] /A << /S /GoTo /D (noweb.NW2i2LRH-CpvDB-1) >> >> endobj 297 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [157.076 489.795 168.035 502.86] /A << /S /GoTo /D (noweb.NW2i2LRH-CpvDB-1) >> >> endobj 298 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [416.702 489.795 428.665 502.86] /A << /S /GoTo /D (noweb.NW2i2LRH-1uBDkH-1) >> >> endobj 299 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 365.095 51.396 373.893] /A << /S /GoTo /D (noweb.NW2i2LRH-1fgq0k-1) >> >> endobj 300 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [192.847 362.457 204.81 375.354] /A << /S /GoTo /D (noweb.NW2i2LRH-1fgq0k-1) >> >> endobj 301 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [416.702 362.457 428.665 375.354] /A << /S /GoTo /D (noweb.NW2i2LRH-1uBDkH-1) >> >> endobj 302 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [40.437 154.072 51.396 162.511] /A << /S /GoTo /D (noweb.NW2i2LRH-27nDss-1) >> >> endobj 303 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [186.76 151.434 197.719 164.499] /A << /S /GoTo /D (noweb.NW2i2LRH-27nDss-1) >> >> endobj 304 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [416.702 151.434 428.665 164.499] /A << /S /GoTo /D (noweb.NW2i2LRH-1uBDkH-1) >> >> endobj 308 0 obj << /D [306 0 R /XYZ 71 791.866 null] >> endobj 309 0 obj << /D [306 0 R /XYZ 83.504 755.001 null] >> endobj 310 0 obj << /D [306 0 R /XYZ 83.504 743.045 null] >> endobj 311 0 obj << /D [306 0 R /XYZ 83.504 731.09 null] >> endobj 312 0 obj << /D [306 0 R /XYZ 83.504 719.135 null] >> endobj 313 0 obj << /D [306 0 R /XYZ 83.504 707.18 null] >> endobj 314 0 obj << /D [306 0 R /XYZ 83.504 695.225 null] >> endobj 315 0 obj << /D [306 0 R /XYZ 81.963 666.1 null] >> endobj 316 0 obj << /D [306 0 R /XYZ 72 663.373 null] >> endobj 317 0 obj << /D [306 0 R /XYZ 83.504 666.1 null] >> endobj 275 0 obj << /D [306 0 R /XYZ 81.963 571.441 null] >> endobj 318 0 obj << /D [306 0 R /XYZ 72 568.713 null] >> endobj 319 0 obj << /D [306 0 R /XYZ 83.504 571.441 null] >> endobj 320 0 obj << /D [306 0 R /XYZ 83.504 559.485 null] >> endobj 321 0 obj << /D [306 0 R /XYZ 83.504 547.53 null] >> endobj 322 0 obj << /D [306 0 R /XYZ 81.963 493.518 null] >> endobj 323 0 obj << /D [306 0 R /XYZ 72 490.791 null] >> endobj 324 0 obj << /D [306 0 R /XYZ 83.504 493.518 null] >> endobj 325 0 obj << /D [306 0 R /XYZ 83.504 481.563 null] >> endobj 326 0 obj << /D [306 0 R /XYZ 83.504 469.608 null] >> endobj 327 0 obj << /D [306 0 R /XYZ 83.504 457.653 null] >> endobj 328 0 obj << /D [306 0 R /XYZ 83.504 445.697 null] >> endobj 329 0 obj << /D [306 0 R /XYZ 83.504 433.742 null] >> endobj 330 0 obj << /D [306 0 R /XYZ 81.963 366.181 null] >> endobj 331 0 obj << /D [306 0 R /XYZ 72 363.454 null] >> endobj 332 0 obj << /D [306 0 R /XYZ 83.504 366.181 null] >> endobj 333 0 obj << /D [306 0 R /XYZ 83.504 354.226 null] >> endobj 334 0 obj << /D [306 0 R /XYZ 83.504 342.271 null] >> endobj 335 0 obj << /D [306 0 R /XYZ 83.504 330.315 null] >> endobj 336 0 obj << /D [306 0 R /XYZ 83.504 318.36 null] >> endobj 337 0 obj << /D [306 0 R /XYZ 83.504 306.405 null] >> endobj 338 0 obj << /D [306 0 R /XYZ 83.504 294.45 null] >> endobj 339 0 obj << /D [306 0 R /XYZ 83.504 282.495 null] >> endobj 340 0 obj << /D [306 0 R /XYZ 83.504 270.54 null] >> endobj 341 0 obj << /D [306 0 R /XYZ 83.504 258.584 null] >> endobj 342 0 obj << /D [306 0 R /XYZ 83.504 246.629 null] >> endobj 343 0 obj << /D [306 0 R /XYZ 83.504 234.674 null] >> endobj 344 0 obj << /D [306 0 R /XYZ 83.504 222.719 null] >> endobj 345 0 obj << /D [306 0 R /XYZ 81.963 155.158 null] >> endobj 346 0 obj << /D [306 0 R /XYZ 72 152.43 null] >> endobj 347 0 obj << /D [306 0 R /XYZ 83.504 155.158 null] >> endobj 348 0 obj << /D [306 0 R /XYZ 83.504 143.202 null] >> endobj 349 0 obj << /D [306 0 R /XYZ 83.504 131.247 null] >> endobj 350 0 obj << /D [306 0 R /XYZ 83.504 119.292 null] >> endobj 351 0 obj << /D [306 0 R /XYZ 83.504 107.337 null] >> endobj 352 0 obj << /D [306 0 R /XYZ 83.504 95.382 null] >> endobj 353 0 obj << /D [306 0 R /XYZ 83.504 83.427 null] >> endobj 354 0 obj << /D [306 0 R /XYZ 83.504 71.471 null] >> endobj 305 0 obj << /Font << /F57 146 0 R /F55 149 0 R /F73 223 0 R /F61 150 0 R /F18 190 0 R >> /ProcSet [ /PDF /Text ] >> endobj 371 0 obj << /Length 4405 /Filter /FlateDecode >> stream xÚåk¯Û¶õ{~…¿tÓm{U‘âK-ÒÃÖ¢kP E€bMúA±e_-¶äÉòMoÒì·<¤hJ–%ÙæMS h^Š:âãð¼ÍV³höí“Èü"Yª¯gMíû'}þä‹o(Ÿ%aÂ0›=_Î8AfœÄaDèìùbö"Hn~yþ G½‰fÕj¦+?~«@PŠBÁù^Îonc,[ f/nYœÿ5!š…DCÔÕ¡ Çêe‚,U×-”k(]Ðò½—nn©îä×Qð KlonE|ÝC Ÿ‚yغþ,CÁŸ[ úR®!§¢†¯§¾ÇIç²5l$‡å o‰Õè%Ôø‚=cA(KÚù&t}ª^À“_ üÊ|¡»]jŽT¯aе½†}˜ÒA“[ÄB%ò… 5x‚¢3EÃ@”“¸Á( ààÊÊõ– Ãxni€Æ”}5n1vØYÊ]àÔì_ÜTáÇÕßO‡G ¶£€î¡ÜÀ²ñ!ÜÁÇ{Dî­šl‡)Ho.š°3 Æ¢Ù\=— æUA½4-'fçñ²8øÁ..…Å¥Á+(3(«I¦|IRœ²¯‰:»é °\…ö–…þ±`õJMjKóÔ×g¹&k€µ°–F° ´È2uZR賺îY;õ žî|paY æÃa8,XA½‚23ŒÈÛv°à ±ÊÊ;h”tè•·qÙhĘƒFЈMD# VàÓ r*“Ô§MÁðùì‰b‡‚Éu+à(.:€Ú¿0)%Q‚49E¡H˜™$«î»b™UY1W`%MÌwê—Û¬Z–Õ&[èæWºyž®×y±Òõ]¦[Ù.¯š®¹ P²¬¾+aœ_|ƒ„;œ8B!eIJÝb©{µ-'Ÿ ;yÉU(ArÐx]¥yadz)™XboÌ×Óù<ÛÖúqªÀϤ&BÂå* 8¤Ä¬Â2Í×ûJ½IʹԿz~²âNIþi¦uÅ#ÒÂü‘ïv{Ó'U?Ið&­Ôøº¨‘!ìr g±¯ôØeýîA®ø6­R <«tÛ:ÓoÛLw¹¡4¸…‰DíØwåb®vj¿èêTÿûMVårÕŸñaÒêÙ&}ЕW¦!­ël³­³Åçêoɨöµ~ –¥o;%vEÄnçFnÌ&›õí)a$hÓs¹/æu^q¬ªtž-÷ëõƒÙ7š„'íYWÙ¼¼Wë…¹–U¹Ñµú.ß…j×ið7³Èªy+‘4‡o}nÞ8¬€Þvý²ip¶]ýÙl;<ÚéߴпYU•Õ BHbhß‘„òDK|W'1–kÅÔ\DÈb¡çÂR½V€u°d žŽÃgpp ûœÔ\/œ~"•¡[Ì:c<¥™Æ9ržü)AÄRx¾ÌlÌ7¡6–?UλàöÈ ‰¦˜][kÉgغ0»ZuÒaá•i¿±%K.çÕsÁ°9‰9”¯¡ü»³(ËɦšhÀ—¡Ü<¿Áý6fÔé3OaǾWªû‰ø¤¹¿±©kwÍØÎà3v† "xfåVÈæéîˆÍ'ï[ìÁĦb‹AÑÐÎÙãPp« ï34š€K#}áàæ Ü7<¤Ùvu$Ô«èàvùjº¬ÇÑ8; µ§r&r¨28pº~+Öà|ªV‰¨©rp'INóÊÏ ? þt"µ]QÉדªß–ϳô¬Úbeí°†“iê‰=g~]³ÿ‚Å-¡Üûô›¤Ö¼Ý”ÚÐ*ÝîÑ!ѯàˆA'™¨à\5´È(ÆZUÎAU֞̕OÏÈA˜¤£Âä•óé ”®F ”Üã>3>*T^åâÁ;yˆ—‰¢œ©@YÞB;‚ö÷P‰b2™a‘é~£|ÓѦEÄÙ¤;Áhܳ3‘>*X­á$g ´q?ÃÁ\ƒ·åësã2 _ÌÙ4OK»ðÊŠîö#D‚¯á [¹H)A<ñe5¦‘‘KKÕPujäûºÇ÷{¥å•;L’Gsî()”~L×M0Mq0M‰Ç0M‘GPI)3ø¥ò~ì Œ;ñOÌ¢Ý[b[ü¾wú{„0¬ p|~Mo~Lº”‰.sÊE«~ 9Cñ`$R—¶'!–˜nVºklqŒÐ&^.w‚à–7¹›úÇb«“D•'yã–,ƒ25–ËÛ„éÇ ³ÆðIùÞ½ 9”´‰`Á‘:s FÁ¯ Å@¯è •FOüªµÜ"vƒ¶„kì¾—l èn`|&»N\þx&?Fãpu•CÝ÷=ÒºÒŽE´Z*1µ ¨» dciŠØ:’_ܨ¨Å~¼¥Â¨m?Âþ¢ñæ€Ð–².ûÑéXXÐM„æ.B#@hì"4u+„6FZJ“…6Iû¹u0ßÁ Þ;¬Sß,e[ú¡çc›à>Ï@ïÉd†49±ºâ<\qˆ_®œ$~'qˆ_WtÀ3†˜@†/!~­o–˜Ÿ¿ñÌ—V aµÇ#Î'Ƈ`jAÇ“„’óãá;ž9hù)hü:\`n‚. %O:cÃReiÄ–Ô±<¯ä‘Ê¢^ùZ¡*–:ËÁ|úF¾&&öYÇAÀh|ç˜,–VFíóõ…•\eÖHŽBKè@h‰ J³“¹c#ÊÎøÔ §»”éDnñ&9YÑ×ð2<&úRÌ~Ó«Tð¾ˆb-û™otë‰dúÈÑm¢ñ.Ä ý>rÔWn”X­¸n Ü:ínTÚÊçj˜Ué3Î}nCjš¬7½p¼Ùdr0b¶sŠü…2ª—›œÂW4Cí0ƒÔÉ~(Ž Q«Kd!E–ÚüO×á‡U…¤Çm¼wœÇõMËQìÏ»4o¼KR »¿&ñO€ª$åâdäOÁa×›úg¾Úµ.ŒÍK‚˜N»5ùDÁÒ4šŸ4=t"q ¨>'é¦mNN†QÓ£=Ó3ŽœX“ìLOîá4«p,eð:CŽÏÏÎSç2¶žÚ{ÇÌ\Yã&o¤KFÇ-&*Sí©VãÞ ‹ÃR¿ü–,…%ûáhIµr¹½=G_"Ùûq!ŒZàŠ{‡Fx!1$|yãu8s]AøÇÂ/fWôÃÄ8SS¿Œ-Å?"&—ŠŽéG”²öÓ Í'RÆ*·*_(iÒvdSº®²taZçåf»¯UnšzTd«´Îï3ýl]®tó&­VyiQª9­ó»²4/Õeç/£W»Z·–ûz ™QêËÕj¿ÉŠZÑá( /Lµç˾Œ-›5§ˆªl[VµI.ªMvÑ6­ê\gk Ù½Êïaôæé››¡É:ªMº“M^j'”™Wr“âTí‹ÂfEåEºS¹ÈB;î$ÄÈŒûŸ6‹JAržA%°»Ø"Öp©¨ŸuYnu7“±%Ûv›t½ÖÕWi=¿Ë DXrÙ3½/sm“mÊêA×·UùjmLçeYéެ鰖;‘PÙ®î[ý]Vï&çjQŽL®7!Z¯Näj©ž“sµÌz™´³lWwÓÔv=¹WHŽ ›¢Õ?y9™Å¥.È8$P]ÅÅú²¸zr¶„ª¹Ü@Y¡¶6Õ¬yÊTõT3ß~ƒ³P¢ †`ȧ_iAà™coH±êKcxü·Êñ“8øº”w üõé²”Y—Á$¥«¯;0=&N–y%¼µì®‘¾v=ñntjk;ˆ[Û1¢¨ŽË9‘DÎ[nÀc5)¬—˜XÛ‰'·¹kûÉ­<¦ Hl[ƒ-M¼‹ Ø(µÛ#»ôÎãIE¥r.RÐW1¸-;§^ÀÓz$°ðø*†‰§P°d(Ïe0 €3÷Ö!|ÄQvˆU}£–6qÙIT“7E`á„äÔGù,«4&9o¡H…ã×õ·N|kežzÊ!ÉÏ,ÛôÜ*nìdê‹Iÿ&~Hºd…Ï€-:ém–:âqj2”ö'3Ÿû‘ü×þ Ò©Ú})Á§á˜h±f K®èßømS6ÜÝKÄWL½™VM¼máýýÞ×¾Øß3?€tzäsQUøf#HùK¥ö`Æ"Kï\×Q ï^ÖÓœ³/ôƒþáèGK"Æí›åŒ|­ŸÅÂ7-éã¦ñÑèÇ¥{Êοõâ`-sð?pG2Yfj,p•³ØÙÉe&žC˜–Î"Í>ËÊàr&þÉ1¢*ªË=ÖVG4ÅR«­Æ·Ózrô;‰×öF³^ñÚŠÔÓòv¦ª·ÖÆÉú’·W½u¯5ŒAn%£×jQ„YÒ€Lº›.£«¤tÌå¦|âëþ Ÿ×Pí¯DÆãÖGS®ðû‘8âÄïÈئK¸ïé–’íQÕ=¬ln.­8ï ðò áu5Ò¼1N/“'UØ#$ªÐák/ú½Ñï;)¾Ÿ¨òðËîq;i¢÷ ¼ƒy~ å{ EJl¢øÈ+~]½ }1ÿ‘Šní*U RëˆY9²/oÉò‡J|† ³ï’dúâNÕ'©¹IëÎ9W—Ý·ËiNûÇ{§i@v&±WÙœ¢$P+… ž²ò´»맰Wà»Q-‰Š Q ¡ÿ§€4ß_š¦y(ºÑ0—\œ<ÅÍ75ú–âs2lÚqÀÏokçfß¼1½i_Â^Å&L_r»¾<-ñ;?áI#~ƒ†ÃI\¥ÐÊI£öi~ ÝžB=jŸoÛç*³#K’Ä·<³ó%+<ówO˜G9‹™Œõ­ã*Éz"±¯°kÖNªxîä 6”¦¹‹jmíÂÓçmøkz§] ¿NÖ‘þr]$?ëMA¼¾08U]Þ9'õ툺,Ùð¯À}åK Vš/»·SÁ|ùÈ„qUM»²üŠ„ÍÒgþg“Ö•:áD©¯œá­ÃízóêoÞÅ—ÜÇ#ÈJcä «±Ìµ/‡xåØ kk§„”ª&DN¶!ËbâëÙŠc’øsÓö£ ñ‹‚Â/ ’¯ðýnͲþÊTçà4´²0u/ã V¤Ž xE*=º˜Œ1î2»÷ónº·\Mž^yÆÐÿ7™OŒp±vœc™£©}¹¾ä™­C\ËAÇv -|]áY;¦¢&àÈSÌÈŒOò.Ÿ¤cgò žjkÉ4³}óŸZrÞZNÓþcvCvcí³ð}Ì!ªÄ­ÿ?©;X¯Nbä³iLÓ_üŠÆš°·5Ómã–; ÿª{rõç7o®¥iþ;.÷„²–”ùâ6›†;™}NE3c…1á³[œ„Bˆ>÷ßߟ?ù¸"á® endstream endobj 370 0 obj << /Type /Page /Contents 371 0 R /Resources 369 0 R /MediaBox [0 0 612 792] /Parent 151 0 R /Annots [ 363 0 R 364 0 R 365 0 R 366 0 R 367 0 R 368 0 R ] >> endobj 363 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.935 624.2 51.396 632.639] /A << /S /GoTo /D (noweb.NW2i2LRH-2o3xLS-1) >> >> endobj 364 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [120.695 621.563 132.156 634.628] /A << /S /GoTo /D (noweb.NW2i2LRH-2o3xLS-1) >> >> endobj 365 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [417.204 621.563 428.665 634.628] /A << /S /GoTo /D (noweb.NW2i2LRH-4XnUPC-1) >> >> endobj 366 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 237.758 51.396 246.556] /A << /S /GoTo /D (noweb.NW2i2LRH-OTtxn-1) >> >> endobj 367 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [193.436 235.12 205.399 248.185] /A << /S /GoTo /D (noweb.NW2i2LRH-OTtxn-1) >> >> endobj 368 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [417.204 235.12 428.665 248.185] /A << /S /GoTo /D (noweb.NW2i2LRH-4XnUPC-1) >> >> endobj 372 0 obj << /D [370 0 R /XYZ 125.795 791.866 null] >> endobj 373 0 obj << /D [370 0 R /XYZ 83.504 755.001 null] >> endobj 374 0 obj << /D [370 0 R /XYZ 83.504 743.045 null] >> endobj 375 0 obj << /D [370 0 R /XYZ 83.504 731.09 null] >> endobj 376 0 obj << /D [370 0 R /XYZ 83.504 719.135 null] >> endobj 276 0 obj << /D [370 0 R /XYZ 81.963 625.286 null] >> endobj 377 0 obj << /D [370 0 R /XYZ 72 622.559 null] >> endobj 378 0 obj << /D [370 0 R /XYZ 83.504 625.286 null] >> endobj 379 0 obj << /D [370 0 R /XYZ 83.504 613.331 null] >> endobj 380 0 obj << /D [370 0 R /XYZ 83.504 601.376 null] >> endobj 381 0 obj << /D [370 0 R /XYZ 83.504 589.421 null] >> endobj 382 0 obj << /D [370 0 R /XYZ 83.504 577.465 null] >> endobj 383 0 obj << /D [370 0 R /XYZ 83.504 565.51 null] >> endobj 384 0 obj << /D [370 0 R /XYZ 83.504 553.555 null] >> endobj 385 0 obj << /D [370 0 R /XYZ 83.504 541.6 null] >> endobj 386 0 obj << /D [370 0 R /XYZ 83.504 529.645 null] >> endobj 387 0 obj << /D [370 0 R /XYZ 83.504 517.69 null] >> endobj 388 0 obj << /D [370 0 R /XYZ 83.504 505.734 null] >> endobj 389 0 obj << /D [370 0 R /XYZ 83.504 493.779 null] >> endobj 390 0 obj << /D [370 0 R /XYZ 83.504 481.824 null] >> endobj 391 0 obj << /D [370 0 R /XYZ 83.504 469.869 null] >> endobj 392 0 obj << /D [370 0 R /XYZ 83.504 457.914 null] >> endobj 393 0 obj << /D [370 0 R /XYZ 83.504 445.959 null] >> endobj 394 0 obj << /D [370 0 R /XYZ 83.504 434.003 null] >> endobj 395 0 obj << /D [370 0 R /XYZ 83.504 422.048 null] >> endobj 396 0 obj << /D [370 0 R /XYZ 83.504 410.093 null] >> endobj 397 0 obj << /D [370 0 R /XYZ 83.504 398.138 null] >> endobj 398 0 obj << /D [370 0 R /XYZ 83.504 386.183 null] >> endobj 399 0 obj << /D [370 0 R /XYZ 83.504 374.228 null] >> endobj 400 0 obj << /D [370 0 R /XYZ 83.504 362.272 null] >> endobj 401 0 obj << /D [370 0 R /XYZ 83.504 350.317 null] >> endobj 402 0 obj << /D [370 0 R /XYZ 83.504 338.362 null] >> endobj 277 0 obj << /D [370 0 R /XYZ 81.963 238.844 null] >> endobj 403 0 obj << /D [370 0 R /XYZ 72 236.116 null] >> endobj 404 0 obj << /D [370 0 R /XYZ 83.504 238.844 null] >> endobj 405 0 obj << /D [370 0 R /XYZ 83.504 226.889 null] >> endobj 406 0 obj << /D [370 0 R /XYZ 83.504 214.933 null] >> endobj 407 0 obj << /D [370 0 R /XYZ 83.504 202.978 null] >> endobj 408 0 obj << /D [370 0 R /XYZ 83.504 191.023 null] >> endobj 409 0 obj << /D [370 0 R /XYZ 83.504 179.068 null] >> endobj 410 0 obj << /D [370 0 R /XYZ 83.504 167.113 null] >> endobj 411 0 obj << /D [370 0 R /XYZ 83.504 155.158 null] >> endobj 412 0 obj << /D [370 0 R /XYZ 83.504 143.202 null] >> endobj 413 0 obj << /D [370 0 R /XYZ 83.504 131.247 null] >> endobj 414 0 obj << /D [370 0 R /XYZ 83.504 119.292 null] >> endobj 415 0 obj << /D [370 0 R /XYZ 83.504 107.337 null] >> endobj 416 0 obj << /D [370 0 R /XYZ 83.504 95.382 null] >> endobj 417 0 obj << /D [370 0 R /XYZ 83.504 83.427 null] >> endobj 418 0 obj << /D [370 0 R /XYZ 83.504 71.471 null] >> endobj 369 0 obj << /Font << /F57 146 0 R /F55 149 0 R /F18 190 0 R /F73 223 0 R /F61 150 0 R >> /ProcSet [ /PDF /Text ] >> endobj 425 0 obj << /Length 2360 /Filter /FlateDecode >> stream xÚíZ[oÛF~ϯÐË¢Tb±œg&»íC‹mÐMž £ûäAQd[°, íÆ-úßwøÍ…CJ²({éÂÀјäÜÎœËwΙbt9*F¯žî—Úü]|ëõ³Ο}û“#ë’–£ó‹Q)sÎùHr–\ŒÎ?ŽÞf¤¿?ÿO¨éJF›Ë‘müòªCDc2çº4/Ñû·ñ„óìÊP–-Ð^‚Î;ƒŽÞNJQf«ñ„‰l :­Ç]fÿOJâÞn]Ç\Œr·ÛüåU´Æ}oÍ$JRšý£·ŸGTr­ÍNÊlZÎ@ç [GSL&Ivc†S`¡j8áÚ:û€ö´^ÎAÓ,ÂA…#¸H2œàX/Å)ØEÍ“$+e`[†æ˜†'b…‘Õ¬¨R±Ân zçØ’„’#°}áÚèô¾3Í>3ѳ§àBr-¼‘az&Ê\æMÂœûhtšéì;óCuö®…iiˆ¶åÁ6"1Q<{65ÛÆKo È(çnóh:m§Í~Í$ ÐÕŽÉ)ý¼*š—¶óšç/ªô‘ ÌcP>椬å#„fg»5‘…[ºù÷Ÿ3†“[¯];‘è­#q[€®@kÐDº¨Òê¢:`×Åv}ˆÚ8%èªê©d]µ%÷jsÕ|úíO%‰¾`¹’ÁÕÞL¯Í ©4ØÌ?.fõ¢Zmm'³e:™ÁL'YæLm”ö‹ÎÄ<×J´êÚ,qa 4WDõ¶¨YEýTû•Ðp»±LE–ƒX͵#÷‰Uí¹Sú /L—žêKžH…2#$REx*³f»ò¤®Ð ò-è ´Ž„z“FwtZU,a4ÖËa§¼ÿ”j‚ĈfIpRDSÂ*)X%‰Ã”8X †HL¦@ý“i&–ŸÏ?xZáHmt`º~”u5†¤czhqzä"‚1:l?†‚$Ñ_9}=…î¦IÝ}JA;4¢) ±)²Ë‡Ð!¼ ´Ìz¸Å·+´»,Ý.é€%öÝ€Ê <Îäw º át²iÀ­ {vbIâåcc¢ È0 IÓGi@3&Ü‹-^ÐöEiøÜ< =tá{Ø[ô£ñPoñâ}ó™5¬¡-UÇL¤Ìþ|ºßá¥*ÒE²S‡ÇZjÑÚ¥Ãl‰,‰‚·\wž$™Fw$ç2ȉDð#]H¼B»v¶ódkc嵫Ýìtí–*²6F<–‘¨Ìesd~QþwÔÆ¢á$…Ï”ö‘“µcÚøèßÇÕô±RäºÐÄaî\5Á8e&(Wv¸ŸW6 ˜ßÍ7÷¶¹¨ç›iØ« û[_¹ˆaú¡ºsÍeU­Ïló7÷hVݬoëy¯OeàÁEµq£-—îÛù¶vW‹U½=0û‡i=»Ê÷n¼»˜CÍ…j’:TöC˜ˆ3îÃà)ú„ç”óS£'ÅòRìOau;ñâ]aÝ~oÊu®XˆÂšh¦ü€d¨2NrÙ'ŠµÌek½…*^Gú³mõGe¯}Ò¢¤ði´IÕ<˜m)î_ˆÎ`7l®Åýø«ð©dzÜWÐÓÉ„f`ò— "o‡J›¼ŠùèÖèœc3l£Â6îb3ðɼ¦n*@ÂX…ßͲù"LbcA7Ùǧ°{â;ÓÆ%šÎ kt6[|‰Í¹tTÔnKÜ™tmÉ‘CT2ÈlX&p<ò1Ç£à5}’Ö+˜Í‡Dy.`Ë‘"7¡KFÚÕΣÿ Ùý3»þ†çÜIÜÉÐm+#³P‘îÉû2 ¶´Úà“Ó±²G`H'VÕç'¢ˆØì‹]’ã ª}⚇RÜεä”$:V£%oÒ äs„‰*^…+©m‚ˆIWj[õ¢ k±qi§ñ0ôûTùÅ[èµÏ2&Jâüˆ-_…@ +d”.— Î0ô>U1t=7Ñ©ÜDEÏ:Šö6 Ë ¢“{>Êt¸:Ù°„¡²ÀÀ*¦%ÑóYãù}*€áŒ¿üï°ï ™ RP»†€bO´†Á);c(,Ðœ¶¶ï×ñ„ÐBÚ0“4öN xT½#ŒÛF{h‡ê…¾&:¹ _=îÉP4Œ Šg}S­<%.ô°K·êâ']Ey›"¹Ú™”´~;TD 2eÑÐÈAŸãíëÃUÈAA¯åwç€Ê#ĉ좦-ÌŠšÌÚ¯#ƒµµ1Àäε Ò|oÝÍ®Clã_ÁÄÜA¯æ͉üÑS=u±•'ùt,ÄRa! •RÀBò‹`!I£Ì·Â‹(W´ 5 j'çÙ¬u¤NžŒ‘4ÓCÓlä†fEù•c¡PP¨ Vuª+_À v.ÀŽ!ËèŠWŒjV;À>ƒ²[oˆÔº(0±M³ÛH%kŽâ/G­É|co§m£/íÛÞVh¯41Ý5êË»!sÒ/Ž¥[Ø@7dæk}3‡3Ò3âQŠì ¼Õsx«]GnS/æ!é&³¾yz™I+Y~ÉØ]L%Âl§’áÝbß_ɽ×&OöW.·¤O÷Wm¶)}šæº:u‰ƒêÏžºQ ê2g¶*ýéð<”v‚ÎÑqô*F½´…gg@k@š¸ê‰hMj‘®Æº9p óÎ=OTü´î _ðì ü,R„eªÌ˜½¹ßåغû’I¨ÜE´!h>8ª's€,œ™‡êié¢FÛB ¤†ì)áîê7|{-Í×}ÝÅߨ¨ˆjnÆz-1ë:*.9ø]m$dÜûbƒÅ{m`¿Ãq£%ÞÛâÄw¶÷½-îÇkÜ/êøÊÛ™Žêo×ýë04ºc«ƒûK'û܉°×§ÍªŽ€6»'ÞV+Býâ„ì†=¦îÁþ%ïsþqÛ7Ô„{Û½‹¬+Ué„1Þ±è c¨¡,ŒD’/t¿äÐ=J‹œqƒl™b9/Õ¾TÌ¿ÏŸýØ”H^ endstream endobj 424 0 obj << /Type /Page /Contents 425 0 R /Resources 423 0 R /MediaBox [0 0 612 792] /Parent 454 0 R /Annots [ 419 0 R 420 0 R 421 0 R 422 0 R ] >> endobj 419 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [174.281 715.648 181.259 727.753] /A << /S /GoTo /D (noweb.NW2i2LRH-1ep4xx-1) >> >> endobj 420 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [44.418 591.338 51.396 599.777] /A << /S /GoTo /D (noweb.NW2i2LRH-1ep4xx-1) >> >> endobj 421 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [158.898 588.701 165.876 601.766] /A << /S /GoTo /D (noweb.NW2i2LRH-1ep4xx-1) >> >> endobj 422 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [416.702 588.701 428.665 601.766] /A << /S /GoTo /D (noweb.NW2i2LRH-OTtxn-1) >> >> endobj 426 0 obj << /D [424 0 R /XYZ 71 791.866 null] >> endobj 427 0 obj << /D [424 0 R /XYZ 83.504 755.001 null] >> endobj 428 0 obj << /D [424 0 R /XYZ 83.504 743.045 null] >> endobj 429 0 obj << /D [424 0 R /XYZ 83.504 731.09 null] >> endobj 430 0 obj << /D [424 0 R /XYZ 83.504 719.135 null] >> endobj 431 0 obj << /D [424 0 R /XYZ 83.504 707.18 null] >> endobj 432 0 obj << /D [424 0 R /XYZ 83.504 695.225 null] >> endobj 433 0 obj << /D [424 0 R /XYZ 83.504 683.27 null] >> endobj 434 0 obj << /D [424 0 R /XYZ 83.504 671.314 null] >> endobj 435 0 obj << /D [424 0 R /XYZ 83.504 659.359 null] >> endobj 436 0 obj << /D [424 0 R /XYZ 83.504 647.404 null] >> endobj 437 0 obj << /D [424 0 R /XYZ 81.963 592.424 null] >> endobj 438 0 obj << /D [424 0 R /XYZ 72 589.697 null] >> endobj 439 0 obj << /D [424 0 R /XYZ 83.504 592.424 null] >> endobj 440 0 obj << /D [424 0 R /XYZ 83.504 580.469 null] >> endobj 441 0 obj << /D [424 0 R /XYZ 83.504 568.514 null] >> endobj 442 0 obj << /D [424 0 R /XYZ 83.504 556.559 null] >> endobj 443 0 obj << /D [424 0 R /XYZ 83.504 544.603 null] >> endobj 444 0 obj << /D [424 0 R /XYZ 83.504 532.648 null] >> endobj 445 0 obj << /D [424 0 R /XYZ 83.504 520.693 null] >> endobj 446 0 obj << /D [424 0 R /XYZ 83.504 508.738 null] >> endobj 447 0 obj << /D [424 0 R /XYZ 83.504 496.783 null] >> endobj 448 0 obj << /D [424 0 R /XYZ 83.504 484.828 null] >> endobj 449 0 obj << /D [424 0 R /XYZ 83.504 472.872 null] >> endobj 450 0 obj << /D [424 0 R /XYZ 83.504 460.917 null] >> endobj 451 0 obj << /D [424 0 R /XYZ 83.504 448.962 null] >> endobj 452 0 obj << /D [424 0 R /XYZ 83.504 437.007 null] >> endobj 453 0 obj << /D [424 0 R /XYZ 83.504 425.052 null] >> endobj 423 0 obj << /Font << /F57 146 0 R /F55 149 0 R /F73 223 0 R /F61 150 0 R >> /ProcSet [ /PDF /Text ] >> endobj 461 0 obj << /Length 3989 /Filter /FlateDecode >> stream xÚík“Û¶ñ{~…¾tJ'>šxÒÉdš6ö¸yL¦¹&3y| %žÄ9ŠT)ê|×__pPNw†äÕYJ<ãH ‹Å¾°X&£é(½ú$±¿ºüäÅËTtty5Êè(ã,Nx:ºœŒ~س ’$Iôº¾*Ú¢Ï.h&£ïŠnÖL–Ï~¿ü—n!‘$V‰"} Éè‚%1K¤iÀ{‘)Íí‹ú†EykŸ¶ÅrÑÔËòMU˜_®šÖü2næ‹UWÖSsÛÍl…ß’4É‹¶¹-çyWè[bj,šeW´¥~`#Ò‡eYL•Ð0l}Ý]C 4N¹rµž?»à”¸®©ßõEÿ¶¯R±Ô» ^„×<°hÕÅ4ïÊ›¾ Ê£ª™šÇó¼–u^ÙÇåuQ•³¦™ìžHgúÒ‚UW¿ì„=‰yJ]¥¼ž˜ŽÊniúXämWšÓh¢1up-wöHY,3æ›<Ô¥ˆ“[oŸiÅí3BHwqlz5ÈÓ=ÎîE«aÈç…C^ªç$y–êô îH.M è¶vCD¬È€ W2º4HMsÓÕ¼¨ :TÔ5ö¯ù]EW«zÜ•Mmî …ê îÓÉ¤Ú :M[Œ÷«"¯ xC÷;:Èdœe™{Þ !Þ¾ƒqs“·eÞ/C@:×ÈÖ•6ö±®9I2 ^·44.”OÓœ%öBS>ŒP¿ºs€$kz çL¯Ï6/k`}7“¼Ëwu!ULÅÐÃíÎahîÅR¿ý]“Mc*‡•q·«4,óÈP&,ú±¸)Ú~õõ0–kÖØ``ýMOxfyôÌY7¸1Uå|Q=à ýŤXŽÛò{ÞÍÊ¥¹Z€ö¸ouî/€.c%ï!¿Ð+:•ý’ÔØÒë ú”fŒšl.ý¾¢#ySE ÒÑSÊ] =d+:âù€³uÇBiN"]å=òXÉarÊßJ`D=möÿ®Gîê3*¥§F¥™?ãLŠÄ*¦²…8ÅdÔö·æò߀ÔkG“ÉÌÿÐÀã¿Ž~½)‰^ë‰Ï¢Ê+({ºQ;\»_Çð¼@é8K¢y¿P¡uPΠl ìWW-:ÓÍe,ú,®¢x@P®<@qp hßS0î©Qnt”@¼ñ(‡I]xe e¥U]B9‡ÒÔï‚IÁô Á.€€` ×Gª†lËa ý”ÉâýêГÕÀdá4ª%dŽ…–W@™9 yåž,„dð«ˆj¬lFë-˜±·¦]÷a4¢•·š!ŒÕrDÃX­ GOƒ±êµóÝ;ù[v‚¦Âðæ•·¶ù·´Ë¤‚ò§c¶Á ÿ¨,oX?7×5p”Î{2Á‚:p$Xy<{̳ !~ŽBˆ —ƒøõðNœ™¤–C™94zÈ×p};p®ÌÎmw|ˆ(OèŸCÿFL2èŸ#ôÅÉÕmÕË_ºÒS½2$ AËŠp A)J ËŠ"1!øKñ4“#Ì·Ú j£ó¿×äÛ:Ÿèðâ"ft‹{äöaÚ9œ‹í‚ƒàÂq ™œ¦ÕŤէ—ÒàÖÿ‰eVÎÁ|Ìï—%”·§hV¦ÊîTµÞ^bé¹÷ýÀÄ)ÖÖ´AH 峿»ÇÖªÔ¹ð…»?9_ØËío\4 ”•G:sKR¡í É°ÖK ~íBZC‰tXBî£t\&‘Úí48ú?D 7|Îv9ö߃zQñ#-ͽ#!Ó{Á8|žÈž5%çbµš-I¬½Û?ˆÕ*¸¥Ób.'hjï× VÌÿ l™˜´ ÊUðöQ¢î²&€–ïÌ*V…xYŸ[ë<åë3œ‹®ëöžb5:÷NÕµÃh6ÕÇÜFý)â?ó< ¡ÿ32Àv¤·ùùn¶óžg:ë”<9"ÄÔ®GøMð6‘Í•šFŸbÇx ýØqêÀø®–„—[;X¦ˆó0t)îöÄQ܉•uæxI9¸äVF;õhЯô2Ö8n€yÄÚf€ø¾ìl.ÿõ™ÎÛÐä©)^^—oO,‘‘öå×ÂÕ”;Aö°bMýO±y¹V?=´ŸKÂÂ¥& Ôãnõþ í„";‘ˆ}FN$þ^«%”+/«såÒ¬qšð©Ïäü‹=~´(p- s¼-“¨†ÎœýD;¹¿}Âå6<Úe•–‰çk*<üˆñq%ÌDWžPlmºÖDOƒ5i5âû}<.ÉNÆñ¥<ä‹f¦•³b(¨ì7FLù÷6΄ØÈRy[ýø]ì±ÁOi«g®q׋`Hs¯iû-ìõkËßJeî…U1´°*{í$㦞è(‰¿ïÌÞ¾óÏXØDõ*oGÃí\L‡ÌßÙÞ1þ&Åõßp$ë‰äp~…E ðí^ðÓÐÝ´¸ö$“óöÀRÓ8.‘4±yš­/ÒÍé¸hŸâó?F·ÜøhèúÈ×Ë,©NG&kÙöÙ>nÿcÇ †J{é¤=Á–ö\AX¤-HÄ!¢ŠŒÌÔ^àÁÒÌ“c~8t—˜8‘|Ê,=ëó©!ð6•6tË7  üï)Jûí¶Æs6ÍùÚqžÄn›Ô'®ÐêžlvN°‰÷EYÌï2ÙoN?ô ¸ûŸ­yèpgùÙ›™ rzaâ7ÅrG¨ðÁþ{q<)uÚÕúÃì넺白®½éÌ=Õ¯ÜR ƒ¿QMˆÂó³)1ñœ“â±ÄÜê°ÅðýK‡ÍÖ³é:Üd-¾Òþ!•/¡þ—ös*¡ªâK;ºòâô+ë©é¯—Á†ˆ±ˆëGô|«©´G¶+ï|pƒ0™¹öÏÅÁº=’s>;Ýp30ñÅËŒmö˜×!ªñR¯‹e6T˜.æz.i&£q³¸kËé¬3¯h0¤~Eðþ!ãDq÷Nš›*ݪ˜ åj”[ˆr€êFI«Dœ¢±âltAYL)3¯~ßôºÕ쥪ÌEY_mQ s;/ºY3YÚJ­}:n拼+ßTöþmÙͶکÊë¢*gM31÷W«zÜ•M½|®ï™ŒŠxÇæ§â6w;{/mÇM]Á¾¬ØaqÊ•Ñ,— ’:(^å+ýC^›çkPú>iÒ]›jM;)ZS§kÌ£EÑ^5íÜ< éa¦ZÈçcÛA‹©U H?¾ÉÛ²Y-ÍóEÛ,.vA]´]Y, 4Wæo7+Ì…·þÞÇ[¿,àÖÜðØ’’¥ýLëîX“€cm6“¤ÿ¸ºAtÿæëÑŽþ> endobj 455 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [44.418 638.203 51.396 646.642] /A << /S /GoTo /D (noweb.NW2i2LRH-2yIl95-1) >> >> endobj 456 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [144.345 635.565 151.323 648.63] /A << /S /GoTo /D (noweb.NW2i2LRH-2yIl95-1) >> >> endobj 457 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.478 181.505 165.939 193.61] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 458 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [264.989 131.537 273.047 143.838] /A << /S /GoTo /D (section.4) >> >> endobj 462 0 obj << /D [460 0 R /XYZ 125.795 791.866 null] >> endobj 14 0 obj << /D [460 0 R /XYZ 72 754.004 null] >> endobj 355 0 obj << /D [460 0 R /XYZ 81.963 639.289 null] >> endobj 463 0 obj << /D [460 0 R /XYZ 72 636.561 null] >> endobj 464 0 obj << /D [460 0 R /XYZ 83.504 639.289 null] >> endobj 465 0 obj << /D [460 0 R /XYZ 83.504 627.333 null] >> endobj 466 0 obj << /D [460 0 R /XYZ 83.504 615.378 null] >> endobj 467 0 obj << /D [460 0 R /XYZ 83.504 603.423 null] >> endobj 468 0 obj << /D [460 0 R /XYZ 83.504 591.468 null] >> endobj 469 0 obj << /D [460 0 R /XYZ 83.504 579.513 null] >> endobj 470 0 obj << /D [460 0 R /XYZ 83.504 567.558 null] >> endobj 471 0 obj << /D [460 0 R /XYZ 83.504 555.602 null] >> endobj 472 0 obj << /D [460 0 R /XYZ 83.504 543.647 null] >> endobj 473 0 obj << /D [460 0 R /XYZ 83.504 531.692 null] >> endobj 474 0 obj << /D [460 0 R /XYZ 83.504 519.737 null] >> endobj 475 0 obj << /D [460 0 R /XYZ 83.504 507.782 null] >> endobj 476 0 obj << /D [460 0 R /XYZ 83.504 495.827 null] >> endobj 477 0 obj << /D [460 0 R /XYZ 83.504 483.871 null] >> endobj 478 0 obj << /D [460 0 R /XYZ 83.504 471.916 null] >> endobj 479 0 obj << /D [460 0 R /XYZ 83.504 459.961 null] >> endobj 480 0 obj << /D [460 0 R /XYZ 83.504 448.006 null] >> endobj 481 0 obj << /D [460 0 R /XYZ 83.504 436.051 null] >> endobj 482 0 obj << /D [460 0 R /XYZ 83.504 424.096 null] >> endobj 483 0 obj << /D [460 0 R /XYZ 83.504 412.14 null] >> endobj 484 0 obj << /D [460 0 R /XYZ 83.504 400.185 null] >> endobj 485 0 obj << /D [460 0 R /XYZ 83.504 388.23 null] >> endobj 486 0 obj << /D [460 0 R /XYZ 83.504 376.275 null] >> endobj 487 0 obj << /D [460 0 R /XYZ 83.504 364.32 null] >> endobj 488 0 obj << /D [460 0 R /XYZ 83.504 352.365 null] >> endobj 489 0 obj << /D [460 0 R /XYZ 83.504 340.409 null] >> endobj 490 0 obj << /D [460 0 R /XYZ 83.504 328.454 null] >> endobj 491 0 obj << /D [460 0 R /XYZ 83.504 316.499 null] >> endobj 492 0 obj << /D [460 0 R /XYZ 83.504 304.544 null] >> endobj 493 0 obj << /D [460 0 R /XYZ 83.504 292.589 null] >> endobj 494 0 obj << /D [460 0 R /XYZ 83.504 280.634 null] >> endobj 495 0 obj << /D [460 0 R /XYZ 83.504 268.678 null] >> endobj 496 0 obj << /D [460 0 R /XYZ 83.504 256.723 null] >> endobj 497 0 obj << /D [460 0 R /XYZ 83.504 244.768 null] >> endobj 498 0 obj << /D [460 0 R /XYZ 83.504 232.813 null] >> endobj 499 0 obj << /D [460 0 R /XYZ 83.504 220.858 null] >> endobj 500 0 obj << /D [460 0 R /XYZ 83.504 208.903 null] >> endobj 501 0 obj << /D [460 0 R /XYZ 83.504 196.947 null] >> endobj 18 0 obj << /D [460 0 R /XYZ 72 116.995 null] >> endobj 459 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F18 190 0 R /F73 223 0 R /F61 150 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 507 0 obj << /Length 4688 /Filter /FlateDecode >> stream xÚí]{#7rÿ?… °f½êåûqÁ^.¹³‹}—øn‘ ¼g²¦g¦oõ²¤ï$H>{ŠEv7»EI-‰s1‚À0ÕÛÍ.’Ū_=ÈæÑýˆŒ¾zEÂï?¾õöK©G”–X:z7Òl¤/ˆ£÷·£ãòÓt¶›?ßL˜•ãÝCµ-n&ܪñoïü­j~·þw6ÏË[ýSµ{ðWSÿ3¯>–óêaµ î—³]µZú­vå¦nhº¬¯JñÕôq»­¦Ë7Ð>!Q»ÛÇ2´]¿Sn6«Í ¥t:ûˆ|ÿþŸ`ÌÊ )¬[MÞÒf¼^mw如7¡êÛ/5‰Ùb)¾Fü‹?†*6®¢ Ei]ãÏD_Iv*‰‚ ^Wºó5L§%^(ÞÔøËoR-QZpm¢¦èÍDJ5~—¢Çô‹Õ•ªÐ¸Š)¤UPZÊk’„¤¸ kõhÕLAÒ4-þ%EˆœñvL^é.CµhȼI÷ÇȆåÿšš.¬ä½¡ÑTc¦ Ìv†† rP¸w½©‘¼¢”p4&Œ‚ Rf ZK™ =ç(à‘ª»þà«(ÚQר–woáÿ/°/‹fmÛ^¤€&Û„Ùk›¬€ëȈWŽî¿£úêk?0 Ä™Âi[PåÆEŸÊS =–ÀÌŒ6÷#ñdŽŒh¸ŽkÑŸ=~„R—x=Ãr‡w*¼^…§q3£%éøÃÍDÆB†ÕØx‹%¼®¤AŠdäá£r=m”×#ί}YEO—YšÔatœ‘œÌñÚßÀrÕ@Œ:Ý<4 Å¸È#vük½GÚ–k,kƒ²ÌÕXÙÀs}´8gÖ©!·¿^`Yb¹‹žVxýÑ[þ©¿¿Â:Ë“˜@Rh\?\ô¸úy—‡åFÐÌÙ\G"ÎQ96Ãq9,Ç9â¡Îf˜‚Pʼ¿˜0w8ä~¦ó`/JT¯{¼žî¹?O•¬É¼¶&PÞç"ºˆÀsƒå}4‚e32Íg„Vh¡Õt ·…Ö`oóŒØûpÖŒÛLD¥g ’¶ÍÕÓÛˆ9›Èî©œ^°nX¶ËesvzÁö²—IRMúÓI¯D¡“èHT*‰‚&t(.—Þ}ž#ÊU¨ÏP€êïBCýäUr«„üëÿ7AH D/ ÑÐ'ß›–XNCÏzDá ŸNR*8kqGC–×g~ ûh_ׄ{uÞËg1X"û0P':\ç¿”ìsBTßaàÞ1KÐÕgìnÏ×&UHbºÂCÉ™™IU‰úSÞ›ƒfâX®OùÑl½B F‰ Q‡&TÍuà(ÖiåX]9ÍÞ™öÉL– ¤jú=€{B´9Ê -‹Ñ”çJªñïÑäý3–ï±üb é4ã_äÊ -QÒV¹üŠ²ÉTëÄ™G ®Y4Ó»\nÞ Y;4yÂ<)Pì32'fÅG¼îgk°Û^‚A?ºÔ_IaF‡XcÓJÌ`oš,ˆ@WI$ ˜4+ÙPL©¢1¤«ÓêW5®ÃšQ¼ôœ|C]EYæm”ù!RÎ2g³§rèW’ÿ*Jçt5¦…–ibÍîšfO$|êû§>Çìß·Wà–ôÜ—!8ÀLo”žþÕ~‘£ÕÅ$~¶NÅQG\œ5þ€yãWUƒk³ÄzÜ ’T °ËÖÓž~øØ5,º2\ ¥—ÌTèxw¨røP¿8"l"àùBé¯ç‡|âÍj“ª…—´ÿ'ðPŠñݸʎÒ…„ü3JÉýÈ×£g¸|¼Âåãh]™¤yv¡a¥L«\N™¤!þ9 ókây™+™<‹#΋n¢dò42T³€?Yš'!,˜î >¬ût‰Öï˵.׋—”kƒÃ­“NÓÚkI·< ÝӈʥöœøÝT‘YÆ%ÑÙdœ¿æ)ò=çQ¨ÕʸΗ‹m¹N®gZbxB9-£@×Kñ*‘á¸X~õp«U^sBä#3»Ì,óY¢“âσqþ³/òµ‰‚‡ ¿¡6!ÇÞ—æA¹“ËDW[Eòm´YFHÕ,ÙŠ\‚j:Z±IìÍhÃØl¹ø¬‹ä‡¢Ì´ÏœÅ£³Ã±ý›Ø.CTÿm®ÛCu˜­à…qç T¿Eîk“~¢=té9±…\mZÞ„R¬”Ä^§$F3•ÞÝD#uø7Vú·M £{ñæÅZcÑË`}OaðÅ#3Y5Pi-µŽ—î,T¿”ãñÿ‹í6ãLú½+" “—ZCFN™6Û,Õ™ëɯꔎÃsÜòêSÖuèÈðÙ<ŠèJpß5~Hå#–{¦”R=þ%.!\µƒÉu¾ïn3.ô Œì05œÉàþÔMVç³åöe>,XDù¦vC~“ZvÖV²˜ZÚù¦écäð[Ÿ5.iö·Y0}Ì3Æ:(™÷}²õ»:Ó‹F©qfŸÑž M^à:š\ż(&r屘O[䂺À–›“é¯E%NgñkѾû«bSr°R_wìÙ£CDêod«Häï'þýªMÃõs ^ÿÜów‰ }}Ý zYõȆÚŽç6ܹ©¨ö¹P˜Nðâ0”áõ•‹¤ŒçCËÉ@k¨_à{iÃZÒ4 mî3æΣ-¯ç‚CPõ.8˜ól²£Ay6›LzÉP›Ì®s÷5ÏÁró³¶ÉòÅl²ÒG¿™²Ñ7Sæä7SǾÞýû‹ÄD³+ÞG¸-„ѵ0§O?€0*lŒß_š¼ó¬Û¬œmU%^(ŸGk+³ÈÙɺhnBdÝlôk+Ù2#ô>îW"_{ô“?“å“¿ bœ'ç%AOÈq{ηía>MâBbÊâKì—@Üû™¿éžžrÊh›IyÈs”—<ž·¸¾ß_׸ âðô˜Ãˆ»^soµñ`ŽN³—Òz8Wi¥”åÛ­µŽ6J•Í*QíÞ/šmèMŸÉ« ²Õü8"ÑxËøËèNéÌ2wy‰þð¦ó3·Œ:TÑ™: %DU¤>fLî«ãÚáWó9N¢ÿÞ»Þöud«—Ü3(¸/1ÞÚ(0G[¯ç]þóãí²ìèú‡$…µô¸4ý¾ùm—±âWí9;ÃvQð[\Ö©aD\¿E,,pJ_ŸœÒì»Rqßž²Ñ¾½ýÙ?¸+µ¨0[7‘˜L\¼ÅÎ2¼¾tD¡+<â }»n‡v,ʵñ”×ztµ(÷¿ ¹Ž&Û/¾¿>™Ø5Ÿ£s†bqeùöÒÍÙ·¢ž'õmzk¬2ñgŸ×`¤Îƒ‘æX?ý´W7í× SÛ¼Å!!£&qÄ$ñ;Šé=˜Çw|j¾j¨eÕ }c•ùY`UÝ{~rEfêN*Ù#ÈË|l“8òÑŽ|Ã̧¥iaÜŽ$¿ s æ¾×%£L×óé¬üìÆ⺯û.ÿÖëÍêSµ˜âɧ‰£(!>b®wÔûrÈËt¬º»+7årWM˜—þ^{¢êÖŒjhÛ¢ßiå*MãCµö^jáI«î¢=þIÇ?<*þ§945q ê :=¼õîüoµ #_¬nËðhê@û³‡jyß«9{Üo÷‹÷¯~ F’‚FÒBS òËÅh¶xõá{2º…‡€Í…àjôV]€ åNßþøêÛÔÁß -Ý„–Ûƒa÷ç•) ¥0âÚ¹Õ!S0IÎ)”Ž>PK*‚µ Ñ:’n?902¸ùËTc`%(h_Tíë<¢ã·6¦5i&°­¤ Û"MKƒå¶.¬l¤'ÙÈ{ÒÉxC®ƒàÿ‡ìAX\÷|b<ƒ!À‹wH‡zÿ– FÀ ×€ ×ÑyC=¶Ck|îºg=Ž÷Í€”Âu›ÛÐÎ|y€Y–ªvë^ÒD ®Ï¶çÄÈŠÛzTh)ºãî“aÀí¶ÖiCx—hÌ£O¢µ¾XöZë@nKÎ ê©‚Ú½˜¥¨M [@hÕ{çQo’š!,Þ²¬™žœkÚά Õ:>§Oåtš@žÃd2õæ9%'(Mä\9> ïS³¤c·¢pëÜ Âj€êƒ91ÔÅE¢$¨Û{bè¾ß‘S ŽˆVUÀ_5ðHû¤E!9@ˆ½D×T‰dçÒDµjßµ£Twù ØLê#f̲Eß<²£&kR0 V7ú«»»Æ$µ#›MàŠa Ò±I}]) ÕSðëa ²=^ÿÙ2À’Z'1tÏ©Ãh·!óy²7 Ç´éÍŸ’ŽG¬‘züS!)‘•ÀMÝ#Ôe8¨R]\KpÈ'Q½w.c$Ó¤V9‡Áñl€òø ýŽlq þÂÙ‚€É#²U“Ò¬ÐFײ•è—- Ä PK˜¾r´• v—…¨ x•6˜¢ . =i/õË@Ü‹ÚKþâö²FÞÀ¶Ü<ùÄ‹OñW×›rVmñÏá¸Ûw>£êÄŒî¦^ XKHÿc#$Ltœ»qŸNïçšú°á¼¯~仧8ŒŠÖõšpKÎ!ÄWZ²¿1‡¨}øØK(½oÄ;}ö’yºÏô¨ÿi íR-Lƒ¹yòSìÎà‡(½ZNÝúèü½¤d$±,oF÷ÝÑÉÖQ}—òÒ.‚ø˜ÂÔööéØÈ@¼7‰ª½´gņxVüeÜôñ¬e1x»“hÈ…ñª×P_ÈœÃMÚÔ“ÇîÆQ÷§Š †r0kRÇ®‚ûƒYáD endstream endobj 506 0 obj << /Type /Page /Contents 507 0 R /Resources 505 0 R /MediaBox [0 0 612 792] /Parent 454 0 R /Annots [ 502 0 R 503 0 R 504 0 R ] >> endobj 502 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [44.418 710.261 51.396 718.7] /A << /S /GoTo /D (noweb.NW2i2LRH-4ZZkMy-1) >> >> endobj 503 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [146.167 707.623 153.145 720.688] /A << /S /GoTo /D (noweb.NW2i2LRH-4ZZkMy-1) >> >> endobj 504 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.478 624.173 165.939 636.278] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 508 0 obj << /D [506 0 R /XYZ 71 791.866 null] >> endobj 356 0 obj << /D [506 0 R /XYZ 81.963 711.347 null] >> endobj 509 0 obj << /D [506 0 R /XYZ 72 708.619 null] >> endobj 510 0 obj << /D [506 0 R /XYZ 83.504 711.347 null] >> endobj 511 0 obj << /D [506 0 R /XYZ 83.504 699.391 null] >> endobj 512 0 obj << /D [506 0 R /XYZ 83.504 687.436 null] >> endobj 513 0 obj << /D [506 0 R /XYZ 83.504 675.481 null] >> endobj 514 0 obj << /D [506 0 R /XYZ 83.504 663.526 null] >> endobj 515 0 obj << /D [506 0 R /XYZ 83.504 651.571 null] >> endobj 516 0 obj << /D [506 0 R /XYZ 83.504 639.616 null] >> endobj 517 0 obj << /D [506 0 R /XYZ 83.504 627.66 null] >> endobj 518 0 obj << /D [506 0 R /XYZ 83.504 615.705 null] >> endobj 519 0 obj << /D [506 0 R /XYZ 83.504 603.75 null] >> endobj 520 0 obj << /D [506 0 R /XYZ 83.504 591.795 null] >> endobj 521 0 obj << /D [506 0 R /XYZ 83.504 579.84 null] >> endobj 522 0 obj << /D [506 0 R /XYZ 83.504 567.885 null] >> endobj 523 0 obj << /D [506 0 R /XYZ 83.504 555.929 null] >> endobj 524 0 obj << /D [506 0 R /XYZ 83.504 543.974 null] >> endobj 525 0 obj << /D [506 0 R /XYZ 83.504 532.019 null] >> endobj 526 0 obj << /D [506 0 R /XYZ 83.504 520.064 null] >> endobj 527 0 obj << /D [506 0 R /XYZ 83.504 508.109 null] >> endobj 528 0 obj << /D [506 0 R /XYZ 83.504 496.154 null] >> endobj 529 0 obj << /D [506 0 R /XYZ 83.504 484.198 null] >> endobj 530 0 obj << /D [506 0 R /XYZ 83.504 472.243 null] >> endobj 531 0 obj << /D [506 0 R /XYZ 83.504 460.288 null] >> endobj 532 0 obj << /D [506 0 R /XYZ 83.504 448.333 null] >> endobj 533 0 obj << /D [506 0 R /XYZ 83.504 436.378 null] >> endobj 534 0 obj << /D [506 0 R /XYZ 83.504 424.423 null] >> endobj 535 0 obj << /D [506 0 R /XYZ 83.504 412.467 null] >> endobj 536 0 obj << /D [506 0 R /XYZ 83.504 400.512 null] >> endobj 537 0 obj << /D [506 0 R /XYZ 83.504 388.557 null] >> endobj 538 0 obj << /D [506 0 R /XYZ 83.504 376.602 null] >> endobj 539 0 obj << /D [506 0 R /XYZ 83.504 364.647 null] >> endobj 540 0 obj << /D [506 0 R /XYZ 83.504 352.692 null] >> endobj 541 0 obj << /D [506 0 R /XYZ 83.504 340.736 null] >> endobj 542 0 obj << /D [506 0 R /XYZ 83.504 328.781 null] >> endobj 543 0 obj << /D [506 0 R /XYZ 83.504 316.826 null] >> endobj 544 0 obj << /D [506 0 R /XYZ 83.504 304.871 null] >> endobj 545 0 obj << /D [506 0 R /XYZ 83.504 292.916 null] >> endobj 546 0 obj << /D [506 0 R /XYZ 83.504 280.961 null] >> endobj 547 0 obj << /D [506 0 R /XYZ 83.504 269.005 null] >> endobj 548 0 obj << /D [506 0 R /XYZ 83.504 257.05 null] >> endobj 549 0 obj << /D [506 0 R /XYZ 83.504 245.095 null] >> endobj 22 0 obj << /D [506 0 R /XYZ 72 220.902 null] >> endobj 550 0 obj << /D [506 0 R /XYZ 84.372 125.58 null] >> endobj 505 0 obj << /Font << /F57 146 0 R /F70 201 0 R /F79 203 0 R /F59 148 0 R /F80 204 0 R /F81 206 0 R /F73 223 0 R /F61 150 0 R /F55 149 0 R /F18 190 0 R /F74 202 0 R /F77 551 0 R /F83 207 0 R /F82 552 0 R >> /ProcSet [ /PDF /Text ] >> endobj 557 0 obj << /Length 2804 /Filter /FlateDecode >> stream xÚí[KsÛF¾ëWàf¨ÖÏû‘”·¼[e§6›J9»ªbû‘ ›$h”¬¿=˜€C¤hE®ÍEFÏôL÷×iáèc„£Ÿ.°¿þóêâÅ¡"‚‘Á†DWÓHÑHq†0ÑÕ$z/Wù×lž”éärĸŠ¯ï/?\ýü⯒LÏê›?< k“‚„¬)ÞcFB|EJ‘>)dv#!ªd4"H æ¨~ù‡'3N )ØнtdšöŒDiÉÙYwUà–1$"\y(Ã!Fq!ìEÕH+Cs©8Ý,‰òãˆr^MCl8ºÍ†„$°lXM4 $–²7P—‡FLêšâî&]¥N'îüu]ø›òÆßÐ…W›|5IWîöêRË8¹Ÿåþwúu™,Š,_¸ŸIÙá¡ãy>IC󥚃V°àT†&=â’! è C‚Gý1»Ma@ª´×èˆs¤á‹ö¶=á®fþ2hB AZ†ö°‰Ö·ËI"Åûs£,þ[³ÛµSc²5ÚÑWªzñúêâË…2ŒD`â°®ŠQÄ¢ñüâÝMàÕÏFœÉè®"œÎìG³è¿¿5PÖÍ3¢¢‚‚j@A IUĒ숙†×ãƈFÁýÁH*züª:¢¶€n6 éÈþDf$ àå†êwƒ¬ªFЧQ`ó$2$a-5¬=ÒÓ]EÖŒ*~Ð1÷"êüVõ0@o¦*Qn4MÄ€ê{û%YÎ’qúìRñ¸°À軬Ýz»T©x‘¦ÿ:»…W·©°^:š2w¿Ë›l5q<®Û§Ó|U¿NÝË ¹«Gï1¦e™->º0ú8¬WÍ[MïJ˜‹w¡zž–7ù¤µ{Ýá*°VdSkÞ$8HXTÓ >㊰„ ½Q!º0–;Ó~ÈWÁ! øA|ÞF;*8’ÀC*بöᑨò¾-¼ë9Äš—„Ý÷áÛ«i`^Öxà†Ho)ºfÀ èY³ücp-$b›˜m¶C+È1nÐG™º¡ð&üø £‡=)9Ê´že—ÒÔ#X<ÓP¼ Æ”€§üürD¤ŠIu•1õ¿™·l »eá•Â•z}J“ðä0(ž]‡-—ÃL‚2 ¾Ç$((€¶5 Ñ2‰¾ºKX[I¦‡5ޯýúî™Q šì#±W€Í4¤ó „Va³ÁÕše{ ›°kÞòa‚°ëšWGΆ=È®ye×ä°a´ku”¿û–vMÏcלðuß°©3d†‚Œ Y ¢½‹Ko“ÙÚåØ6®·)½zGªã¼[ø‡³|ìÜögÈ ß›ÄÙ€PÔA*®ùu‘®në Øés.ÃÛ¬Ï}ˆ{3‘ÍàD|ÕÈ‘}NgÙMžûqÆÉlV¸[«=EêéŠt\G&:æõv6e ‹‚à.5·¸«Èë|ŠÝ‹7uƒF¸­'÷nb·¢ºxBêZß°î楽H;ñJ½G$¾¹_>wïíO?õ¾ûàože‹©¼,Ù³]: "‚¶0/‰èûF`kŠ’w‰N—¶…Sµpµ`*¾÷ש®'”•Ç…”þqv”p«´\¯üÞ¦_“q9»oÔ½H[ X ÏÓX“2Rà Ñ!ˆjÝ‚BÂ|0Æñë¯KP›d£8oWù2ùØ2–îä€e+%åœÆ×I‘í-‰³Iš¸‡ùÔ=é b_´©(ìR¾~ëaKî¹ žíïUê•¢zXºÈ£Ÿ’uQd‰çÓ¶KQ¦«y²É^¸|Þ¬çöÀ´T÷ÙP݆C:ïd·ÑCþÚìm—.BÄN`º•cÄÄ›%gÇÓõ¢Â"œqj$eãíÊ^ˆ›“Ê=ƒ$ßòf]É ¦ÛÕ ªÂq ãÃ4íŸï°•'óÀx"E}êÖVÅÁ ß…‹feÓ¯AK$5ítÀíY5q8 Ü’âÊ.«:M‹*T¨H—@@¹¿ÊÆ8 ŸqCì®> ÎÖ°"Ö@H·Î†Ãu6~P?Ôé++ÔÖÊnbõPn ~ªžú]4ÔFÖ “Ã0c:á•Å þa’ɧuQúgU.“U2Oßø¡Žî”aRƘñ|f·G[Õw] — WƇÅz|Sûü­`wšÏfù¯iëƒeVúá&ŸMŠªPZA Ímlú®Ôîz2¬‡™:²¿Âbóå}™XÃK¤ÙþCªîâ”AÕ¤[ÙH› Et©g¡Ó+‡Óˆr0xÒM¯¬mF¬!ø‹e7ÿé±±EƒÙ—@ErZH‘GÛPÕÓÚVMy°xÀ! 9T †yoŽ¸ÞcLÂKÀÅÿ{ ðQd[ Ì ï´SÎ~_º?Á7ˆÕ®þp‡k)dî@2KøÞPÍlðÈÎz.À#ªâõä‘ŒNn«XɾšdE¹Ê®×»SZ«ˆçæk$›öÜCj&GükˆFÂÈS\ußZ³:lópaÞ˜ýZ×ñ©ÿ{/›ömM½­Pœ Ÿ!þ:L HÐ*ngXuå꤈«FÍe@…pÏkNßX¹ŒlGšù®O¡}bUßÑ· ¸† xÆ€ëSh9·ÜÝwí?ö>)Ç7îpH æXîEæy&îÍb=‡pÄ–îÝó"™¦îÕ]ri¨í\¡ ǯߺ÷+ˆn²UÕòÛ;Ý4<„kÄOñªNÁ0{¤fCö¬†ôßjÓ£¶®ª"j“ôþ6=r¶æ%©mè«»›Îô^7¢u%ö(”´&WŸDîñT®)ÀÓëÞëá£kËm—^w–hi·Dû}öúéÝ3ùN²Î#j?}°à?iÈvî>Ên×Ñ%¨“éŽ&Qä‘+¥[½Ò¾bJ©ûG/€"IÅiYÊÀ÷{™.lõIö'wj糪@Bïï·Øø ¼`e(͉}ÏýÑè€bz0¬°'?ì‰ük;¬ð§SOf +b¬ˆÃJÕ&MûmÒw7YÝ72NêÿÞLëæ2Éuóõ}¿ƒúLÇçmÄ%ÁŽ\Ïl‹m¸+WÆóu«ï¸ ­ÖÜ×oŸÙWúqù·k5ž k/ž¯ƒBXÞ>ó¯v¶£ªx³mo„àHÚººf€Ý^‚;$Îÿt~eÿ endstream endobj 556 0 obj << /Type /Page /Contents 557 0 R /Resources 555 0 R /MediaBox [0 0 612 792] /Parent 454 0 R /Annots [ 553 0 R 554 0 R ] >> endobj 553 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [485.286 607.318 493.344 619.62] /A << /S /GoTo /D (section.4) >> >> endobj 554 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [342.585 194.354 350.643 207.004] /A << /S /GoTo /D (section.4) >> >> endobj 558 0 obj << /D [556 0 R /XYZ 125.795 791.866 null] >> endobj 26 0 obj << /D [556 0 R /XYZ 72 479.555 null] >> endobj 555 0 obj << /Font << /F57 146 0 R /F70 201 0 R /F73 223 0 R /F74 202 0 R /F79 203 0 R /F82 552 0 R /F77 551 0 R /F83 207 0 R /F80 204 0 R /F18 190 0 R /F59 148 0 R >> /ProcSet [ /PDF /Text ] >> endobj 564 0 obj << /Length 1763 /Filter /FlateDecode >> stream xÚíYmoÛ6þž_¡O«Ä,ß)bèÐ¥k‹iСA,ÍŦc5¶”ÊÊÛ~ýŽ"%‹*í6]¶¶À¾Ä´t>ÞŸ»{ŽÁÉy‚“—;Øîï<~!tBÒBÐäx–(š(Îæ"9ž&'#†øî˜`ŒGïv5eUžÕyYd‹Ý1Uéh?»3«ÝÓãW G%#5±zp2¦qAœšýle¦î'eá>óbš_çÓ«VÕ¢¼1•[ž•WÅtåÖué>͵©î¼d~aù¼,§ng…û;k‚x1¦ IÝö—^P÷Á<&ÁÐFâ=8¦#Êy+tç$R>”yQÇ[rØÀë¹W¸Ìªó¼£Û»3ÁqNšö̤t~ü ,ÙºÎùžgÂvL øhÊUé ‡RËúrßÆ–±óÖПè—cë]¼QhÁîÕo~Lž4ä&=K¢l«s£!×’„Þƒkµæ½&fù¤+AJðöÆšö£`—MËë ÿøÿð?$IÞzöYÐË­öŽ™ÄHPÝÔfÎýèTÏ3Û (ó}gÆ}.³Û|™»Á¾ÞØU»„ªÑî˜QéÆ7xéš,®ƒ\fU¶4µ©VQà(`VJ}¤Ðö€Z7[«ÑsSØ©±1È6AÆ[3h¿K5Vä+÷¹(ÏÇÐ'Ùµi@hû­BCT ß[\ø¨€FdƒÆ#ûDA %£ª<[˜¥ûžÑúÁI¿ mäá´;Xk¤ÝÔ~r‰G¿™:ËþUºŽnH¢k¼‘EÊ{ÿª¬î21+¢°¥&ÖÓ·™Ñ¶Þ¤… ý™q>sTÉ>Êý«ËìÒâÛ‰Ý9\1.¡úz\ù©ãM¶XZÆŨ™ÚeP¶@ÎPp›L­Ë¡âÁåŒÕðüµÓðë⼬òz¾\¹ç@ Ýâ¨,Æk¾hE³Úµ{Ûé,»'¯Ë©YD¯{KÑú ÙS££ƒ7o÷œ^ #«s›c˜ÑaÖ Ò)sÓ££|r‘¯&sû+<úÉ=|kÌ9D%}“â•Åµ¹u?ŠD»ë*ç,ˆ3S™bbÜKx~|Öø­-³v¯_B5©²Eþ—ñ¶æ…É*çUkHàÕ–p©~oðýàÙëC¿„piäaâDÇD#J¼âÏÀ/«lRçãuãùøÕÊŒþÈÜÔ-iÏZ,¸¯•ùx•Wyqç8€P† â~~›Î#\¼†ù$ ;QšCa°œÁ¾)í0î烧‘-eS”Ò²·_8°Ó¦ñ„ëVoØíÅ#áÄ>(D­²4Eœøæø´WF‚xP¤`Èaˆ†c@ÓLí2‡|?U9(“{q³Ó´£-Ó³8 Rox€¤EŠŠ‡ÁC§ì‹ñ@ï‡mýŽþ›×=[ØeD@Qo%.6\Âéá4ÿÉ%]3= EH5"ͧt—c‰m&ˆ@ß³U*²;`‚8,á÷S’ã!×\ìI\’XöoŸFJ[#ìížN·L"FÄV€µÊ8€‡`ŒÆ†S$\sC s´–~ËÛÄÁàA%LBĪ€ ðu3Ï-hšVåôÓ}{×uv7¸‰ X}sÕµXøÿ#Y7VÆË­ÌÄvû…·vuÿkNGBý¢)ØããùžRH=~AÒW‡eëìÉtn!,€NϺ՜tϺ՜vÏè©[=±ÒºàoLŸß]z©»=÷òäÔ?8ÏüâQ^ÌÞí?Úf{%Á¼W"H ®‹¯r}ÚsBY'b¨Ö9pÀz¼ÿÈ¿Ê79C´j8I7òNp€·¼CçðçGÂ3†Äý#¸â endstream endobj 563 0 obj << /Type /Page /Contents 564 0 R /Resources 562 0 R /MediaBox [0 0 612 792] /Parent 454 0 R /Annots [ 559 0 R 560 0 R 567 0 R 561 0 R ] >> endobj 559 0 obj << /Type /Annot /Border[0 0 1]/H/I/C[0 1 1] /Rect [136.128 570.114 481.764 582.857] /Subtype/Link/A<> >> endobj 560 0 obj << /Type /Annot /Border[0 0 1]/H/I/C[0 1 1] /Rect [183.722 556.483 543.547 569.308] /Subtype/Link/A<> >> endobj 567 0 obj << /Type /Annot /Border[0 0 1]/H/I/C[0 1 1] /Rect [71.004 543.282 110.578 555.759] /Subtype/Link/A<> >> endobj 561 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [328.021 421.523 336.079 434.174] /A << /S /GoTo /D (section.4) >> >> endobj 565 0 obj << /D [563 0 R /XYZ 71 791.866 null] >> endobj 30 0 obj << /D [563 0 R /XYZ 72 754.004 null] >> endobj 562 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F70 201 0 R /F79 203 0 R /F80 204 0 R /F30 566 0 R /F81 206 0 R /F73 223 0 R /F82 552 0 R /F74 202 0 R /F61 150 0 R /F18 190 0 R >> /ProcSet [ /PDF /Text ] >> endobj 570 0 obj << /Length 3299 /Filter /FlateDecode >> stream xÚí\[oã¸~ß_áG±¼‰—óÐÅ^Ðb»˜)Ptf[ŽµU$Ã’'ãßËdI¦-%c§Y`_bZ¢ÉCžs¾sáaðìa†g?‡ýç÷wßýé§XÏGŒ :»[Í$IÎæñìn9û8ç7ÁÏÉþ›æÙº,—7•jþÓ®XÔYYT7ŸïþƒÈÁHcMÌ x1ŒVnŒ¿ÜDL«yÞÂ|_ù!ÜŒuG X¢XiÉŽðÁõ!ÐI"-±í#r$˜tÝ>aF]O©»£ Äõ:ÆØwÄÝŽQΛI÷®‡Âý¡¸j:ü‚"FYÓc¢†ÃÊhÓ(!¡-TЉw:᧬^»}[ï7év“l“Ç´N·U³5¢·èÀña€fcú³˜-Œ{¤Ø²Ê}&æCÎe±Ì §’Ü=_¦E•ÕÍþÐÞÄ”#+ _ æÆýgˆÁ)ÅfQ§ÛïŒÇ|”ÇdŒÇ=J–¡‰ÂþìCX±j»¼MCâq+p–’YćU ØZ†b®= é'Œi‘ñ|UnMƒÏ«E’'[÷0Oê´¨}‡Fƒí·/I¾K«Ð2ˆ’HHX…¨”ˆË¦CRx Ê]½ÙÕÁ1c£ƒ½J÷$—“ù_ ·¢zzqSÝ_0ÂÛ²äçÿ%4.g-±uYæ÷å×[3°œ?¥Žò]åY¶´sãþU!ˆ¥HÓر$ž¦bªRˆ¡RÄúyJk>ìÌo¡1Ƶ‚!¬7Æ–¼ `LGhì¥ó~HÞÙ « –ôjo͉±@ðKFܳÌOÐÝD†$;V›î&kÕWš!Ã@V@d­Ô×fÖ8‚õvÅ >L)cÖ߀á0¤ÿÐk:Öu'sæ,0Û³CEÖS[6¿[m’džl6Ûòkö dàyV¬ÒmZ,üû´xÈ ß^–€F¶U”µýu“g‹¬Î÷¾KVÕYñ°Ëªµ{pŸÖOiêd("D£˜Ð>F.ò¤ª2ƒ“‹ÄaØ{‡X¦±M¶)th^t”þÏ7gµ{c,|®v¹!´þé6[ø÷~Uƒ¡üœy^>õ¾[é>-ðØ×yÚ(Eo%å2u h¦2Ÿõº}ÐÙU5¯êtƒÚ!4¢Ä{vC¯°1InÜ7Tå>‹4]¦¾Û}Y¯]k¹ÛÚ&uFOë,÷Ï6Ût™- ËÐÒqL¿Ó¸²~ï½~ð¿kœÍ#w• k#w•êÃ\Ö¼g¨©‡ñ×­¸~õ]º@Ш`öù8!]h/»q@CC‹c”ñÄ’: é6!à:Ä­ú!è:PĹ8``Ä9‰9$‡­ÅÇ-8íp·aàR19ëÎtMö)€OAw­N*·Ùã&OÁ;i¶++BF^S¡vk6è1hä©Õ”¾±‚1' †Ÿ²\ R9ô`F4 è{XF¢¾˜¾JÍÓ¤z} Ú‰¨b¼ä´'VûVøbè%…>ŸéEôˆð‡æ’BÚu€Ì/É6KŒZ†He`Ÿ;q cAS*ÀL6ÐÀb@¡ t¡…P˜Où-øX¬ÝJ’­ÝyÑ<‚_™ú®¶å£k96B‡Ÿ“€xRxF%[0+&ú±£´6¨Å“¾òEDD€åÖª‰‚ É;89Zveù”½? )eÈ fèÅу˜ñvLþdöU€m@º¤Žmñy¶q«Õà7ª2Ž ß"n¬&‘}´¡C·ÆmþŠ1Ô¢²EÇ/YŒi€DÀªˆcèìýÙÍ ˜dÏ›}@l7ñRóC0Ð$ˆI5Ý<×Zö÷àˆ(‡Ñ„bŽ(˜B’“ìÒÔ‹ì‚&)ɬî¶ýN2ALa¿&§Ÿ¦bÅ” ´ Éë¥aîÍ ìØF,C;„I&9•G³ZϹMÏ>B˜{Ç5µ¯aÞ~’‚00mL…øš¶˜ÎHHã(1Âý‡Â}»Â©Óò‡_ñŠ~ìw,Èå< pdb2ü’pq-@ש)~„&ü†ßKÌùlÒo6{Ãb9,Mtfò±wÆÍÓò¾N2âšo÷û`øÉa5±œÉI¡D|ž ¼Hމ™Jˆ'‰ œx¡œâ½óè$hÃAF½¦Ÿ(§¢€|E³+§¢ª|ŽÙ mDòXè³öר6¹©CGN‡ јȋÐñå¤þ@?SèõðKô„þR8¤Àð(ˆ(&|m@1˜Þ”^´Ø5¼f3ŽY>Ø1°Ñ¡Ó«øLàw üR0Ü Ôœ»J›¿sÈwפÝóA:}‘äyßÀ(gDbÌûGŸÀ8íý„§k6 û˜on÷»Û}E?ßD1‘&õk>`v# ëýæÖ=øøÙ7VÐÝ·*z*ÁƒoûÇûÔ30/šƒ#SUú…†9%¶*íášHˆ±ô§žz€×÷Y‘l÷ VWiÿij²çQMûÄ áà·IúBIí'gŒ¤ö‚p<4•yr¤NƒëVy¢°Rtœ'r<^„æA`ÞE j,Ô騇p‘O|¨OhÏŒIˆ"ªÁg}Lí¡ä‰ aSþÉ{¤ ƒ”]Š5rQz­°ø’s¶#_Öô™ŠƒÐ° w“,$(ì‚óI¦ãHÚO3T6n\ùSMÀs¢,ESе½n}u‚×p[9 ñ̹0L˜ƒ›aòzŽ à1Åxô˜¿mÆæ&"´è¢MÈ8Ý­ Ç"U"Ïrd„bK¦ÅË<Šóiý·y.óvO£;å­ê\y+îçwØ)|iI˜AFî/eRÍ‹xvìB HïÁ ‘®Öb¸8}Ú ýà U' Y†¡˜ôüšÏ¡Á‡Úc3–,9e±ÁRsàgd¢¬' ÅcÇB—H z IÎ^ çÄDœWÀ¹a"Õø œ»îLÓŽóFÿUìßQøz&¯ÌCÒŠ]ݨÓí"¤«À>ÿØ—`Ÿ½xxôÂÒ)ŸXMpŠéUiš$“ëâã&©µZGµHtºEòù&¦(ÒÂœŸàÝós¬mF-+vå®òßmÎÉ´lJÔ4º—ŒìƒeºI‹¥»&_Ë`Æ2ø¡4c¬1¹“i‘…¹ãÑ‹,‚êHqòü«”'êSäv« G‚̪¼Î9XøˆÈ”JøéÊÂÞežsI^•®å2‰ÈÜQóï}·¢¬[®Öëðíx)AÝd¿ý­’À]M§€øáijKÙ«ߌêÖ½’‚w'^D.þ·_Îx®½õÏ \>d£–êYk¥™¿˜É¸r§7ŒƒR=%×r7¬ Qn—ædÂ6Wî3+êôaÛÜô‚î¦ ä.Zu~¸4`¾µ—ÜA‚½˜Ô¿Ü9z‘@‚ú=OOñ’\¸ÐOðIïÞàsëåÂùèñB¿·v€?ûþ€¹ìw>éLÀ Û¨_!¡Ôeƒ‘øü1T'GŽ¼¼Á½p RO/½˜œ B#ت“¹^&S„¥˜ÿ—T;› ˜ºËßFúE¤Þzuj[ßÏ%\(ý’úñ}ø G…Ö(O°çlå%µ¾ÛŽƒw»°ŒOÕëÅWú¿/ª¾__0Ž­—ÞÔUhDµç˜ùïLao‘¡±*ýi£ãX8'ž?eyîZË´N2ß^—O¦A¼7ýÜÝAÓìݾ6C·ÿÇ~uÕÌÐè]B5ƒe¢uœƒ{„éØÿœˆš~=OÞÿï ˆ ¤ÀóŸv[˜d{C± S»`N„_0'rnH³ÎŒyì‚ôXd¦L˶׼“|_g {»ÒÖ‰müúª9=Yº»Ð¾œ¤4,žkaÝ«jPÈNÜ+>ª±„öúüx÷Ýÿ)ÚïS endstream endobj 569 0 obj << /Type /Page /Contents 570 0 R /Resources 568 0 R /MediaBox [0 0 612 792] /Parent 454 0 R >> endobj 571 0 obj << /D [569 0 R /XYZ 125.795 791.866 null] >> endobj 34 0 obj << /D [569 0 R /XYZ 72 754.004 null] >> endobj 38 0 obj << /D [569 0 R /XYZ 72 622.102 null] >> endobj 575 0 obj << /D [569 0 R /XYZ 197.823 509.887 null] >> endobj 568 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F33 572 0 R /F133 573 0 R /F79 203 0 R /F70 201 0 R /F80 204 0 R /F136 574 0 R /F82 552 0 R /F18 190 0 R /F74 202 0 R /F83 207 0 R /F77 551 0 R /F73 223 0 R >> /ProcSet [ /PDF /Text ] >> endobj 580 0 obj << /Length 4145 /Filter /FlateDecode >> stream xÚímÛ¶ù{…`¨®è©â«¤ý°®I×-mŠô¶aHƒÂgË>/¶åÚr.öãG=¤äÇö5¡”G¶Ü%“õ¼¿1LÑàÛO"÷ÿ×7Ÿ|ñD¥ÆÂT)>¸™ b>ˆ¥#©7ãÁ‹@†üêšEQ|·,²õd8Ê®^ÞüÍü,°(L£”•?‹×\‡R1û«›»ìêZ(Ìg¯²ùì.ÏÇ›²#†kwg¶´fÄÂöÕù¦°7GwÃù<[NgË©½•ßþ'¹›vÓ·]»!«y¶È–Å°˜åËðêZ%QðÍl2ÉÖ¦³úüeߢ\6¡’©]öp>Í׳ânaË ÖÙ¯ÛY¹`®x0Þ TÞnVf%îÁ|bŸW(;vïmÿ.rûÀ­»?Ê«m‘?/ÿŒËß­³I¾vwëafõžÃ¯gÕº†åzƒÙòu>•³Æ‰Y£ýœÃÙ¼ê½ÍæùýUʃ°¡†êø´I˜j-Ë/{-X¨bf¦ICÅc; bÀ=Ùgîì#šáG 8IY=a6éÉv9*?Ö&\Ô`µ›;¡Ð¢züáIÓ0ÝÍ9û9â Þªì*ÿ½TW·ofž×\Û%:T©4oàï&9Ø“P B6X—ÚËç°C c'¼ZŸ<x÷ÝÁ‹k]‚ÐÕµL‚´¯L\ïúc³Çåu^·q0&™>Á†ÛB»„vÓpm§Ïë»I°!™X‹`h†‹‚¥Ó £T+3¨.u«ÖÁkôcèÉ7P¥3Åpf]·0͆(–™v-9¼8§zqåVú–  €‡;Bx0”؉Š÷Wþ€‘ÊßxÒŒ¼ zï å!j„€ºF@¯Å¤þrïIéˆÛä´4²í2ÄùIõ{÷œ%‘Š­à‹) ä*x íÚI÷«P\'ÁÏ‘Š<%‚ X tmîzˆ¨ühÎÚ%´#h hgÐæî®×¢‹‚Ïip/r˜6‡ví¦n•Ã=‹Ÿöz„ž,P¿e5Þo½aÇ+a9´· )Ø´™Ì &S F¸g=$È4#X²ZöaÄB;=æ2ˆ…Fâ] Ä" ž"!oŠörSËcUϨû5r‘ȤDÛåW¸ƒÖößPÉÆ?Àpÿ€ö{h¿F?§òR<H{B%|ü";þ íÐ>†ÖlŽmÏŸ«žúúûêIóÌÍþ¯Lû“ç§ÑÁKªÎLà+K’>?÷ˆ˜ÑœHvQð•ª]Ë#žòŽ‘·¸òœŠk™cSÒ]k¡ 50~+g| S¬šLÁhp;î‰ð–|´ÎvfåÈ:«¬³Yg5‘.æbE­ýªŠÛÖYvTYZˆÂ6‰5 ¡µ–‰)•|X Í™£v7™¦šL8«óï©´†Å^ÈTÙ™A,ê•V•²]A›$´F=Ch×èzÏdp]T¿ªŸ!³&Y‹EACôÒž=k‰d5Ã`‡™ëÔ.}ÈÛKÏjõ²ÍÉ]P!¸-µ§Ë?SŽ¸æµš—€gª2ç-"¹›„2x©J•`‡¡4Š­Ô¾6#~ãýä#*rFéCà]€†ãvõÙÈ¡³<ðØǤ¦Ö:šoÒß<=Ùe$à¿ ²Ë(GHVuÈŸT•¾ñNF¬ I(?z#î>¡wÄ{ìdí&°¬eîÌ©ÙIº—°q(wiR¹Kz{#bu”üMdgyDõb§í5ï‚Ðð“ˆù >5«Ž‚èêÚÐ6{’XÂ8ë‡ÖÎyOÖ!zãúáãúé<ÀIiÚéÒˆs”ÅC–¡Œ•‹èA쪂4v aÍÀß%$NábBîí’àØHÀ]‘€ë"6ê@yý´žÔó z~ R€¾Øxí7Ð>ö4­¦D>.{BÕEøLb =ktxÐá¤ω$‡n’ñ¡Àð+8ø˜¯ I‰„bμ wð(Ñ”¬‡¢Ìó t!ã§æS’z­_¢¨’¸öýÀ1Õ†â<%d-ʶ'†/ƒAÖ¹µ âÞô²¿>Ôn)ƒ ã}¦T1àæë,Ê*u¾PS™,<ÔàíIPk–¬ûNÁãžØ 4ˆ‚Ö`Íêr!D¿!Ê1ž£ ä{ c“–S©S^£ sœ;D¾[2Ãt'ŽÕQf¶Ž!•›„ÄíÖr–ºÉ¬OÐößÂõö`)iÃôR¿¨]/6óÔ}ëx#Ô›™(Ñ,Wù-€ý/ÞÖsàü?ïç'­Æ'“© 1"ŸÒQpIkvÒ“È™–5L©òˆ=\"¡Š)ÑÎh—í•Á"²ZÙàåü(#­ ÚÓ:ÏÅÀ똊ÂýàínDŒ&]#›:I.ªZ½º”E¥·,Ê ÕáÄ%Þ-©â)†(Hg…˜míÄx(xçñ¦t "•Ì: X¡( J Èø„ƒÖ4œ£'áG*rFð!*Ûö‹k$®Â=N!zž¤8z鿤ˆ)ž˜ë-º»«yC%I(Þ Çè^‡nÆ1p ñ‡ä”ñ£{¢pr$ '¿! W6(»>¦ZŠ?'yè„Öd„]€ó™Å‰ ~íjS:;¿¢òÓüÌ„ôöyÍÕ øC[ 8X tgVƒò`}8\(/<³=ÙÊ35lMš_¼ó=?£¡Ç¼žG!.‚/¤‚6()¥EÈ¡R'&{€.8¦Þ‚'¥ft#ÑA¹))ˆ|Š99Ç¢GÙíAuô˜*Ê6%µÊ¼™£([â€ã `œ< ¦Jb›P}ü…¯á(U$3Ü î%UE9Êps£FÞš°ÇêÔ^ZæCQ.­Ùº¼vÊ4çŽÀä¨ÌýaÕâJUÆɳcÊ#'É=BççÅ«sóâF>Ô3þFt|B]†üš¸ÐÛ…«òOWû².7¦ô- jO%8Á± BÙúÃľܺþpÖðD!¯¡qÑÑ1•Ÿ®‘ÓuxT”l‹HSÑßSÕ‘ûå8rña÷ËìyØ!®÷œnU3‘jÐ?®5}Ö—!1åJçä(¹3£Lœî¢ ˆËݯa½¡,Œ9”¾”Ω6d‰RÕÐÞ¢Dö5eMp ”Bæ ™Õ†Œ‘#}…®—ÞŽt¯Øøœ¶Ès¯l!­IT|!$*F6[ÆNôÚfsXŽ½õJ.áqAÉØÈi¡s* ÿº¦Œ U¯ÎO ¤•®fG¸¨´ÙP%.ˆÛ®wAÉ´ÙÞy¤ÿõÚ†ÒÔi?2×eOÎ’¬7µEÄåÔa®VÇŒ©%mUîŒ6¯÷MW¥ÌˆKÌWN”µ«†XM € ÙCŠ„Ó.¬´î[½‘ñ>ð䫪×2©rTk¥•è˜'AzÌ(îT­V& ê¾|ç긔íhו`bW ÆÞý‹« CòÍôT|Iˆ*¾ÈžÔÂ’}©…yÃpÖ3|j6HNQIì­ |ö¹—e$Z=ô± Ì;KþŸ©Œx,¸ö ‹ «ªàËI0ê´N[­CÊTë#ÕúHµ~×TKØÈpkÍf¾T‹ì‹~[ç•´–ÇTOäBÝ“r?¶ò£Dî#AzªÍ¼"F©wb§W§AKWŠ|¸çOóÕÜ}5çÚ‰´¦ÒAg€ËKÀëSc´‹«ª<áÄP[ëéí)˜!ŽKvž¨Êú`MO¡gˆÊØÏÛ³gLÒÖÛ3ìO¼gd³f¡Çpý#-!¡Œ¸¯tòäƽ½–þ í×'>ú@ƿ㓾”õWÎ!âj?ÝÚ”Á~Úpw—Ç-Íž ú½¦¬Ï—ÆüêÈÅ‚e(‚y¶ŸÈ;;8þ'¡;ŸæÞ›æ†ÞgÝÁ˜阾ŽjÌÖNù­†&2\?xØY}lÙ;;«ŸÜ?ì¬]¹ÂÖÄ<é‰r“öÆIk«¬°³D¸¿ð6Ìl\Hx‰ ‹Ì9š7ÈDÑ X‘ ²³•ã©Ë%Û%]òTÙ6zT{Ï’3á¥s_r'=¬Zú®.ñ¤á¦gæîUÉö6¦ÑåVÉæÞ¦2ºçÀPQog;g ;rìUgu õÃyC:Ú,ó†¬ ªJÝzùzuüy–ñQže|gI¾ß¹¶•:­êÚõù™]k§ô.Õ =íÓÑåmåzÅzc,jº· âbòx•h X±ž´¬UâmúÞf îmúh2fA:¦×ÖtUï˜.¹’9çÌE/âÄ£‚ªÚȼæpÚÌÄ'gÌÐÝ%UNѤµ­ËþD‘Ea¥¬$fœÅ¡”rp-¢PóØ¿â÷øæ“ÿlMD endstream endobj 579 0 obj << /Type /Page /Contents 580 0 R /Resources 578 0 R /MediaBox [0 0 612 792] /Parent 634 0 R /Annots [ 576 0 R 577 0 R ] >> endobj 576 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 668.883 51.396 677.323] /A << /S /GoTo /D (noweb.NW2i2LRH-1oKncb-1) >> >> endobj 577 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [148.611 666.246 160.574 679.311] /A << /S /GoTo /D (noweb.NW2i2LRH-1oKncb-1) >> >> endobj 581 0 obj << /D [579 0 R /XYZ 71 791.866 null] >> endobj 42 0 obj << /D [579 0 R /XYZ 72 754.004 null] >> endobj 357 0 obj << /D [579 0 R /XYZ 81.963 669.969 null] >> endobj 582 0 obj << /D [579 0 R /XYZ 72 667.242 null] >> endobj 583 0 obj << /D [579 0 R /XYZ 83.504 669.969 null] >> endobj 584 0 obj << /D [579 0 R /XYZ 83.504 658.014 null] >> endobj 585 0 obj << /D [579 0 R /XYZ 83.504 646.059 null] >> endobj 586 0 obj << /D [579 0 R /XYZ 83.504 634.104 null] >> endobj 587 0 obj << /D [579 0 R /XYZ 83.504 622.148 null] >> endobj 588 0 obj << /D [579 0 R /XYZ 83.504 610.193 null] >> endobj 589 0 obj << /D [579 0 R /XYZ 83.504 598.238 null] >> endobj 590 0 obj << /D [579 0 R /XYZ 83.504 586.283 null] >> endobj 591 0 obj << /D [579 0 R /XYZ 83.504 574.328 null] >> endobj 592 0 obj << /D [579 0 R /XYZ 83.504 562.373 null] >> endobj 593 0 obj << /D [579 0 R /XYZ 83.504 550.417 null] >> endobj 594 0 obj << /D [579 0 R /XYZ 83.504 538.462 null] >> endobj 595 0 obj << /D [579 0 R /XYZ 83.504 526.507 null] >> endobj 596 0 obj << /D [579 0 R /XYZ 83.504 514.552 null] >> endobj 597 0 obj << /D [579 0 R /XYZ 83.504 502.597 null] >> endobj 598 0 obj << /D [579 0 R /XYZ 83.504 490.642 null] >> endobj 599 0 obj << /D [579 0 R /XYZ 83.504 478.686 null] >> endobj 600 0 obj << /D [579 0 R /XYZ 83.504 466.731 null] >> endobj 601 0 obj << /D [579 0 R /XYZ 83.504 454.776 null] >> endobj 602 0 obj << /D [579 0 R /XYZ 83.504 442.821 null] >> endobj 603 0 obj << /D [579 0 R /XYZ 83.504 430.866 null] >> endobj 604 0 obj << /D [579 0 R /XYZ 83.504 418.911 null] >> endobj 605 0 obj << /D [579 0 R /XYZ 83.504 406.955 null] >> endobj 606 0 obj << /D [579 0 R /XYZ 83.504 395 null] >> endobj 607 0 obj << /D [579 0 R /XYZ 83.504 383.045 null] >> endobj 608 0 obj << /D [579 0 R /XYZ 83.504 371.09 null] >> endobj 609 0 obj << /D [579 0 R /XYZ 83.504 359.135 null] >> endobj 610 0 obj << /D [579 0 R /XYZ 83.504 347.18 null] >> endobj 611 0 obj << /D [579 0 R /XYZ 83.504 335.224 null] >> endobj 612 0 obj << /D [579 0 R /XYZ 83.504 323.269 null] >> endobj 613 0 obj << /D [579 0 R /XYZ 83.504 311.314 null] >> endobj 614 0 obj << /D [579 0 R /XYZ 83.504 299.359 null] >> endobj 615 0 obj << /D [579 0 R /XYZ 83.504 287.404 null] >> endobj 616 0 obj << /D [579 0 R /XYZ 83.504 275.449 null] >> endobj 617 0 obj << /D [579 0 R /XYZ 83.504 263.493 null] >> endobj 618 0 obj << /D [579 0 R /XYZ 83.504 251.538 null] >> endobj 619 0 obj << /D [579 0 R /XYZ 83.504 239.583 null] >> endobj 620 0 obj << /D [579 0 R /XYZ 83.504 227.628 null] >> endobj 621 0 obj << /D [579 0 R /XYZ 83.504 215.673 null] >> endobj 622 0 obj << /D [579 0 R /XYZ 83.504 203.718 null] >> endobj 623 0 obj << /D [579 0 R /XYZ 83.504 191.762 null] >> endobj 624 0 obj << /D [579 0 R /XYZ 83.504 179.807 null] >> endobj 625 0 obj << /D [579 0 R /XYZ 83.504 167.852 null] >> endobj 626 0 obj << /D [579 0 R /XYZ 83.504 155.897 null] >> endobj 627 0 obj << /D [579 0 R /XYZ 83.504 143.942 null] >> endobj 628 0 obj << /D [579 0 R /XYZ 83.504 131.987 null] >> endobj 629 0 obj << /D [579 0 R /XYZ 83.504 120.031 null] >> endobj 630 0 obj << /D [579 0 R /XYZ 83.504 108.076 null] >> endobj 631 0 obj << /D [579 0 R /XYZ 83.504 96.121 null] >> endobj 632 0 obj << /D [579 0 R /XYZ 83.504 84.166 null] >> endobj 633 0 obj << /D [579 0 R /XYZ 83.504 72.211 null] >> endobj 578 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F73 223 0 R /F61 150 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 638 0 obj << /Length 4887 /Filter /FlateDecode >> stream xÚí]ëo#7’ÿ>…pÀa5›¨‡æ {Yàr˜{— vã>Ì$»eù±c[^INÖ‡ûã¯Xd·Ø-¶DµiOÛ‚pÚRw‘,VýêÁb‹Œ.Fdôíâÿ¥ÐÚÿ>ª«ÿzóõÉ›wß52…‘LŽNÎGReYŽTÉ RŠÑÉÙèãX°·?ŸügMˆŒ 1*èhiÿt—?|k鈀•g n@ ÿš@à1ߎ>N¨2||ývÂÅøîѽ=ÉÍø«,„J7.3^`{‘@Ô>&Ç?AòL…$ñn,µW{3?$“ý¿dšç‡ •æbЭ˜¾´”eâ¥|BiaD¥-ü%h cd|övR/›w¹Dý«\ rö,˜¡ìü ¯N7Y²nòdÝdɺɟV7ûÒbfü.—q8C[sž3ÊϨ¿AŽ´ Uƒå‚ù¼Ð`o¥ã¿â"±, v´ã/ÐŽ?‚V&¬ÐQ¬è|@‘¨ØöÆ10lᯠ[ø[ŽØòò°…çÁ™[zC ã@u§É‡C‘œ2õ1±×*è¹C<þ¶›/±}ðß&Q£D¿L”¾»Qè^ºA(„ÆA(ÿí%¶þÛTÊ?çâbæ¬Ï¶Ÿr#ºE7mYÅq½8Ke•6IË•LÑTh1ÏÖåù3w ÊüÙûó[ø#Ѳÿ.y @pÿ-´ÖîÙëï°b{‡Ÿ_ŸÌð“y¢öR;kR¿Ìa_õÏŒÿÒ㿈kÆKà¾îÛví=¶îó[lWY:S´5v¦±D¸þÛ)¶klÃÏÝý«\Žë¯É¾X‘,Ë' ¹ÎJ3ÉX£3°ÈÃiŠKIóM’~'•(Gj€t¢`Ä.Q°½«…Îx!½Ãv|â®o°uw®ƒëe'Ѽ„ļ&%¢¹D.ItŒ\äì¸*‘«2Wäl^ZÂ^åc ÜwP$ÙéJª­ð”¡z Toê-Pù*¿Hu¨Í·‰)É0*AŒu&`ä5¥ÁÀ§"¥ø,HÉ_ÎÖæ)û"e÷ÞFoW„N:Œœ¨dLnöE97ëP›Éñ‡DÉ5©ÉWᶠ(Z–J_'Ó—H¿dÚÉÏ8} ÖóSnëù›LßÂêÜ pßgWØ1{îŽuj¢WÜ)Ð6EÈ•MôÂõ{¼þsj*—ØT.ùòªr(uI•6Î}²*ѹrÞÙb…Ÿ(/“snkéÛlyY¶×›ÏÝ錪ڳ\sz}'Ñ*wÜ.}‘'0yJ‹ã‹3VGRoqç¹2´sŽ&¿-3ÿ±ÓÓäÕù}òê¬r¹‰“\û*Ñŵfk0ÿŠ-Ãö¶Èó͇‰ÌÐô-Qc&þ%‡»ˆ<7JT nˈ\¨œ)býgË–žûOþc ÜB¿;pÇøÆ4——3 gsÜqô½í߈Pœ>¥P$î5¹§2¥ÅÍ0ê†4XÝаÏ}¸c—C›Ò·øÉ]2…ßì»VÖ¡Ð!ס°¼u(ô•Ô¡˜ëP°¾¼Rwšã°:Âð4Ëwæê‘ûÒ=”Bmå µ7€Å°Ê\%Ú*—sarÇ™O‰ºdõã öúo6åZä‹ž¤˜¢{P[¹µ•{Pùvxgpv-wætª¨^_TIR¦ª¤ö¡3€C3z G4Ì[}ÿøÛ›:]ÕM=­ª«í²-½U¶¥£e[³œ¥ßøXæÄPî_ˆ<j ÝÀEPt9÷¯NθX½e™kƒì´µAVÖ[cîPOA`¸SbpÙ+(Xí¶çÁ¡ÐYpç:¸gmêþŠÌ…ÌÒnéúhãºräPu}^×h²úóÄbìl¿u{ÃÖC<ÎϦ.ð50·Ák`rn3¿†®ý”…Kÿceë\Ô:pS³•t¯pÊWØ^D^Ïöˆé笧Ÿ¯º‚ùMÜiPÞp±Uü°ÊU|”^)ú‡\Øð¿*,½Œg<PUT¸í÷û Ò"[}ÈÔ¾'Xö°Îe,øÜÿ(LÁÕ²½Ëà dˆ«þË“«Öb ´L)×ÁûpçÙ«frxh÷sþê¯@Ϭ> ¿ÌÈ„Ü© áðÛ:Œ¨Î†ügA4õáH7—ûNiÇ;j+ÞQOpL8„ð$%^°N³zú.) ý›7îk_Þ¿¼íWw•kÍVõÝTÛ,ˆçÎëµuf™ô)vâªwz8ïhœ£sîª g>É"/ª#‘ð<Š¥hPú×"uALY=#¦î–F·¦(¥©îh§ï« 3ra89à SòÑ„Ó‚0÷dYð·Jÿéæîz~3¿]ÏÏÜ¿»ú4¿¾º\,üßßÜßÎÖW‹ÛU=`J C µ´ÉhÂdQ êÈž\Ú³N€ªç‹ëëůW·¶@Áfj§§×þ›ùíýÍ|9]ÏWîïéõµ»°[²³ûå†rýàÏ€Ø-Šp|ö¾ë`|ØU=>üs}9]»'gÓ[÷Ñi=¨û[Ô;lÊ Q7ì©_ ªÃ¹QVÐZ¤ ×wðÿ¿}ÿïÿ Œ¿ÿcqã‡V…0´zhzëy¸¾œ_-Ýår¾º›Ã€™»¿Wóµ»Xœ»/îæË»érz3_Ï—žë”Ë°FÊBׂòá,6Y€]ªn*¢²òþÄ«)3…Òz4»yó7 ,œSé¾ ®Ý—þ!n¿) t ß|ü™ŒÎà`qÁýŠ÷݌ʗا®G?¾ùËI JÌ. p¯[1G z|¢¬ÖnÆýõÉ›ö:j=QC¬9fØe|ÿGäp`3ÊYA4°ZËBuÀx°ß&û)u–¼Ùór~°yõÊMÂGôAyµdþ–r¤ £ˆ_ÓRÄ %Ê$:ª`x°¹ËËoÜ—œ¨ª·¶ÙÁ+,4õdG³7Q!ËÖBü¹oÞ˜#ØG;IÉe[•iJ8ç ¶á®®£Íز`eÙ‡¯ºM†5øú÷Xg lIÝÙy¤³²šEWÑììÍ¡¢‰¯¶«ä0Áô]¬”¥(4SÍÕêZSHSMdh‚•EÉûÈц†–„4ºÖT4$ æ´{¥*äP U”5óÊÍüÛéýjµTúuÛ€BM¤ã«éí î×wSúk ûþ>&¬ Ú˜§näytÒ$é$肋Mþoéä—ñ©iAóÌkR ”ãgd Û``j“Í]-hQd ÷ÜØ!²âýt["¸½KÀBÿ[jTÂòWòÀ`‘©¨µåA…ÝÕ”,¤÷™ï"r Óx`ð–ÒºG˜y¨·•ô=V¤ÀQÑi0®¹–%gûic Bm²|‚yqžÿóÎ mܨ eº„ZTt‰Û$&|Þ:ñ$C#¼ƒ)ˆ¢[ZNÅ.‚ÞV7( ¦YAn¹ÖM-á 2 :¸iÕ7£Ù¶ŽlúÉBwk_2쫪LCîo¨Ù©)PLmB'¥%<ÆÆ,*A\¹(¬ ë-ˆ©Æër·ìÄU7ä¹áÚB ijÉ ,®±"ÆTm¿Ýa¸s`tm·CŒ>É>·£«¢›¨7j/©ÚæE³G‹cU”»Ãô<¹¨Ù–àO?Êôà ºØ= … ¿Ïô Ðå#Èç³K¶Ãô÷ë¶Ãô·;ž¬þq?]ÎÏv˜ÿ~ý7ƒ*jaW’‚Dתí߃Ðê$šÃàá_¥ù.µul™¦M«IQW· •j<[¬.#ÚÉàN43gMõl¢”äÛ*‰L@à ð^Èæ¤[€hÌéiSûà\¼ 8¯&2&ú0¬¹µÛ(¡ÛiR…Ö!ò–qÏRrÕ5ÚbJ ø8AÖˆX, O&È(cÀ¨]¦eC LŒâ—†[匋,Te5b>•R ÀÓ.ù.§ªê²sì5©`ì|‡•ê%V*„²!X©ƒ'µ0d7‰ýV 3qk€`Õ-ü7)“˜2&w¥©7´À¨Ò¼w±>aqÀ ²÷YwhW¢ÚfBØ®¯hsQ²`šÉè¦SœÑĦ¹ Y¶Ýºöƾ ¬™b¦ñ`ycJ¾ˆ,™‡TÇ$4&¡ywl·BhÙ+±,t“•¸-©“ŒÀ}W©f»Dlª¹OFDsÍ-¾xÑaµŒM",åT:‚0ø§n$Dk6ì•SìQøé}‰Pªr¿S×Ïêv8u¡õOwêZÞ91æ -Óͪ´µÞˆšÒ$ºíE€G´±0Û‘#‘2íØ Ý çK—”ípYAƒ¤Îä²¼tQwÓÖF=Îe%RìQ44¬+.Õˆx½¶ÆO›´ê¦~½Æ‹›Â~g×ÓÕê 6›®;*œ$ÃX>]…“&ÝNç1BV3ƒ”,‘ñ ¼Ôª¥/mÑ%›Z»‹no¿ïã•T!ï÷R©ÔB*u,¤ò&¢§rÆë¨ÂµzlÕÁ‹áŽbƒ?ÅÐß=QHºº”ͧ{ïëýò¼·úuØ›.¾:·*&îVà~ƒÐ,¦ÀZn¢õ¢ºdð SÉÎwˇo9Äý}k[;­h¹f›n$̳Õ¢&xu6ræ…¬Šî;ËÉdŠoTƒÅ:¢4ù4gQºFënüè')MëTÁG(®èF†ƒ»Œ©µ"êQÈ ¬S¯÷‚ô=À Y!„ÇÌïW«õÕ¬zu‡hljøÐwAò@ ÌõÛŠ¡öîaIQp¹sÿBNÊLÂÀ’âMuå]X_DK…ßì^5Rv‰…ºKÀš¨M44ÐëL(Ä•¿—Äu?”ƺh1½Ü¢úñ£#i¼ê8¤Ñ !|g6 WȪʇ6ßÑ“þr× endstream endobj 637 0 obj << /Type /Page /Contents 638 0 R /Resources 636 0 R /MediaBox [0 0 612 792] /Parent 634 0 R /Annots [ 635 0 R ] >> endobj 635 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.478 297.217 165.939 309.322] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 639 0 obj << /D [637 0 R /XYZ 125.795 791.866 null] >> endobj 640 0 obj << /D [637 0 R /XYZ 83.504 755.001 null] >> endobj 641 0 obj << /D [637 0 R /XYZ 83.504 743.045 null] >> endobj 642 0 obj << /D [637 0 R /XYZ 83.504 731.09 null] >> endobj 643 0 obj << /D [637 0 R /XYZ 83.504 719.135 null] >> endobj 644 0 obj << /D [637 0 R /XYZ 83.504 707.18 null] >> endobj 645 0 obj << /D [637 0 R /XYZ 83.504 695.225 null] >> endobj 646 0 obj << /D [637 0 R /XYZ 83.504 683.27 null] >> endobj 647 0 obj << /D [637 0 R /XYZ 83.504 671.314 null] >> endobj 648 0 obj << /D [637 0 R /XYZ 83.504 659.359 null] >> endobj 649 0 obj << /D [637 0 R /XYZ 83.504 647.404 null] >> endobj 650 0 obj << /D [637 0 R /XYZ 83.504 635.449 null] >> endobj 651 0 obj << /D [637 0 R /XYZ 83.504 623.494 null] >> endobj 652 0 obj << /D [637 0 R /XYZ 83.504 611.539 null] >> endobj 653 0 obj << /D [637 0 R /XYZ 83.504 599.583 null] >> endobj 654 0 obj << /D [637 0 R /XYZ 83.504 587.628 null] >> endobj 655 0 obj << /D [637 0 R /XYZ 83.504 575.673 null] >> endobj 656 0 obj << /D [637 0 R /XYZ 83.504 563.718 null] >> endobj 657 0 obj << /D [637 0 R /XYZ 83.504 551.763 null] >> endobj 658 0 obj << /D [637 0 R /XYZ 83.504 539.808 null] >> endobj 659 0 obj << /D [637 0 R /XYZ 83.504 527.852 null] >> endobj 660 0 obj << /D [637 0 R /XYZ 83.504 515.897 null] >> endobj 661 0 obj << /D [637 0 R /XYZ 83.504 503.942 null] >> endobj 662 0 obj << /D [637 0 R /XYZ 83.504 491.987 null] >> endobj 663 0 obj << /D [637 0 R /XYZ 83.504 480.032 null] >> endobj 664 0 obj << /D [637 0 R /XYZ 83.504 468.077 null] >> endobj 665 0 obj << /D [637 0 R /XYZ 83.504 456.121 null] >> endobj 666 0 obj << /D [637 0 R /XYZ 83.504 444.166 null] >> endobj 667 0 obj << /D [637 0 R /XYZ 83.504 432.211 null] >> endobj 668 0 obj << /D [637 0 R /XYZ 83.504 420.256 null] >> endobj 669 0 obj << /D [637 0 R /XYZ 83.504 408.301 null] >> endobj 670 0 obj << /D [637 0 R /XYZ 83.504 396.346 null] >> endobj 671 0 obj << /D [637 0 R /XYZ 83.504 384.39 null] >> endobj 672 0 obj << /D [637 0 R /XYZ 83.504 372.435 null] >> endobj 673 0 obj << /D [637 0 R /XYZ 83.504 360.48 null] >> endobj 674 0 obj << /D [637 0 R /XYZ 83.504 348.525 null] >> endobj 675 0 obj << /D [637 0 R /XYZ 83.504 336.57 null] >> endobj 676 0 obj << /D [637 0 R /XYZ 83.504 324.615 null] >> endobj 677 0 obj << /D [637 0 R /XYZ 83.504 312.659 null] >> endobj 46 0 obj << /D [637 0 R /XYZ 72 289.302 null] >> endobj 636 0 obj << /Font << /F57 146 0 R /F55 149 0 R /F73 223 0 R /F61 150 0 R /F59 148 0 R /F18 190 0 R /F136 574 0 R /F70 201 0 R /F74 202 0 R /F33 572 0 R /F133 573 0 R /F79 203 0 R /F80 204 0 R /F77 551 0 R /F81 206 0 R /F83 207 0 R /F85 678 0 R /F82 552 0 R >> /ProcSet [ /PDF /Text ] >> endobj 685 0 obj << /Length 3538 /Filter /FlateDecode >> stream xÚÍ]Û¸ñ}…_ŠSg…¤ø!µ½—¹Ã]¯Àõ²( $9@kkm5¶ä“äÝ,Šþ÷‡”LËZ[¶dBS5Ãùäp†2YLÈäÇbßÜÞ¼þA$JÃD6¹½Ÿ(6Q< “Ûùä}ÀCþjJ !ÁßÊõ&mò»|•7O¯¦LÅÁ›¬y̲Â<ü’ÊVù²,çµH‹¹éüTÜgUVÌ2óø¬YêIoìjBI˜„jìd2e2ä‚ä·KûÉ}¹Z•y±0Mz·²oVyÝ´øV+ÓÙ”uw3fåú./`åea'–÷í·íŠ-–m1ÓÓzËÏ÷—¿6ËõúoÞÞÞüqCaÝdB21QD…Œ³Él}óþ#™ÌáÝÏFI¤Ýi¼‹t”Æ_Á5G]9HýD¨(õ¤ñQ*eÏÃïší<+šo^)ØãdsDùÏ^Ç€ò_£ÕåËøñŒ»wùqÄÝûÀÙ)ÿ œòû|§ü.ÒQÊ…¨ÍÇAœV~À@¯³ª)÷d>LC-ßVUYõ²-Ï[Ϲ«0ž‹yfãp 9¶qøÀÙÚÚ œ­­y@Ù™Ú ”©yБÎÒ\œ›ª¼Ë£$£¬îr5qÔ]¨ëŒŽˆ0†ÒÉ$Œcec¬r‘×M>m5g¯dÀl.£æ»q©9f7^¶†sik8>pv–sgg9>t¥3é tå˹BU½—ÇAôLg€Å&"!¡ˆc7<ä^òkó[«'Ók*˜kêúñ^o/ºÓåÄàI9©þ:|5•TŽMš|¿þÜÎ&]8M¹Èšefá=æÍRsR×(,K†Â(J‚Æü¸å…ˆ³vÏ‹¥&©A ày[Ô-ð4ßffzSš¢,¦ ×é¬,f郮½$,"‹ÛK7@Èç|Ý4Oœ•‹ý4P']þ˜ÃhØ3Pg…-ªè§»lüµ©Ò,3óѺœg{C2˜m«‡´ÙVí'º¸²÷Í¢Jç9pÅN7‚ÓˆÌﶘqž} „²˜ö˜ؘÔ`TbÞ'fƒ Åf–ÐèV°îw°ýB ÎñBµÙ*ÛcÃ4/狨§¡Ð¢0"Áî *¨˜µªð§Ž¿E¥p$Q@Ê´bnäeL½Ò»ð³+ÝÙƒQ´Â Jðø+Ôþ[î û»qT]è—]«‚¹ôl´ðE±Œ4JØ*4PT~€rt(1²Î¸•Ø‰)¶¬O“vBü`ë¹PÅ¢àé^"2_Ô*?c»éìqêc_å_@¢Bâñ3dÔH¯ªÛžä/ÄG~]£‡½No_¢ç~Ïfª4<_Ÿ©'i¦1ÏìKÒ Z?_ô 6¸5¢‚fäL¶‹®•5†MžW†–ˆçyð1ã8Dlø^Š!Óx†ñd,Ã@¡§£…ð4råÔaÊEð;¶ Û×NÿÛq b‚9-qÆfó;²š!«½éð_ûdlüm…íHYÅÔjPÆÊoGseƒ\ÌG+ÅHÈ’)‚òA-€Ö¨‰Ò'{”|{øÂG†Á]ÈÄJš{ ZÁ“— ngoÎìHàqÅeV’a›b[øâJ l羘0Þ$FÉúK‰ÜHæë\R;Ÿög®ÝHåóôE´²§–‡.>‰-²ÆIu\w^¹Ô)¿ÞèÒeÄ/Ä)âéÍ·âI—è4´l3l«vDÔˆvÄô×íLxÛ|UûZœñ.ÕhŠÑBƒ?{‘{òÔO†ëq(¹`Žoý3»ÒÓè |P?G–tùà/y”c Ï&ßíb¤NÊ3Î<žJê'LA=‰ƒ×)u[7 ~•úâu°b\Ið7ú Û§_u€ÕóâWºoŽÛz„a€ ²¹/Çmv ŽóÆnœ#_‰­ff” ?w‚Üù8Ó¡”öNNW8²{Ç/xË,Ïœ"Õû'cÔ8Q[a³Ï^'˜°âZ`ªÕŽäÝñLàâ¸mç8Rå¾ðÄ}a¤ß¦œ+'ÁùÐ…´1²k¼MQ{aš°Zïó0'ƒuÇe‰ýÖ1Ø6I˽aôb¼a…D·NI÷7݈@«Hº°*úà¼5*ú4V ¥?PÛ@ÕcyE¤ëãƒt}ì¤ëë,æ£5{|‰É$1×¾‚÷È-ŸÓn|W¸"ëÖ2ÇàÒ®B)QãóöÌS5èûÉcš»ÀÍ4k}˜íS´EŠÓQˆ¶*¢|Y•DkoóV•Ýä½Ìýf†Œnp\)GS¬+ü¸gþ2ܳ½«÷$O|K‘WN ‘ÛÐÔS4S8AÓÚ)yVN’kæÜZùBœÛ€h×nªö&>¢Y,Ø[‚÷‹d;Ý;Y‹žlÚ‹…âzcÔ¹ ¥@™Ý>;ÛCåÄimÌvÙöp±³_ù„ݦ¯ä‹I_½³A?Ù‹ØnK+lk{2õ¢Vn€—ØÀïìo»ó^‚NË=Ø$Ýíš±yØ/þ™‹žY½TÑ~/Öîa$òð.ïnBƪ›°Ø¬Wí¬µyªòŲ¸Ï+ã$Ýõ_‘v÷ywP“ˤ»Î;úÖg¢Bž$=ú’³¯}2ræõ_râ,WÇ÷}éßæÕG|s®v=øû¹-Aö.ïêäÎ_ñ’,Ã; ß¿/¬ã«/ ëëUߛ˸ÿÅwFC¾vý®ôÿÍÞ÷Gæ‚÷¢cçÀ5Öø?d÷_웾¤píÊ9v·Çk®˜ñ+z ½fD¨Jb_[B¼Gï¦Ë³´ô6¾Ò­…ÃJ“ιsW¾ö¢rØ¢.½¨×« Ò¯ vÕÀÊ©®j`¯2(Èe‰*í4öï«lÂXÏ}x&Ñmu{§ÁÜr«TÏ9÷ŠZ”ôªcÚR{õ¨»Zõ-^é7qõ“ýHïÄæJOgªcY$zÛ‹Îwçüe»sŽî\t—>/ôQÌüÝ¥ÝÜöž;9ù]t¦¾ÄÝge/ö”{õ¯©‘{_ùùܹ'ä¨Ç‹=…SÁi|u7º—Äïo4{Eq,Be*’Q­²yîüO®í&ƽ?8]þÁé.‚¥B…°_w’&‡bXÊÁ{ñóƒXF¢P&}·'Ï÷aªóaÙÁa.»H2ª/uÞªDþ4r¶#БG§ÚÑ8UåA jîØwa×}ãÛÿ¯U‰•ˆ&ÓˆkôƒÉƒ··7ÿ¸‘ endstream endobj 684 0 obj << /Type /Page /Contents 685 0 R /Resources 683 0 R /MediaBox [0 0 612 792] /Parent 634 0 R /Annots [ 679 0 R 680 0 R 681 0 R 682 0 R ] >> endobj 679 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 386.293 51.396 394.733] /A << /S /GoTo /D (noweb.NW2i2LRH-FUd2L-1) >> >> endobj 680 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [146.789 383.656 158.752 396.721] /A << /S /GoTo /D (noweb.NW2i2LRH-FUd2L-1) >> >> endobj 681 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.478 168.699 165.939 180.804] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 682 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [255.686 85.013 272.132 97.118] /A << /S /GoTo /D (noweb.NW2i2LRH-2zwz7b-1) >> >> endobj 686 0 obj << /D [684 0 R /XYZ 71 791.866 null] >> endobj 50 0 obj << /D [684 0 R /XYZ 72 754.004 null] >> endobj 54 0 obj << /D [684 0 R /XYZ 72 504.684 null] >> endobj 358 0 obj << /D [684 0 R /XYZ 81.963 387.379 null] >> endobj 687 0 obj << /D [684 0 R /XYZ 72 384.652 null] >> endobj 688 0 obj << /D [684 0 R /XYZ 83.504 387.379 null] >> endobj 689 0 obj << /D [684 0 R /XYZ 83.504 375.424 null] >> endobj 690 0 obj << /D [684 0 R /XYZ 83.504 363.469 null] >> endobj 691 0 obj << /D [684 0 R /XYZ 83.504 351.514 null] >> endobj 692 0 obj << /D [684 0 R /XYZ 83.504 339.558 null] >> endobj 693 0 obj << /D [684 0 R /XYZ 83.504 327.603 null] >> endobj 694 0 obj << /D [684 0 R /XYZ 83.504 315.648 null] >> endobj 695 0 obj << /D [684 0 R /XYZ 83.504 303.693 null] >> endobj 696 0 obj << /D [684 0 R /XYZ 83.504 291.738 null] >> endobj 697 0 obj << /D [684 0 R /XYZ 83.504 279.783 null] >> endobj 698 0 obj << /D [684 0 R /XYZ 83.504 267.827 null] >> endobj 699 0 obj << /D [684 0 R /XYZ 83.504 255.872 null] >> endobj 700 0 obj << /D [684 0 R /XYZ 83.504 243.917 null] >> endobj 701 0 obj << /D [684 0 R /XYZ 83.504 231.962 null] >> endobj 702 0 obj << /D [684 0 R /XYZ 83.504 220.007 null] >> endobj 703 0 obj << /D [684 0 R /XYZ 83.504 208.052 null] >> endobj 704 0 obj << /D [684 0 R /XYZ 83.504 196.096 null] >> endobj 705 0 obj << /D [684 0 R /XYZ 83.504 184.141 null] >> endobj 706 0 obj << /D [684 0 R /XYZ 83.504 172.186 null] >> endobj 707 0 obj << /D [684 0 R /XYZ 83.504 160.231 null] >> endobj 708 0 obj << /D [684 0 R /XYZ 83.504 148.276 null] >> endobj 709 0 obj << /D [684 0 R /XYZ 83.504 136.321 null] >> endobj 710 0 obj << /D [684 0 R /XYZ 83.504 124.365 null] >> endobj 711 0 obj << /D [684 0 R /XYZ 83.504 112.41 null] >> endobj 712 0 obj << /D [684 0 R /XYZ 83.504 100.455 null] >> endobj 713 0 obj << /D [684 0 R /XYZ 83.504 88.5 null] >> endobj 714 0 obj << /D [684 0 R /XYZ 83.504 76.545 null] >> endobj 683 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F30 566 0 R /F73 223 0 R /F61 150 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 728 0 obj << /Length 2913 /Filter /FlateDecode >> stream xÚå[ëÛ¸ÿž¿Â_ŠhÓµ–oŠ¸¦E$Áµ÷!(÷åÒ;(¶v×X¿êGîœ"ÿ{‡CJ¢dÙ–m­³A‘`V–hj8œùÍ‹&½ûé½Aü_ Ôþ{ìåWÿ|ñæöÅÍ;©{&6Š©Þí]OéXÑÓ‚ÇDÈÞí°÷KÄ’«ßþ£˜È~•ö÷=wñ¯÷vÌÁtlñÛƒ«¾`QŠt‰4«LË^L혱¸†I{¿ô•ÑËöƒ£LŸDS¤w@uô^§HçHÇxßÝ mÍNÂjÜ4‰´&ϾFŸÒØÈ\ ¦2ÍÍ;̓¯p•äâ{°CoÞ)Žˆ­ò?¥óq:È^^iaåËtM²ÕÃlè®­ÜÕût½\ŽÒ©û4=fãÑà Æá €ë^ „}M’XQ–¿‚ªOnP…O*`E#\R+‰K˜¼"NNÐ1vq£¨cuL Ž‰è-P}h;‡6§jQ]d´C-zûÁ©Åhz—-²é ;G{D´¤…öè´Y{„ä'h“1IMì»ÕžŸQ{Þ\^{x‡Úó³Ÿt1JW£Ù4;y“n²ewZÅ8‰•–¥VuŠIx&w"R^Dâx­*_kå·šDÓ¶ æªn™<ŽªÀÄØa~ý­møOà XKbÊuî!J§$6ÄPT)¶Õ("b&ŠÙp4°Jtühë$)W#kˆ?Å‹G £MsI3¦óa‰$T¥­5‹ ëšÿ§Gî#[+uºÏ%`Gu ÂÍá.sxRlÖ3¼ÎŽ+SXh!¸¾«¾dÑ檯D„ ís½š Õ Ð2º¾êSF ä¹Évñ!×—âݨPàNY;'ĈŽöÄ8Æ]™áÝw2)ؚ˷HgH?Ù¸»9y~½jÝÊ9ëã7±c‹¹.üÿwqîŒÛÚ8`~b åa¬yƒ`Ò¯ú’„xH ‘þlº»sÃ05/Â/xª©½ðÀ"ûû½ŒÞ½EöyîçÜä0ðÕ …],—>D—Âwaÿê '4_)³œeH¸BL.AÖ â…$Û¥v¸%Ó”G›¦á&yÃñÎÄ®-^ð^¨s‹ C Õ]GHßã¯×H—x¨æyT–@m5fƒ¦DÒWŒzݼv±ØõÖ:&¿µÄ[Ìß²ÌïKSaëÚç´.ãVD¬¯âylVºÙ“ˆ adGH)iÔ2Ü©a^¼³ îÌðÎôøü¯Áôñ¶@y7騟«ÂNr(„-_¾Á—Oðåë+t9¯ÝŸÿ⓯>GÜÝà0?¬y0ÂîsEG`à"§«lèfà bX“¦ã±0˜MæëöÅ–nTê^¢€…iÎ ¼x°#èÚ%Øò`òiãVíöz»•æZý×îÃ2 z'<¦²&ÛRf^ÊÅQÃò£ŸÏ³?F“4×”ƒÝ!Ða!‘‘X'lûAµ=$K$ÞÙâOyÔ‘ Ôíà´ÖWÜÕ-Ø N¾|Ý¢ Ô´ˆé\©ÝÒV4'¢¿à;ÔYA’"yuTšÆ@¼,°žYÈ$>ÃÏŠÌ_ûÜþsà Ü߹ˬ›×k ›òvÇðô*ÖþÔᨩRÖz>]H%\ùK­…²:°¯·£µrjm†( C°eUR̵•‹hs¡ónÌO†Èõ—9ß_K1^_XP—ïÚ÷^ÿŠ ¾oþtM[ÊÌŠ "³sêó }lÞ•Ò¤¥4“kï «ñXøzƒÔ×Éó›¯`°¬×ºËF!Ü=PšW\\®6Ï°6/}.[䨻kó'•ÞÕI¥÷³â‚E²&ëÝ[Ðç<6¬v™ŸÜa׉ݑívŠ:·¬Í´ìª¬Í%®Êõ˜¬¹·jé 쳫êßÔëcMÊzÚjØStSE¥Gúè—_íjÛ*z­3ßœ8ºÀâ´ýœÆ·g“΀"ãÊŠ§šUÄtwFb˜×½“mË莌‹™¶G4óqó°Ë¸YV«éçàÀjëþ³·j<þÔщqv†÷{v¶}’U[“lß¿2,Jî#e~ B1i|Š2õ)Šoºç'±:=Í—4;Û2"åɾ,2îªê·ž—Å9 D˜‚—Î/ ‰²HNNƒD"$ï9Òã Qu‰œløÚ5¥Ý5$æ[MÿϽ‰]:’N«f¹‚ÖÈ*&6æÉÎB'pÜW@.úôl’^:g5¤íž(ÑM+Ìϵ¯i|8’ ÛÈç±ó=6.[4 Y.Ÿ¦Q¸_éó”zØ>.ËàeõˆyÍðºÆ#uÕÄ7?RGƒÈ½tSÎYûi´}­+­Év?;BéUwá„_ù”ÙÖ¡#§ž7¦ ãÝ¢«çû»2ƒjàyUÔØ«îcÿÁPØlÆ‚Š¦À8Ðß‘>àÁˆMþÔž}íqú*"Ç1«ˆuþc£3ù)Ã;ö 2¯— ­ ﻹ»8hi{šO !m¥ï&w¼T ´;6JgìÐïíbhÉ]Ë¿ÍÏS¢Å̹êtÎV‡žJv¿‚¨þðcy¼³vÈQuúÖðJ¨ª ´×š9^ÿV`v!5žÂqOsGÌï_¡ƒkNñ“'RYm¢þƒ=0#›Ö'[ˆvæa=ª“U“îùÅí‰fg| tðÝô¹¥ìô(Â3Áß,˜®€™]äü£Ô…î}“´¿Ù›O&Ï+ªï+Bî·AÈù{Š äZºø”’ÄÜ`ÿuÓ‚%[†OÖM8n.û£ûú‰6FuÌ)ä±\Çù±¶ZœùööÅÿÝ¢¡ endstream endobj 727 0 obj << /Type /Page /Contents 728 0 R /Resources 726 0 R /MediaBox [0 0 612 792] /Parent 634 0 R /Annots [ 716 0 R 717 0 R 718 0 R 719 0 R 720 0 R 721 0 R 722 0 R 723 0 R 724 0 R 725 0 R ] >> endobj 716 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [298.004 727.603 314.953 739.708] /A << /S /GoTo /D (noweb.NW2i2LRH-4GHEiM-1) >> >> endobj 717 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [277.142 703.693 293.589 715.798] /A << /S /GoTo /D (noweb.NW2i2LRH-1sNM4H-1) >> >> endobj 718 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [340.066 679.783 357.015 691.887] /A << /S /GoTo /D (noweb.NW2i2LRH-10PmPC-1) >> >> endobj 719 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [34.95 640.136 51.396 648.576] /A << /S /GoTo /D (noweb.NW2i2LRH-2zwz7b-1) >> >> endobj 720 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [248.036 637.499 264.482 650.564] /A << /S /GoTo /D (noweb.NW2i2LRH-2zwz7b-1) >> >> endobj 721 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [416.702 637.499 428.665 650.564] /A << /S /GoTo /D (noweb.NW2i2LRH-FUd2L-1) >> >> endobj 722 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [226.152 346.474 243.309 359.125] /A << /S /GoTo /D (subsection.3.1) >> >> endobj 723 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [34.448 293.2 51.396 301.998] /A << /S /GoTo /D (noweb.NW2i2LRH-4GHEiM-1) >> >> endobj 724 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [281.778 290.563 298.726 303.628] /A << /S /GoTo /D (noweb.NW2i2LRH-4GHEiM-1) >> >> endobj 725 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [416.702 290.563 428.665 303.628] /A << /S /GoTo /D (noweb.NW2i2LRH-FUd2L-1) >> >> endobj 729 0 obj << /D [727 0 R /XYZ 125.795 791.866 null] >> endobj 730 0 obj << /D [727 0 R /XYZ 83.504 755.001 null] >> endobj 731 0 obj << /D [727 0 R /XYZ 83.504 743.045 null] >> endobj 732 0 obj << /D [727 0 R /XYZ 83.504 731.09 null] >> endobj 733 0 obj << /D [727 0 R /XYZ 83.504 719.135 null] >> endobj 734 0 obj << /D [727 0 R /XYZ 83.504 707.18 null] >> endobj 735 0 obj << /D [727 0 R /XYZ 83.504 695.225 null] >> endobj 736 0 obj << /D [727 0 R /XYZ 83.504 683.27 null] >> endobj 737 0 obj << /D [727 0 R /XYZ 83.504 671.314 null] >> endobj 715 0 obj << /D [727 0 R /XYZ 81.963 641.222 null] >> endobj 738 0 obj << /D [727 0 R /XYZ 72 638.495 null] >> endobj 739 0 obj << /D [727 0 R /XYZ 83.504 641.222 null] >> endobj 740 0 obj << /D [727 0 R /XYZ 83.504 629.267 null] >> endobj 741 0 obj << /D [727 0 R /XYZ 83.504 617.312 null] >> endobj 742 0 obj << /D [727 0 R /XYZ 83.504 605.357 null] >> endobj 743 0 obj << /D [727 0 R /XYZ 83.504 593.401 null] >> endobj 744 0 obj << /D [727 0 R /XYZ 83.504 581.446 null] >> endobj 745 0 obj << /D [727 0 R /XYZ 83.504 569.491 null] >> endobj 746 0 obj << /D [727 0 R /XYZ 83.504 557.536 null] >> endobj 747 0 obj << /D [727 0 R /XYZ 83.504 545.581 null] >> endobj 748 0 obj << /D [727 0 R /XYZ 83.504 533.626 null] >> endobj 749 0 obj << /D [727 0 R /XYZ 83.504 521.67 null] >> endobj 750 0 obj << /D [727 0 R /XYZ 83.504 509.715 null] >> endobj 751 0 obj << /D [727 0 R /XYZ 83.504 497.76 null] >> endobj 752 0 obj << /D [727 0 R /XYZ 83.504 485.805 null] >> endobj 753 0 obj << /D [727 0 R /XYZ 83.504 473.85 null] >> endobj 58 0 obj << /D [727 0 R /XYZ 72 407.54 null] >> endobj 62 0 obj << /D [727 0 R /XYZ 72 331.8 null] >> endobj 754 0 obj << /D [727 0 R /XYZ 81.963 294.286 null] >> endobj 755 0 obj << /D [727 0 R /XYZ 72 291.559 null] >> endobj 756 0 obj << /D [727 0 R /XYZ 83.504 294.286 null] >> endobj 757 0 obj << /D [727 0 R /XYZ 83.504 282.331 null] >> endobj 758 0 obj << /D [727 0 R /XYZ 83.504 270.376 null] >> endobj 759 0 obj << /D [727 0 R /XYZ 83.504 258.421 null] >> endobj 760 0 obj << /D [727 0 R /XYZ 83.504 246.466 null] >> endobj 761 0 obj << /D [727 0 R /XYZ 83.504 234.511 null] >> endobj 762 0 obj << /D [727 0 R /XYZ 83.504 222.555 null] >> endobj 763 0 obj << /D [727 0 R /XYZ 83.504 210.6 null] >> endobj 764 0 obj << /D [727 0 R /XYZ 83.504 198.645 null] >> endobj 765 0 obj << /D [727 0 R /XYZ 83.504 186.69 null] >> endobj 766 0 obj << /D [727 0 R /XYZ 83.504 174.735 null] >> endobj 767 0 obj << /D [727 0 R /XYZ 83.504 162.78 null] >> endobj 768 0 obj << /D [727 0 R /XYZ 83.504 150.824 null] >> endobj 769 0 obj << /D [727 0 R /XYZ 83.504 138.869 null] >> endobj 770 0 obj << /D [727 0 R /XYZ 83.504 126.914 null] >> endobj 771 0 obj << /D [727 0 R /XYZ 83.504 114.959 null] >> endobj 772 0 obj << /D [727 0 R /XYZ 83.504 103.004 null] >> endobj 773 0 obj << /D [727 0 R /XYZ 83.504 91.049 null] >> endobj 774 0 obj << /D [727 0 R /XYZ 83.504 79.093 null] >> endobj 726 0 obj << /Font << /F57 146 0 R /F55 149 0 R /F73 223 0 R /F61 150 0 R /F70 201 0 R /F74 202 0 R /F59 148 0 R >> /ProcSet [ /PDF /Text ] >> endobj 785 0 obj << /Length 2661 /Filter /FlateDecode >> stream xÚí\Ýo#·¿¿B/Ö—ÓzùMÖu\mF’ô餳„³-W’ë¸Eú·w8äîr?$Q+Ú1Ð Íãrg‡CÎo†3££ëQ1úð¦ð¿¹zsþ­0#Rä¦0dtõy¤èH:'BŽ®f£Ÿ3ž‹œIQÙû_ï矶“íruw6¦Jg߯W÷“k7ð«¿UtË¿@\tn¤ä–ö˜É¼|4¦E®4u jbß=ÿV±?µp“nŠ$ážS^Íxÿ½ciy÷y¾žß}š»>.· ×û0yØl–ÏùÍòËüf¹X­fŽrƒO¢Y.µ,Iïàðœ×.)(é£Å‹œQSNû¥ÐâÜ7Æí_Feï;'4¨tBã$W@aLHn„ð<µä]õõÈu~@É‹€°­eÅÄòlÌDö¹E"'£œ;"® dF?¥ ™l‚íúlÌ‹ìûKl›{¯³?ã7¤” KÓÑÆîÂ/÷<2Tœe_µÙBJ2‹Ù* 9Yd3ƒÎæØZ™(+ èÿ Û Žl±_+˜Ÿäó*»rVcÛ¶MÒ1'ËŸŠÆÁ¢ *­cDiÎ@ ü1ºAþtÒ5Ù%ü2ƒzI<åÀÏž`”ØAhoqúÃÙX“ `Ìe–cû[ ­ð#çØ::ZfÐl Ӥ̾Fš¤ #Š—4 û®ŸGw\žÀ§›q\‡AQÚsÝ:Ë€~5…% ßÜ#WË©ÞµOA%"–eÊe±²ÃQ6¥Å³‹“Œý4úüíãP¢B[QÂÑž`»Æv›Fƒ$àŸUÈ S™m¡úçØ÷ž‰ŠðL*è'BçF¤ƒ =î>0{®û€ö.D =¡§yOH=äè‰:z¼z5ôQ =×0lû, š£½3NTv¨6TÓƒ"E¬¶yÒyÁDÉ“3ΓÀ˜×}…ÀëtÀ«ºÀkõÂ5Ö xEÛéOx¯z# 7„HFì¸)ì¸Ôý ¸ø[xQ’q+jÆnŽ ÞÈœ+ìTÅÓ¦çâ?PQÑ4U°Ä~8:ò|h˜*:âÄÞܺýáØlÂB,p£0HíÛ'lï­=Öî¢`¼Ý­B€~Oè鈜ˆ r"û"'b•Ž1äÐû›“ç„X:ÂÒµ.®çÔà,c‰ù8FI¬š­xûSiNõ˜ ©\îtáYåxV9žIg2ÊTášy/Ñv4¶£ëg ¹MJ3*¼´À³ô„í}¤“'c—$…¿À»ˆö—ã- Sý&X°Wfí€oƒ’.ÒØ8~¼ô°7Ò“aj¤¯,ÚHôDÁâ>ƒõã™p½Ä”ÕäÆ傾™<Í7±ù+þM+5Ý‘¿²S£óW?ž:ös–,¯EÍ…Öu^kº#¯ejÎwåµ¼@ÅÿuZ+MŠe$v]yZ§+¿JÍ }µ0˜V{oåˆÃ »ê© ‡&ñä\\•á8XGÛ–†:ɘ÷¡—>·ÌŸ¶¤§ïØÞa;’O°;¼'ŸðJø4ñ7~ô-c‹Í1dIªÅ\¦rgœ‰ûQ'ÒQÄ{ßîbÀŽ·®Sldlœzª;,üc½<Ç‘Èf­ËLܧE"9|EyPøç‘iGËdpCj¦‹óÙ—6Þ·Mo£}Ûr“²«7G#•ìϾJ¤â66i´@aÍSÅ…§xÒWxÒw>àøu^Ì&í—Ï a0ãN|TÆiÕmäa4ÙŸ’0÷ i4¸½¶0iÄÎLòF¬’D3íTà4#HÕ`¤04fÄk'‘;›Ò<ëtò[DŸ’x,lD¬lújšiâcÇé]¤Qq„~]Fë×S4Íó“¶b¨+.â‹æÊÏVEspj«ð)°uá*GîÎîæ®K,'Õ×‹Û ’yÓS~æÝÇzñ­4.25-?péâÐuP·VPqûäæ^Öÿ"þ•·˜5Yù¬[‡ƒ%Ë2ìÍ1¯ÓÞá#RÁ¸ÎÅ°:žÞò€½;3C†§Õ;½2PÊ—î| FuaVåNë*.?†²²ñÄYy¨òܪEôüwã^^ßy7; ó‚vo9Èû Y¸ðgky¼ 'ÍPÀ‹@Àõ´r ‚r.6A#zƒQÚ쯂’C÷t›²BØ Üáû+l×>à›ä3Ô:u®ìƒh‰ˆ®Þ·Ø>¼ß¥ù oÔ•^{\Jat<ê,}Jˆ:5°ò6©À¶­OÕ D¢Vt‹uÖ ºDž@À…‘1>ÕAèÐzIM2Ùß‘ÓïRÝ[¦ Xß¼ÓáÍí*Hn<ú%ù€r>2ÈœKŸù© •å€”vŸE0GY„*,iptË´té78' žK깺 ‰ …-‚â4ƒºégté4¾ùõ!Å|zÏÁT·ÈßìŠÚì*ò»Šü1.Ó¹´rßxôÁÒ`ô¿hüw õàÿäÌ>wó€Çø *×Tì­æi•B›ŽÔb×Æ•†¾òMíf=ë‘©LÁñ-ŽtXã—±ÕN°Ú}5==ò­ê~Š;µÀ³'¥–×XÞ=ŸYV,GôXY yTMã,uÌ#µê d-ç†X3R,ªßC¸B^Ñ V(fzJ–Z|à´w ÑdE¾Pª|X…¯º"rÊÏL2×]á0VØØSë3">ï’•E•Ë}ÙÏ â}÷µ÷àëª —%½ Æïz*5£ÎÐ þL&®žÎë;ìO˾ÀË•ö©¯½¤~dZ`%#¨ 6 5Ö8.ECI¦hKÕ¨ðÓkT”Ž­ÃŒ.1Ô-®D<:¾ê`Ò sZºòŒ ë,]b¿s$8w–êðª5Æ/š€+ïÙ_H¶±~ñU¹C}Äm¯LÔ…ÆØÓ.ÅQI¶ Z© ; wYjÊÚò㦠|?Qø¦ð =Çk×u ÞeðCTáE:Ȭ‚êÕøÁª³+›TA‘ëÀoSy“T;±) ñlÛOÕÈ*Ì–h^ƪw4O¿2«®Š@ÃK=¯5Ü4¬ñа’‘†¤­év!iÔtó¯é¦æšîß«¦[úÿ%Á)5ݟѼL]s£9(Î¥DåÌšÓBç†ó^[ýþêÍÿÀ ºµ endstream endobj 784 0 obj << /Type /Page /Contents 785 0 R /Resources 783 0 R /MediaBox [0 0 612 792] /Parent 634 0 R /Annots [ 777 0 R 778 0 R 779 0 R 780 0 R 781 0 R 782 0 R ] >> endobj 777 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [34.95 686.288 51.396 694.728] /A << /S /GoTo /D (noweb.NW2i2LRH-1sNM4H-1) >> >> endobj 778 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [258.934 683.651 275.38 696.716] /A << /S /GoTo /D (noweb.NW2i2LRH-1sNM4H-1) >> >> endobj 779 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [416.702 683.651 428.665 696.716] /A << /S /GoTo /D (noweb.NW2i2LRH-FUd2L-1) >> >> endobj 780 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [34.448 464.634 51.396 473.432] /A << /S /GoTo /D (noweb.NW2i2LRH-10PmPC-1) >> >> endobj 781 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [327.836 461.997 344.784 475.062] /A << /S /GoTo /D (noweb.NW2i2LRH-10PmPC-1) >> >> endobj 782 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [416.702 461.997 428.665 475.062] /A << /S /GoTo /D (noweb.NW2i2LRH-FUd2L-1) >> >> endobj 786 0 obj << /D [784 0 R /XYZ 71 791.866 null] >> endobj 66 0 obj << /D [784 0 R /XYZ 72 727.375 null] >> endobj 775 0 obj << /D [784 0 R /XYZ 81.963 687.374 null] >> endobj 787 0 obj << /D [784 0 R /XYZ 72 684.647 null] >> endobj 788 0 obj << /D [784 0 R /XYZ 83.504 687.374 null] >> endobj 789 0 obj << /D [784 0 R /XYZ 83.504 675.419 null] >> endobj 790 0 obj << /D [784 0 R /XYZ 83.504 663.464 null] >> endobj 791 0 obj << /D [784 0 R /XYZ 83.504 651.509 null] >> endobj 792 0 obj << /D [784 0 R /XYZ 83.504 639.554 null] >> endobj 793 0 obj << /D [784 0 R /XYZ 83.504 627.598 null] >> endobj 794 0 obj << /D [784 0 R /XYZ 83.504 615.643 null] >> endobj 795 0 obj << /D [784 0 R /XYZ 83.504 603.688 null] >> endobj 796 0 obj << /D [784 0 R /XYZ 83.504 591.733 null] >> endobj 797 0 obj << /D [784 0 R /XYZ 83.504 579.778 null] >> endobj 798 0 obj << /D [784 0 R /XYZ 83.504 567.823 null] >> endobj 799 0 obj << /D [784 0 R /XYZ 83.504 555.867 null] >> endobj 800 0 obj << /D [784 0 R /XYZ 83.504 543.912 null] >> endobj 801 0 obj << /D [784 0 R /XYZ 83.504 531.957 null] >> endobj 70 0 obj << /D [784 0 R /XYZ 72 505.721 null] >> endobj 776 0 obj << /D [784 0 R /XYZ 81.963 465.72 null] >> endobj 802 0 obj << /D [784 0 R /XYZ 72 462.993 null] >> endobj 803 0 obj << /D [784 0 R /XYZ 83.504 465.72 null] >> endobj 804 0 obj << /D [784 0 R /XYZ 83.504 453.765 null] >> endobj 805 0 obj << /D [784 0 R /XYZ 83.504 441.81 null] >> endobj 806 0 obj << /D [784 0 R /XYZ 83.504 429.855 null] >> endobj 807 0 obj << /D [784 0 R /XYZ 83.504 417.9 null] >> endobj 808 0 obj << /D [784 0 R /XYZ 83.504 405.944 null] >> endobj 809 0 obj << /D [784 0 R /XYZ 83.504 393.989 null] >> endobj 810 0 obj << /D [784 0 R /XYZ 83.504 382.034 null] >> endobj 811 0 obj << /D [784 0 R /XYZ 83.504 370.079 null] >> endobj 812 0 obj << /D [784 0 R /XYZ 83.504 358.124 null] >> endobj 813 0 obj << /D [784 0 R /XYZ 83.504 346.169 null] >> endobj 814 0 obj << /D [784 0 R /XYZ 83.504 334.213 null] >> endobj 815 0 obj << /D [784 0 R /XYZ 83.504 322.258 null] >> endobj 816 0 obj << /D [784 0 R /XYZ 83.504 310.303 null] >> endobj 817 0 obj << /D [784 0 R /XYZ 83.504 298.348 null] >> endobj 818 0 obj << /D [784 0 R /XYZ 83.504 286.393 null] >> endobj 819 0 obj << /D [784 0 R /XYZ 83.504 274.438 null] >> endobj 820 0 obj << /D [784 0 R /XYZ 83.504 262.482 null] >> endobj 821 0 obj << /D [784 0 R /XYZ 83.504 250.527 null] >> endobj 783 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F73 223 0 R /F61 150 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 824 0 obj << /Length 3162 /Filter /FlateDecode >> stream xÚå\Yo¹~ׯè·Hˆ‡!Y<,o°Y`±r°ÇÃÈÛãèZid¯ÿ}>ö5M›ÃÑHöû`4fÉ:¾úªÈ6oÞ6¼ùö„÷Ÿ_¿:ùÓß´o„`^kÙ¼zÓXÙXEŒ+ݼºh~®¯ßv¿^¬îÖ äê¼¼ë^¯V«‹?Ù Å$™fAœq®:‘—ý¼–Oç ? ‹éÛ1·ýh§Xdõ#~âšçä`B¥†AŸºŽ'bÜ0à}N„d$iñ&·Å´““•@á^¨Ó¯Â.rzUÓŽÏŒ:ˆÄzƽÛÝ`"W'së7Œ›QOç3š”:^¿”ãÂcu Ç4éÆ2rzÒTQi.0.k1ãÆŠÃŒwòÍ«“_NÂC×tؼmŒ‘ÌCM¯¯N~ü™7øÇïÞúÏÇvèUƒ‘Ž%\6ÿ>ùgžñrˆ<“P¾Ñ’­'úlBr&±VüÀ„éƒå/9{Yæ¥Ý±û>ß·%aïLcHc¤,nK3éJ»’oéDc0ú•g-&=P¢…Ω÷Un‡Â3Ë1Ý°ûõÛ¬ƒcÖÂOéJ;Aà¢ØU»U ;¯ÐºL)QT»‡~ÈÆjïÉQ°ÙBÓÞ¤®ÿ±¼_°à¾!6Œ‘9cf•àŽ£írŠñàüæ‡ÄÂæî ‹7;’°X»k4” „ÙoþàòŽ¬ëEj¼D]šb°ô‰ó rLVE>žèkÇÇk,DÏj¡Þ;ŸÄ@Ö2Pø²¢¬(‰Å›½V¹À§žÈ@®Ò@¦` ’‚YøV……Œ¢¹’…aÕô¨¬4kÌ£"…‚Ý f¨_Ñui=&ß þ0·tT¶Ÿ-úüH컾•$8\?.*»ì…p†9éÁžAô„~JÕgÑY0‡iÄ­Q¦üa=TLMƒ0ƒèAqñ½Æ¼dA -ÂD2²ÛÌäô•œBCO{È®°Ú~v*áp\øØjY¶ä-ƒi\¶B‰·'rµæB*P! :˜e€ÇPrʾäüæ×ÛÕëMW]¶åà?înn—okËÍÇo'·_Ü­~yXß Uçúz³z{7ùgŒ]_ßoú_ºo—÷÷ëe?âjµ¼¸«Ú»þÛ›«ÕuÿŠã×ïPé²–…ò¶ÞkWøŸ3«Nñ4yjKÝð“:ÝÜt߬>,/–›U÷íL„€ç‘]퇙¤7âw’ cËo‡‰°ÈÝP„'™ÞîÿÊùlÈü $t;ìû||¸/T{§õï¤ö~‘_†Ób"`v™t„Q°»Ù–}óy µ¡×i*ró„éï9I¨±½­Qoª›t¢‰nö—é?qiæÐs©8é£ø2·£¢d>çYsM]t8¸Õ·¼ïöãêò² Öá›Í»>z§-®]÷míÃʼ+NgvæTNøIGX§ e L†y{Úr@idÖÆ¢-ƒ9êy·Ù±ñ‹.K’ƒS¥pržU·fœŽ€…úý:KõzfKfÎgZ„~ÛÍù*ÇßÃÉ4²3ˆ2½ª)u ¨·Løi¼%,ݶèQ]Jç £¨2mPL±Ÿ6„…XE6˜ÁZÙDŽ®B6 Ýìv;[h§O_†ÚI· Òšäòî~Ón)ð\í‚S¾ïóÜoV·/ºnËÇÕÐعxn[…KŒ(†È"¢¸:.] ²°ÄÈg¸…FØIW @•‘Å9a(k¼–‡P¬³`xÏÈš‚Â\[:£äD5cŽ$)ƒ°ƒW/ò‰Ô@¾$æüŠDÚ£¯F|&=ì’ZÚFôÝU°A£Å~€®éL°î¼¤ŒUqýi•i€ $&©ˆBËû>E‘¤eàj{{ú`Òx0 §©Rv<@î_{fEgÙÔw–Íq<@‰p·‰É@9ZÁ£,­·Á-ó©w(ïóu)q߶Õl}Ù=¼H¦Œ ÐÑŽDÿ"ãD"Ö›QiQ½j“[ušq¡K €©8*:*Ü\&b†«é 6‡ÀØL–DÕéE11+ìĘâ ÃA>M¸ÏÌúàE2{˜v/ ÆE_P˜Q´aI™.aÓm—.]]º”}º¤št^Ün¢òîµôõðU×þÞ´I¯4k…M´çóY£;¼‚Héç8ŒŒ·ˆ‚|BÎ¥™¨ò@]¬Îøgȹvz8°6vÚ•¯9ÞLc>œ9„7¯[–9`¸Œ© }yR¾ê!*4[åÞ¾¼Íõå7Dƒ%„7u”ÊÀètŸ± ͛”–…F¡†èn×ÏÄ…·-ñ}þ~ˆP Ûi·3ŽÈ¿f"òeöŠ:¢ ªš„dýiBºr)}õ™U6:ö'×=§5=q¶Æ—Óó—äŽÌ|ßš¢¾õ$ùš§‰ù…Drá2axƒSˆB_[æã*ûÚÒÈÊånx‡ÏÖ$Ui\{=ñuKeôóœi}1·ú (¥.,t8a£óîì´j<"®•çOÁ9ÂÍ‘ªiBÇ)žæ˜ªP±?Û Ur[V„›&”@"w¯ÍÖ¼yTDœO·GÄÑíÔá^D BÉ–¯Nê‡};¯â¸Âû&¶îx]~N(’O E¢ŠÔ—º±=EºŠt‡úQP¤+§±N%Ó¡H×A‘~ (2]d‚V> endobj 825 0 obj << /D [823 0 R /XYZ 125.795 791.866 null] >> endobj 74 0 obj << /D [823 0 R /XYZ 72 754.004 null] >> endobj 78 0 obj << /D [823 0 R /XYZ 72 733.159 null] >> endobj 82 0 obj << /D [823 0 R /XYZ 72 550.051 null] >> endobj 822 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F70 201 0 R /F79 203 0 R /F80 204 0 R /F74 202 0 R /F82 552 0 R /F77 551 0 R /F81 206 0 R /F83 207 0 R /F85 678 0 R /F73 223 0 R >> /ProcSet [ /PDF /Text ] >> endobj 828 0 obj << /Length 4939 /Filter /FlateDecode >> stream xÚí]Ks·¾óWl’‹UÔL4žW9VÅNùª$:%Ê‘(‹‰DÙÅNòßó5æ±3Ìcwf—”ƒŠÔ `Ý_Ýh`ÄîÛØ}}!êŸ_>½øõWÆí¤(ƒr÷ôåΩÓT mvO_ìþòÙÇ›»Wþúô›²¥f'Jémõ§gŠÿé×_9Ñí¢¥7zWt› e릡Û4”ŠªÑçu ½sep"6ªÔŽvº Ö¤]yê4´˜rhz*2=Ù2øv¨¸ÁÅoŸ^|!ñ¦¿“Ò—:X<½)=ºzþæâ/»øã7xjMv÷16}³CKéˆÝëÝŸ.þPI1™µ*•Ç'ðS5'jçeÐÖYn‹ßȱÌd¨…u[µê¯‹¤’¬gA˜P5»º}Q- D$ÔÂ…q>]˜vRÉÂX’“ cJ )éR…Ðï*«×%I¹H¬MIµéj;©ÚÒèžPËGióÙÓWÞ?~T(ç?ûx]ý|þööùõíÝ»«»úƒ··ÕÏ»WõïoÞ|÷úú]õŸï?\ÝÞÝÜýPýïõmµTdK¯ÜN•Nº™µB¢ø£iúçܸRá‰E3{¥ [µvWÀM-§×·¹QøWÙ~ýÏ9MB¢ ‰núRµÊÒ „$!j¥'ÇÃÜ‹n»ë}W÷(“A] ÖCm| *÷¾P¬¢Û4"´\kÚ›ƒÉ®Ô ë'|™ÕBØ)),†vj–¨™P[:Ý¢f/Û‘÷vSMžÊÆTÆJ•Z†ª$5i¥„%"Û7ÓTt)ÈìlPeÕ”——Qz‘Z¨2©¥§6ÜÊý™²Y€®È%>£ç!(QR wJþ/·ún()EÐ¥€HCS’‚KJ½DæM_6¨†é Ê.úÏ(lû”Ï Œ×-ÕPŸè°Ù;À¿çLŠûЩEee°Ž®oQ/r³jDг›.¦¨eÐ;U7y’ÃŽB»Ò QaK¥jœþ1³2ÎE—äà*t+ÃÐð*äw²˜é»ëÝËf-úƒt„¼k¿Gùï†ÃÚÁÔÆ2æèþ0ÿÎél¼O–ÎÉ)`]í¨] ÐÆ­ç”Ú¤mðÜx‰mðôD˜2Ž¦³œq¸qãÀäežÐë¦ñ¹²þÄ ‚ÝØ´ŒŽ>{ „hú:\”¥À–×ÛRŸ%z§,Oyи!Aû}™óûpÖuO¿Ì™¸¹Å€oÀuzßn%\ý"àHb¡di¬ê­7ƉäˆB(Ï‹í¤9#1?s*À¤ßY‹(˯¤"P¨»ÞYãK+ 1‘ŠÔüör”ŠHƹˆ1p:jg<£\ÉEš¾;o—"ÍËœ=TLãäÉ2-² öÀÜÂ@ÐÍ­dZLÚTOG&Êú#Û|Ä™/“ùÌ> MªMçP•óOŽsäÌ4uÓ‡Q7ZMÝÜrÔíJM ¥öeFj_f¤öeFj_ä£þì×ô»¾ ÒS/³hhðH Ñœ*ù7­Æ# VÍTŸ{aFë]ºa†‚ðmà®–Á›v±÷Z¶SehMºÔ½pbS:£ÇcÒ1j9Á ¾WåÀ# w2”+íª(«þöh쯌+ g¶ŽÆ ~úþ0ÿÎ-DAÐ\ccìoÙn1sÛ-Ò¦$iàÔµ1[Ðâ~Ø -IøJ}ì ¬Zèq¶kõ@AK‘ëØnÓ—$5P3ªðfHHVÑ©ËÒ©Dò̦fÂ-QÈòö——Óû3–ò“rÕˆ· O+ÀP‚¾—x«ð†JÔ ¨Q@9ø'€’ø'€ÿÍí'«»–¸Fo×ІqÍÃõ°j‰‡UÏÃú%Öö<ì帇•s.V#Pr°þã|–¾'Ì»Xmñ%ïŽÆéR«pkw±Å¨‹µ÷âb!™i«"8|m Ûöõ³÷±8WùX­á[ºÿ÷±ÇûØ\àßõP4œ¾ËLße¦ï2Ó7#µOݹE˜T¾G‡m÷kdoÃ&ÕŠèZ:ÍŽÚIœ±R09·ßî|“±7˜RÈxã^"!Jßšn\™°^Kø;ÐÐzØOZ¤‹hSõV—‚ÃãKj6ÏœL äÍsR&h™fÏú³Îž „µzÑΛû ＤhÊ ßd¤XéwܦuýŽ’¡4W´ŽÏ•+ƒÕ7KLŽÇÄMô‰¸HHê™”‘eÞŽÏ$ÛµL¤lƒzvÒ2ììnN&lÆJ0Ém “·Vˆ•o­¾èûNŸƒÎ±°Û}ºÛ‰öÛ‰d”sÈH=d¼\‰Œ3¥:ÂËm Q›9ÓRõvsh\¼É&î²MÖ¾YˆÂC‚¦Á½³DtñqÐËø#VBè)Íh;»G§™š‰iËeNÀ­¼× ÈRÿN N‘k7 N`í¬©Yå²Ë•ËnéaÕâàÄÜgY ÓÁÉïófB.;ë8競'*¿güæTnºõ›u£ÇQI° tžýÀ0LVf=7>& 7Y¦Šƒ _7ã|]Qéá#7R:ªÛj8ѧ‡“š¡WYèú/:˪áIJ!ó&­}zÆpiáÚr˜õåÎ Yr²ôauydé-HÓ¬9ý‹_ßÓêtLÚ®fÒB€Ÿ¸Eè4Ϥ›Îî†LÚg 5×=™(æ4­‚4C9}Œovû®íJñ–¦:ËÑ ækŸh7GAÖF }o¬s½™XbÖmùÉ ƒ®”ŠÖ Tj<£ö¼…nxÀÊ|¦x&¤ÓhšT=Ë> –PxvpÖ>É´K1v“uRÜóÒìUÉ –†â9I­\FèPuƾ#Ââ\¹îSXÖeîL©t«V 1"ý,Ë,¶¬p°eÄ8\y“u"Xù°,ëÓÜEšˆJ=3½ÑŠQ…õ–µiT–…ÙRY½nÁ½ºB'ñR%)àmˆ—*9eïëV¥lˆ«ö ÿ*ßòÅ(E·Ý YP=·!òU»&ú¬NÓì)T<†vs'Ù¹ì5˜f¼PzÝÙž¦/¾¦-¸¾ juõo_`•÷Lä•-`ö (²/¯%Á©Ù&8Íl¥e`*r|­à[W:VÄÛV´”@ªp mÃrm¶bÐùå ŒQ¨xqô†zÈàÚÞöõSÕöæ¨<挜œ5o2$kåcš-Ì‚¼ˆg^Ïd7Y¿£ãÑü†qØAÐÍvQøõvQN›1×VÆPm [lûØâ*6î¦RæD¾®0ûs +ù E9·äôn€ìÝ”1´_»ÃNLþ¥Åj|£3ì–6š¸.×Î/ÂZ‚L‚“Øplâs;š#Å‹¶Å(¹&ÈżÇô…“ãÓ9€y½ò’ º+/÷;ɉ¯±+ö“s‰ÎÅTÖÛ¹:FËɦ„¤ÉéXÆXŸ¤ª®xöH>¬‡Q«ùvu^0wÁ€†™‰£‡òZô†ùoÆæ9¸àšL„’mUñUæh'~ãP¤s½íev“XabÌ3€ÐrÉÁ¬ôÀUâ<$&Ð5¬Õ”jK“7[šüQÙˆƒ/3øbÌàõÌýƒ|W †ÞÂàÛ¾Îcñ6ÀLܜœ†Þ“Ú"§¥@9—ÜÀÉí³&Oz(ŽF E=oò€ ¸Ãq°É9Ã"ô 2x¬É×Γ/{÷‹ÎBΟqœÅ‹=Ó\¹ù5dšvÓ´gIûZB¹‰Y¶} Ìòtµ|Y9WÕf4M—ÍÄPó:wy0Gw#Š»q¹KÒ|eàÁ#ÎÇ•sS2 ¦Ô˜²™p r¦îœd0‘{ˆ›3g¯Õ­¯¾X²É B6å@Úðw2–×v¡­»Ò6åÂ+ë‹Ø’ýÅÜßÃB–¾¹¿ãã«ëw×¹qIÆôÎúÈ¿)»TãóÌzK çÈ·DHá:¤ûÅ©¼R¬Bšún=)N"ÒÇDOtz¹ôðNèÎä=ªÆwu;©çæçùR“ÞüFßu!ùŒ¬³´›:äJmj(ðw¹Ú.DAÐ<âúÈ•%}m_|]é “f|¶ä½,´“] ?óïe ¦´p<¬@ûôµ6V õ¢ß%`OÓü{P€Úñž™%1󔶫‘§H߃âsïA¤­µ‹ÆOãik¥1uŽb7PŶ¯ž*>É­H’„Z÷…€ÊƒH¹*g«ÜÉ°x@HJBFÛ¨]Ûק„€ÄÇ…ÛIL[‹€Å ă(†-°íêÌH÷å6ªØôÕWÅXuмڋÉSæŸê;×´ˆciÍF(XZ/ëÄ’w®‰¨ ׃=Þ°Lmøµ-Þ·¦Gj +k»ZceÃ÷­¹þûÖbq&RUpw„/sWÌrÆK/¾â};É૾¢ô ºðäâ«¥¦o҄䥟â–:{M¿@­ö07ªŠÝ$§ìÖk`ÓÕ" TC Lœ•éhàâГišÞb(tÌ©õª!“U6.œd•Óë§)f¥¶Ð⦯¥ÅÀføªT‹§@-·O깸Ûn¢æuWǪyºCèxûv¯æM…çÝÕó¼®&„z÷þ®®½~wóϪŠ4çÚƒ—tZÇú‹ñ½IÞ(óƒ´ç\ÁhºÉÍA°ŠwµY=yJ–·nT¢{ý’lØÑ:„àÕ¦t5ó1¶H<`ïU#ŸgŸa…0Ìb„œKØïLçëíxÖ…5{.OnhÕ¹®¸ÎJÈagÃ÷Z>&¡ôÜýŠ¦H(oôóK„•!Àj˜‘­ä›·\Œ¯ôâ{v6“k±¥\‹ü‹üârG²ðI^®i!+B…XÅ4k+žW~ÚXœEý=cßø¡R6p~PÙFv3 1¦q›èÅܤòŠ‘„2|yrñ&Õô×Q^ü茚+¬d,#ÇôÈ|Û9Ñä*IÞå>±š1;&­³€†¨œåÓ)!ýbBÕæBñFN>DÓÙÈC$±©‰'W:1áZÜ'^åæã PÒ(ÕxrF£[¾ê“(ÚϪäLÓ—‚wï]~ŒI¾$FÌ `¢N(!–2ô`Õž½¶$>a«øMÌmqhd[^äÓÝF ïhý²H÷ .‚îâ=óÒcMä$îkS¦g[lúr>^ÇÞ.Řauí½[P›êÃBÞ¢xoÀÍZŸÔä·{ ž¶ZgQu_2¨¤6:e ü²ç%‹¸À ¿ÓYWñXƒâÈxÒþ&Ã*d׺z;GÇ”¾Ï¦PùØúÜÞ·õ~‘Pø¶¿I7CŽ·@ô¨PBr´€“ÀÛ¨“f4Šm¡NM_³êÄoѶv#} \kïVéSBª‚å"|ò›š÷;/%UÅ(©’†¯­0Û*©Q¿”TÑ2RegIÕ‚‡XNªÆb”TÑ HÕl‘Û눗Q*f$¥C¨aåJP÷ee<8N©ä&-ÿJ%bUoùW·aà]õ>£ÊP)¾ûÁtU}”I¹MjÍÃ"áf‰Tc3ŠAiòݹ6Vñ/±?¬Ä~ ‘Zk5ǼE!Ò(êA±‚F)«©±F19´®€¡îKU¯µ"LAϯá#âÇÄš ÖðX#ÊШ®ìz×Ï/#P¹T”Vq‹f4æù ÄòÍú#ÒãOè‘—ña7Ñ£¦¯9=rüÊ,½9Šl%ô ³üå}r.¥OýþτÈeLd7ð-xs›<ÖÏ( àö|™¡ZÀŸl6-E£µõx—?ýl=e"ç& gD >Hм@`:JÙédjõÆ®@ÆêÂ؉ºòh¦ żµp„ ¹ONò®´–º/vþNŽ¿I‡ƒ cˆjÖVß4"ªƒoíñ #¶ƒ¸±îâBì¾Å$¾¾ˆçÁ)ðYËê\Sµ•zð¨ÿ”ªçá endstream endobj 827 0 obj << /Type /Page /Contents 828 0 R /Resources 826 0 R /MediaBox [0 0 612 792] /Parent 830 0 R >> endobj 829 0 obj << /D [827 0 R /XYZ 71 791.866 null] >> endobj 826 0 obj << /Font << /F57 146 0 R /F70 201 0 R /F79 203 0 R /F74 202 0 R /F83 207 0 R /F82 552 0 R /F81 206 0 R /F73 223 0 R /F80 204 0 R /F85 678 0 R /F77 551 0 R >> /ProcSet [ /PDF /Text ] >> endobj 833 0 obj << /Length 3510 /Filter /FlateDecode >> stream xÚí\M·½ëW´“‹ch²øÀ€a àƒ BÄöa-m¬ä•íUl$¿>¯z¦gš6ÉÙîÕ®€\¤Á —MV½zõÁbËáÇAy&ÿùòÙg_Y?()¢Œjxù¯ÁÓàÒØáåëáÛçWw/väÃó߯߾Ýš¾yÿæzÿáîúÕ»Û×ûϯ¯½ùíêýÍo×/¾ùõg_y9Ÿ\É(b0ÃŽ‚pÆíŸðÅ~$–áEôrè†ZÄÃ:Ž™Íæ„7aØF½½-?TDå°ßqÌ?yȳ?¿|öË3…ï0rPA«òX]PaxõÓ³o¿—Ãküøõ …Ñnø}úÓ $ññíð·gÝË/{ZÀÖT\ŒÂØâw’\a›ŠD° ûTV'õq>'‘ðDƒÚ›ý°ÏÃÌl2Ò‘”…hÍL´éfÉᣆž•PN×6«”°Ae›Mž8Í¥Œ°“f{ÅÏ%à ¯/B„ãX •%•9a•v ‡i¯ ;¶ÂX3CÉÕ᱿DïÝ~Lô˜I¨xô§…‰¼0:}QšŠåñg“eŠ¥.{AÚÖ¼°ÀB§€¡yy&`UBTQûTÀ%¹à±`Ÿˆ( àE } Dçú˜µÛQ»‚€¶ÜQÚ5½^¨œU4¦ó=ªŒœ¨Æ¡øYH½çЉkÚ:²çE„³ã°hC:,÷%Ϋɗ,­Év¬ÉÅ"†’EÙIŒ¶ä’a™–xQÄZ2Óâ¡%*­? "¸ß ¤×¹÷"’ŸvYñª¾âU lÍHÛ ‚»¯¨¿›àa³iàøÌhè( ]ð¾A ™:ßäÙ;- ØÆÿAO‹4¨Œ>´ÞÁO×M`ïökQÄ4Ìuüªe¦{ȶ\ÜÍaÏ#Vb Å ™¡÷EËrÓC¥+=N 5KÉÐgÁ·ÌÑÂélÂ<^3ãI Ú…V¼¦a…’ö"Óep¥Æ‹¡4—ˆúó¢%ÁÊ›”bV(#ÝB5íÉ‚»ýƒ‡„ÜL“§6‡Ø”CÕIê$|À­m¢NB,¯±± ԹĪ©{¡¡*3˜ò*~DÁby~.Ï‚2nD(7O‚JºpXveibôY{ñª*´–Béƒe™ûfB©ÂàJàH$*ypœ,r¶r –·3°niFÄXI*b³±_Ì âù€øÁi$‡HëòU®êUDôˆ!ŽàCØ$Ó„Ÿõ>¦™æ.*=i˜@'ÌA\Š - Étl"¬šö4ö«´šp¬u)q!¤@À¬ƒ^Î…Ç™P‡i]*|œ Q¥sÞIUn„™iá§Ðã›rPÊ!ºœvjå®È%xb<:ÿ° (é˜ÿ§eò2aH8ª²9 ñ?-KÉÒdŒIÿÈkßébvFl9cZ–ÓZgÇÎà8MU‡o€ã)˸€îOØ9HØ_µ™JE ïëÞ^q_ÂÎ'E&ä¼×c‹›'/tõx‚㨣x=Øžvþ_aÛ¾‚3.ä©ë}…$hrÈ'_²œ(âbAZlìŸ&ë6ObíYb£#g.”#²GZ}\S‰Ìˆ²EÛyjd„'7 ŠYh9/ÉHÄŠˆÉÕr1Ô vPˆC’Í%¦l…”vZ ¼ŠH¼Ç…HdQµ£(£êTÚŽt%!!{þeGV8eG§¢ÓRR¥¾õM1b+8˜sªâlé4êÉANÙ:⨅¸ÂšSÄùÖAá8–+m,p‚êM Fs¥žÀ°àçÚ,«"¥›ªŸƒH!MfZMÓÏ…ÑÏQxžŒ£Ó†S#¤J ?3¸/wž²¡(ô9”ï­—ãñK–üxPõ½N‰É€3g]î{³ôÛ•â̧„·INú”ª|Èãl’~ˆÀ9v)¿¸0³Z¡õˆOfþÀe²§0Pj,¯u Uã”ÔUãS5¥\?\ ¾ ÀÊôk ø’ ð§û4ŠË!)¸ýC¸©oá—²É*u$ËF0Ŷhš%á&Ÿ±kŸ·ýR®»)U”¬IèæÇZa,Ýv€Ï"IŒ¶z®e$’Ä5L™–=Ɇ±~·¦CüÀeO·x€=•=-ਵަì D#ÖÖÝeOê+{Ú<¶O‡9ßLb'_ÚÞl+‰fâ×.xªfÁôHr­A«ÍºŠçq.yªl^ñ\ë%}åxýÈ þ!2ìæL ¾o¬üPøÍÊP£7vhœæ*CcM½s¯BÇêjçc€:'$veI)N©~ÒǧÉF¥”Ü ¡¨+„¢ÇÕ6Ÿårzî•ÎÐç\Û=H9–"Ö»å\N«kœ›BÕõÈüAΛiåO¼¾IËÏ9Õ7ãŠú&çA,–4«ê›\pÁŽ¬ÔðvÔ_ß4—¦¼gõMw¯úæܱºY*NbZä I¯P©é4:AÒu(’»CÏú9MÆUÓ‚&S›GÇ$]Ó1é=CÊ;ƒ[é¹®HÏ“rW™|aÊäâüWýÞÙËzW1 å°Vƒ$Ùzµº. gÊZ)×%cµ.¹+·f’ñÜÍm• mñ¼SÙÒo\¶|PzD]”¹¹ÀÚ:¯ˆ(äP^Gl…C—Çpi\³´M|rÍÒ¶îÐe5ËÝv5Ë‚/7’@RÔƒ½Žš%ÈÜ~ÍŠ–Ü¥¸h‰H#ÄõEË44[ê‡5]æA18˜”V¯ë‡åÂ¥l 9@¶§pÙS· ðIÝrÞ‚“™Š?ø¡8Õ.°•D~{[ññ¬í_g†0š‹W§~ñ.I^å÷¾j1Ëw b¢s«ñ¾j5 ­é‡6¢ý[#¬©Ç^ mUªÿ†áPåú+ì€ wÀž>„ ¿¿¹yõfO—¯®n÷~˜îéÞüôóÛì”®·uß¿+ªC#ËC¦Èóo=ú-]Þ-¯ Ñ–0ä×ÝÒ%NS¹Åq’ézm±yIW—.é6‚ñÚ-²Æ=ÄéYíä iœæ>½SfüCÑ%¾Ô‘·¯7#´¸þžeÍIÞw{»RcLÀºmÂ<[7öÀyìˆ/ezn Ò%·[W·QNkQ¾}ƒ±‚‡ºn0ªa~Uê‚ÛGkn0ÖG;®[¨<4o0ÂAïT˜bÛ+Œªæ*´æ&È ±–Ÿ4öï»›Û‹2ÆóÙ?jì*¤V•yeø²%¯s;wËŸq ÜÒ}qк…æ«®^Êñʈ–ÈÃ[y Mj6Š¼øÿKh]B£ñÂ…íÒD3è梟RªÐv«KRÁ¸‚§Ïô:ÞëÄ¿þtYÉÈR Ì/C`’‘åH9Ž¬™Æ–‚åÄk N*Ý#g…¤Ÿê•V¨"PùY¯‚ÎÞÛ€_ÆüV§+d( îBÔ8Uü\bfDx죰c¨~ˆ&]–ϦÜxœ‹µoÂòɆå£×%.;"¨&.†%ã’¸L– ̵y[¾ÕË^xöI‘@hçÞÏèƒítvvW¥8½äcL'}o:é»ÒɺÅE©ïb¶v:i¸þ(e¶Ëê01½3Ë/8à·‘h°¹¢û2[ÆWÜi¾T´q÷IïÏO9’äHÍfe)î}ˆ7͹xv~žqþ‰²²~E]°‰}Áú5o#°RÜ:úšæòSËôe42¥Ø%º6}3Þv]]‘¾\ƒ¾Üvôe.£¯©†e‼Ë6È 0=Ž/Yü&Þ£’”¢V“*â¼ó¬r}±Îöëb90!©zDLÜSQ/rã©lU÷—pF®Æ B´Ó0W]ë"â—Ø$*/Y³4"àwÂtäê-ð|îTEÓ4!ÀÑ÷êB6©5ر*=·ö³lÐŽ_©Žª@G `j3rƒœ*«EÞEò¥”¡C\É»‡¹$‡ !¡Ä»\Ò=šB^Ù¸S<ÍÅš 'EUz‚K‘ãYXe!‹(5ª“ñoPZ×>Œ­ =æXÙÔ«+qdûûC8EÜ7d6ÍÈ' &×)’Ì­žña^µÊ/ľÕCˆ•ß_§5b¾ÃoÿxáÍóCiÿêíÝ»ý§Ûëë×eí9ðÀ¾[ÄÅ®œ©È¯‚‰;ø&N—¶Ðßq®ÇÔ_~: "]Ë€ÇÇõê‰ ‚B¢”f¶;Y}dÓùÖIZýÖIÆòöù™MÖí~V=¦òùœtî\Û—+á.ü;_SèÑ´GlYož ûj+ZZ÷VÆ»¡æ*²JÚ¼Ti¢g%cbv”äv¾d3ûç ýb§¤”Ïÿþ"Òó«_o®Þß¼»½:¼´ûË«ÿ\ß•ž‰åÃöÍ”vσŠ¥PQh ãµÕ7åŽGìE xßIez'õìòÕÔ€˜;~òG]Mï"¿y—¿”ü.}…yòRó…Nˆd#Ê”K”¬qÝn©µÛd¦?Ö2½Ð`Õ°ì$L/Ú¿ù8†õ?Fizƒ endstream endobj 832 0 obj << /Type /Page /Contents 833 0 R /Resources 831 0 R /MediaBox [0 0 612 792] /Parent 830 0 R >> endobj 834 0 obj << /D [832 0 R /XYZ 125.795 791.866 null] >> endobj 86 0 obj << /D [832 0 R /XYZ 72 137.809 null] >> endobj 831 0 obj << /Font << /F57 146 0 R /F70 201 0 R /F79 203 0 R /F74 202 0 R /F81 206 0 R /F83 207 0 R /F77 551 0 R /F80 204 0 R /F73 223 0 R /F59 148 0 R >> /ProcSet [ /PDF /Text ] >> endobj 837 0 obj << /Length 3495 /Filter /FlateDecode >> stream xÚå\I“·¾Ï¯è[¨’ˆ`_’JÊN*N•Ë—”uòr`† 9æp$ëßçzaDo$G²œ‹¦IB¯·~ï=´xWÐâŸ7´úû··7üÆЂQâ¨cÅÛ»‚sC”a…aŠ8S¼]?Îî_ýüö[ tíš«@%Œø‰*Z Š¨IÂ¥<b*OI Ñ¢Ä^Í“³¿”c•i†- ükuù¸vó·7¿Ü0Áë n,±Ra –#‹Û‡›¦Å?~[P"….>„¡X-ž6Å÷7ÿÊò£¦E-‘RÔ“¢™Ÿh½B¦Dkf '[Ch1W W1v›ã™%L+ÿ笟qŽÐ0ù0èuŽ]xÔ†ÕC6Û*Ξ 2& ’禭 Õ-}èd$Šê^Âcö$ÚµäD;ÙVÎEÍxZ ªÒLücvNÒæ&•è±ã®˜·†ebE¼BèrkÜ<ûJgy='–{ß*@“Å\bí°MËûªrÍ1 Z eš·†õY¬x!‹Uƒ«­ƒ•™«X¬6,4£,–OµX—X™²®4Y‰=ÑEo|”2ªqʨF*£©Œ<¯Œø`¬)æœÕvŠ2fµQû<Úȉdz¼#»DHâª2j¼¸ÈÊÈ2¢½¿àŒÐc˜b2Ç08؈a/Åú˜`Ìúü€y36Mi+ª03FœòÆ †pîÕÌ{PGŒ«8"‰y5g”ÒÙ÷‡çåj{øÃ+ÇgOˆÆÎåŸïÖÿ]mÖ÷»Ý2§@p  ªÐüö~õj.¤˜mŽÿ Ÿåì~ñTþpø°+î?>®ö‹ýâauXí«_=÷ÿ½;Ü—ÿi¿zܯž0­Õ²üy½­ˆ”¯‘³Íî]ùÍr÷°h~Ý•?®¶OÏûj>»§õaý~}øèÅò'¿?wï=+¾š\ËÙrõn¿Zy&h1ÛÝ•_Þá+¼#ïÆ<ðÐG±s“ã<9k”v±]–t«·ŠÙÓíb³ÊÚ4ÀdD䕽ãžCå…Žür:+uÑõaí9í§ñ°ZdÁ—òJšqÑUÑ^¼¢w»}ŽGÜR<âôËÌþš#ŠÁV5º5˜hÅ81{¿Ø¯ÛÛU†–N½KâÎÅ2‰£¬”ŽhLIA¸²½Q>©$̶ßÙÐÒ…c©*D6øyt_çÅ E47ÕF‹Ëõ"@ÈÖ Ÿ¸æ%·ûÄ Žh}–8¹'ɺm¡Dp&s¡kªù<æ]­—Ï”|«Ò^›D7Ù€ýÿd×IoHÜåfßlyìô½ßÌ:}d`J5²ÿ¡"b{2Gb´ræЗHúÒó×ÿ± ÿyŽñ–PÚCYn Hæ\”9ùØaµiãÔQL¾ËXJu¢ð(ÈIÜF\JË‚ŸBk͉O4Lƒ­=FïuЊ~l $¨…jY}¾öÑ\¨³ v¬ Ü ²@L×G _îB~(cîËÂ×3¼…… ÛxgÂdÆm‡Év;¦×Ç 8ÊD($f½À¬^¢¤ P$a½`îèo9¨²˜õÑ»€à¼ÄÉ8WMr~x6åm€ÈVÌz“ÕA$šfšØÚ¬0kàPÁ€íÇ„¼Fý± 1Ôa c"˜¸p渇Κ©ñaåõ(”˜ÈÇ£xm–À…`ÅÄx½KRÈ FfŠ]$D‘9°“ó´Ñ±×K&gQP¯v«½·ë3Û†—!Ž^MN,+§d!Ó /7øÛ»/QÞÃdañR åã«=lö\61—ñ“…MÂIÖ · Rò¹¼·6`×7c¨¿ nº“÷ ©˜÷Áe².·‰˜KWK´1Òq ¾€e2¤î`Y“VåAß-7‹ÛU;úúñq¿ûuý°8¬wÛ¾¨réß0͹ä¶ÇÃÃnY=ëüpXoßùÈJö»ÛÕIÊÿÙÍ>Tö+Ÿñø§åj¿~7¿ ¾x~¬ˆïê—¬÷Ë’ÚnÑƳ’ÇwÇiÖ)ÑV"$+C]T™_IÕüñ´_ýò¼ÞWžvõ€õ¯aÂxlÑ&>õR³¯7›ò'ŸŸù¿‹ÍæMùôüäY‘s£^³]IöyTaŽym1‚·ô¢ Ø5• É]v´`µ´kîïCŽQ.Ô»æaD˜¤¾Ók_þYo³ø~·œÇÀ1=G8Wņox@O”AÙáB_ åO^Jr¿Ú[¹õàÍé÷Þæ½Ô^"'ªÏ½#S„O¼ÓI–âÁ #A+J æz%§¢2#EÅ ³MˆçY}õ#­¡¶7‚®åcx 4&T‹=-YŒ` ÀƒH¤õ±¨Ð‘K9œ·×;ï¦u†¥»1™pÄbA£ÖZj»×ZÓJ×ê{^X–hÅË´/ío9"€¤[ ®y‡§Oxù‚6.)à–/ì_ÃÆ=Ò6'f‰u2'•ÄøK©¼h9 ÓFT.ê;ïóÑ«‡ù| G@¹c˜I Áº\ÖÄÀLÙÖ„¾|\žÕ¤Žeà›Ô:HÛ¸ÿ›”¸òLî!(Þ@+oýUŽ{ø›)±2ƒÜ™BÚ®r;Äkt•÷Bw¦ p¸åÄ43Ïbî“¥îÊ ˆN¤éû°D‡×M5ÞSäuæU¶ãŒì„"µÊ_ûL°G¢,-Ã\ ¥Xo-©&îêÚ:;­J¤VuºJ¡y¿ G¬g%ü¢­ˆ‰>™2Ýp:Õý(™¦>ás}•2לGk#Òˆ(¼~ÛÞüÐÝ×3b1ÆøJñP¹ŠÃÏó–ä¥äX¶ðÙª¶„ædF@1)¥ðYõEe(åñ£B/»·[¥RrÄŽSË°'¯¿½K«¥¼²^Âp¬‘½ìlf«n £þ¬§”CÍìšÖ‰RžÆG^îõ]S£ùü5»˜`UPžË—äRÅL¤TãÀ¹ÂL)Ú½hÄ®}KS[èšx`ÍÙ…Ä¡-à§5N¸= $Ú…Äz‰ÞA‰«ô ªƒ‹ÇQ«óñëø˜–ÇȲOË~/X=µŸ¤u”Lçʇe²GŒ°É©¾CÙÂMiŠŒ¨£ªî:*ÓÄŠQž‡ To%ÕÍfÁõP'~וT3õ þù•TdWJ³+VRMoÿWÇ© à·¯äwpöŠ©‰1§Glp±ÓÂÌÃM|>"#Þæðï—¹#…Ýxò]æ¿kÅ Ûþnv­œ\g½Í`LŸ/½Æ>’"k"µÌö"Q6ˆš‘@´×øýÁ'L†!Ar¶÷Ø–áôD­‰i_„´¢Áç÷'¡ý·WG4¾ûn¥™[λœfÚ½/Üj×C®ÖÍ\Zc¹$“gìs¾°)fˆ0xÿ—0¢Õ^ó.áÿ{#’v“zí<öþÉ…Òw4\ØGÌj„(‰­²ýýùáyÎaU×xìÞ­ŸëÛs/õàŽG—zøÏáRÿ°ÛV#NN›9DÝäá¿ gÁ\}FÎÓö7yø‡ò&Ï’Ê©ÙÄ6“áO‡Åv¹ØWÔ–«÷ëÅ/“^œÈ¶·Ä9ûÞëM@úû35-mˆ¶f ‹Å©k÷_Ô)Ðc®úí‹ý…dâä*)µ×'ªÒ$×,ôNìXÑv_t ‡«‡+© G˯!ØšV{áò¥k†äŠ5é¯ô»\°5©É6W”ú@Ä ¹64?KõfÑ(ÌïrêÍï endstream endobj 836 0 obj << /Type /Page /Contents 837 0 R /Resources 835 0 R /MediaBox [0 0 612 792] /Parent 830 0 R >> endobj 838 0 obj << /D [836 0 R /XYZ 71 791.866 null] >> endobj 90 0 obj << /D [836 0 R /XYZ 72 627.345 null] >> endobj 94 0 obj << /D [836 0 R /XYZ 72 505.192 null] >> endobj 98 0 obj << /D [836 0 R /XYZ 72 147.257 null] >> endobj 835 0 obj << /Font << /F70 201 0 R /F79 203 0 R /F57 146 0 R /F74 202 0 R /F83 207 0 R /F77 551 0 R /F59 148 0 R /F80 204 0 R /F73 223 0 R /F81 206 0 R >> /ProcSet [ /PDF /Text ] >> endobj 841 0 obj << /Length 1098 /Filter /FlateDecode >> stream xÚíYKE¾ûWô ¯Ð4õèç )â‚ð-Éaµöfyíeƒ²áßSóðxzÜ3[ÎB 'írUuõ÷}UÝõNúyÍë‹Ù7?Ù¨t„ˆjq«<)oXƒ±j±T¯çFWÀü—ëÇõõÍê««Hó÷Wù0ÿþññiûñþáúÏûíæêíâ•øó]  rÚX¬Ý-îVõo·ëõöù~ó®~»\=Ý'Vãë§Æp³Z-WËúùùn%WQ“ËÝip¶“;7F¦ä¤‡]ê‚íd+³MŽVNÌKÚ˜ˆ¤Ì~\Ìþ˜¡¨Rq§1*¤QÔíæaöú-¨¥|÷JA…¡çÊò¡Mk­~›ýÚÈaoó*W.’†ÐäÅùº' €›"±d“'¶ oó2 о »óØcéÙC_cw±“5c.@A(*åXettT[—Û  Jð1mJ‡±„+çI;Â1È‘ÕRÈõ‚Ũ¬Ý9Ñ!ÁIy¶ÐE]YV`Bc—U ¢û'áÍt0ç;•b¶¥üA3@^£èJ!Æä¦4Û6éÌ÷AyD9'LeàÿnHCke¬¶5)^€ƒa"Ý­œØ¤gY9¼:Énìü ùCåàΙGÍÖœpõ08¬† ê;ó¬ÏZ{£æ¨ŸM=üËõpá}ðÒé½;ÃvaºØ6Þ`à°¥¬E]â3Ž€y)rÇæ€Z„ü*¡ïB²È¯þ\ëpìi|ºt“ÁV[íTj=—zr+ü0p]á|5ïŒÕåã;@vÊ”¾EÊZÔ!Œ^U——Å|yØpÒß /{51šÊäÖ2tÑNœoÃ…ýa°±„$lsIØê—íÿqªp,—cˆ•RyïP#“+¡òß!e+0 endstream endobj 840 0 obj << /Type /Page /Contents 841 0 R /Resources 839 0 R /MediaBox [0 0 612 792] /Parent 830 0 R >> endobj 842 0 obj << /D [840 0 R /XYZ 125.795 791.866 null] >> endobj 102 0 obj << /D [840 0 R /XYZ 72 754.004 null] >> endobj 839 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F70 201 0 R /F79 203 0 R /F73 223 0 R /F80 204 0 R /F74 202 0 R /F77 551 0 R /F81 206 0 R >> /ProcSet [ /PDF /Text ] >> endobj 847 0 obj << /Length 3301 /Filter /FlateDecode >> stream xÚíisÛÆõ»~ê™NèN„ì}d&3MÇ“4NÝX3ÍÔöš„,Žy¨$t$¿¾‹· `AD=È`ê/+ˆ\¼Ý}÷±»$Éû„$ÏHøû·“£¯¾—6¡"åB±ää4Ñ,Ñ‚§DÈädž¼žÈ§Ç”2y‘M×O™6“ï/Ö³|±Yïž¾=ùѽ®JRK,-^'É1')'Æ¿ý­{ŪÉÊ¿lõä4¼ìßÕ$~×°TPî@À«+ß…ržèÔj]lj-OŽijóÝÞ®0£4µF%Qϯ}7ÓS¦š±rÈ_ ÞD)»üÉwáÍ."eZ•]~éÂK¥^Û"ÉÕ"?ó89ûí<ÛžO·ÓU–gÛ]¹nÕ|Ÿ¤ÄÊ@¹êæ06•<ê$©ÇÿæÔ4õž¿ìB¿Pnº|ù´ ¼N ­Ö¹ØE@Õd7›.§Û•á›yö†¶ÎæþËÍe¶-€;NaÄQ\8zòT ëÁæg™ç³«³Í2<Î7«ébÝE?ã¦dE‹€-Ô“ÔÈ )ùÙ4÷@g›ÕùEžíüÕ°Ùõy6Ëa²î¿Ëéò"ë•6冶¸µÉƒž[#š‡é5:9ÞÕ®»À¨”¨6QÝÔ¾édE“JYÍêY8íÈ\Íêu×ÚœøsÞbÝš´4¬ÏÊø+SñÊÞÞB@NlÙmsÚI–jNï`^E+DŸn¶-º/ÖŽº–¡œZºA¡&è…ÒˆÓAR‚Z=Œ:ët—)3A=Ê”õúÃÚ‰þét–Ý¢T™J…£+¼õƒ“+Î$LÝë Ó ž¨^èó—/~ê‚é[‘0ßl–ï6×_:À\]À÷Z»xú˜Ö¦L¤Š}\œ|] Ö©ˆòáâ.E\X;±6ú.E,j¿Î²ù.`sSü“Åê|™­²uî?Î 0­ßä *L<©îųÔÖ\þÍ-ÜK™B²|Æñ¹M{hž_»A‰ y:å*Âùt=÷º]H¿lÐí²„@ØvW VRg¶‹K‡ØËl×59­RZÏ®D¬h`„89rÑ’‹Ô½†BUE½¾é€åÔ¦U¼Àˆ fè¯E¯£g'Gÿ=*ÈëÈ•Pmj©SÊ'³ÕÑë·$™»/Lœ%sû º®œuH h„eòêè_ÞÏjXÁ"iaêý˜I…sM¬v^…g©l±Tnlc©m‚ ÷à”Šd ´µ$À¦¶âÞ )²Ív…åóÿ€4 õX<Ör³vb±ë¶†u¦"×â>‘*ÒâMø•ÿÓé:ª¡GXâ­…9w£ìÅ:çISU›‰7„’»14Ý• ›OˬäoG8"[¾Ë»l¹¹zj™·•›]þ…q NSl‚ÓÔRS¹mL‡i󮥅žaFg¾‹¢·Js¡º+?=­Ù ܨT™Ê:wêtN=èâ a ÿAHò!)Ÿþá—æú;Ù€¥Y÷ ž˜@¶’Ê$¥É¶ø×?þ(’7];îàãß&¯•pñÈSG¿‰ãÃ0;PX hWОC»„I|Yß5ÈÛ(dš€À´ç‘™ÞEvß½û*Ž« ö†‰ûf#fÃ1›N¢U$Ñæ†Dïë%zSŒÃrò³\ˆÕ§g¹;qN ‘Amd ±xþOdK/oðqéÆ9FaU®'¿GR°…vƒåŒ=Š€©¦Z&C`ôapª×—*âT8õŸÑs6ü\e ‹GP’“ÇóêíÐÜ©"‡NÝPÿ •;Íap§Šô¨Žô¨žü=rÑÖUh®¢;çŬ²HœêXhÖ mcÈ[KVX”ƒ–UH|l‰MÄÇ&DÇ?U^~’A; ÂðÜÌ-–ÞuwE}ëà|íÚ-›‰ƒõTÊY‹ƒ}Ááž,ìáìïYÇóž5¥¸óx@†i!¹¿Kmß¿ ’׬÷_VbhBRk}Rsâð>ÖÕ)‡¿oV§ÂËzpŸÞ€RÈšk¥dPz×Ðzb/[ضá¹Nq™* F™\’D2³’AýÅBêšà¥«ß"Ý@›Gí&$ÚzQÑ°kÈ-÷k-Æ>±+çÁ?½{˜cû*ªÃøÏ—½sŒ#rlIÈÉÍ"󿌂9VUà2 _·UuNE1“BÒÒ Vx¡ÝÐË×䩉I…å‰ÃàÛ'/#λ:¬Ä‰ãËaò¢¨_FU×ÙH„Kît‰< º=Kþ2ª%{AšT–ÜqO½‘£ eæÄ:mÃϡׯMÏ'{Ë…:tÓòªb!ÕËבiñ;ƒÉ³À[;,V=\οߖ«½“;úÐ-‹( òþ‡C«›®"~ò²ð++° eØZ´2,†Gù–š8¦bÏfj w«"¸’=DNÆE…óˆ°yxF±Wf$I;ŽÔ°°“Ÿ«tB¹Ñ'®5¼ÇÚ…;»ÁЗQmfÝòUDÝeÇÃö·eh£Îýø)n%êm¾xÉXv¡ÖÜÈC°<ÇÚýü¤gV~_šïšÉW½{î=ŠÖ“¿€&O{ŽEûæ²]Ï'÷ÕuŒŒCç2:’y°qè~'¼¾^Q:Äðóÿ ,¼ˆ ÷è5æáÜ ¤$œ žâÚ%´8gfÀ6£ïé­Í;úž^Õ8Opí|‚vlaíõ|EuŒ_»|È=½Q!ã#Ñ1‡òg\a%,8ñGŽ.ÉäÖxùgxþÚðɳGˆàÇSU=F£Xgá˜Å¢mÈõÃ)âåý4TÝ:kù´ç½ßºaW÷–f9­¢Fµòâ#~Öa„{àqo«½* s•ô׈ç^a'ª’‡Íä#zž»íe-ò¨®°ŒÚzÖ iÖŠ=´ GaQ¹~ù¯8ž«òò zSø2x±½íXzÖ,?™B»žWÐ'ƒç¼|«ê³ÃI¯<ú+õ^ÙÐEdž·½Õ­‰ºá‡khÏ¡õŸÌ‡9<÷N°R¦f4Žö÷‘Ë”,Ó s2…vxÓ² æ±Í‰ {ˆ¦ÎœBuæÜ`Oz׶V7N1Õ'Ju°ßI'ÿ –3¼œá:ä„Šsà1Â'_ô£ÓÈßõV=_`ÔlÊ<¦GÆ“ÑÔJØáÔJÌGnKz€l œ×·Q^_Gy}Óá広gonI‘³CÉî¯þàÙýûæ’h>1…´ë›K6ÉÅqˆ¼îw¡¥¯9Çy\.þÀékçí{SÀ®;œE@ãk½.¢Íø¦Û'†¸žŒÝ⸑+Wl€ØxÈf’Så!ÃÙ²[¬|ì8í{)^ú ªâA¤‰tøj€Ë.õGRK{ë`9’»ø¸‹Óè–£9–¶¿Æ:š‡™Áôy.‘ôÇJP;ÄcÞ9ìæõVÆf¼¾™"¸Æ’~ÔT,ì¼æ°o‚Ãá™›þõŠW¯¹jÝ=ÌQu¼À¢5²¨³X°·)Q£ÉÉbÑ‚‡¢åE 6Î?¬hÖ`rù6Úý•GÑÅ6âõ éDóÐ×¹=Ž3-‘=}te>G<\‡Úrƒê„8=LOTp=’<¤%à“÷•«àÕþb¤+?gê¹GÊQƒÈÔkø .±òõfˆ|½ŽîmÐQ¾¾<É>H¾^ƒ7à _Oûæë-AŠ˜훯W¯D°xüIeM°Ï%‚Ï%‚?b‰üºÆp )?WÇVbÈØ 52áZ«iÄLyt‚ ÜŒ†t!î<Ú&¶î!‰ïæÍ[ŸÛÐçÊ(Á?§þêzŽm®zº¼ïM’õÝÛx/˜9*Ì^yïs…¤÷ª½W…<ßË…ìOŸì‘éJ•¦]ª¬ŠŽ)UV=›¥J$—ÊoF½ˆ~î]$Ê[,sº/÷¶ èûs¢ŸŽ–jøµùâgâ©2'÷Ÿýìú endstream endobj 846 0 obj << /Type /Page /Contents 847 0 R /Resources 845 0 R /MediaBox [0 0 612 792] /Parent 830 0 R /Annots [ 843 0 R 844 0 R ] >> endobj 843 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 589.738 51.396 598.177] /A << /S /GoTo /D (noweb.NW2i2LRH-1KNHNk-1) >> >> endobj 844 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [161.931 587.1 173.894 599.996] /A << /S /GoTo /D (noweb.NW2i2LRH-1KNHNk-1) >> >> endobj 848 0 obj << /D [846 0 R /XYZ 71 791.866 null] >> endobj 106 0 obj << /D [846 0 R /XYZ 72 754.004 null] >> endobj 110 0 obj << /D [846 0 R /XYZ 72 680.789 null] >> endobj 359 0 obj << /D [846 0 R /XYZ 81.963 590.823 null] >> endobj 849 0 obj << /D [846 0 R /XYZ 72 588.096 null] >> endobj 850 0 obj << /D [846 0 R /XYZ 83.504 590.823 null] >> endobj 851 0 obj << /D [846 0 R /XYZ 83.504 578.868 null] >> endobj 852 0 obj << /D [846 0 R /XYZ 83.504 566.913 null] >> endobj 853 0 obj << /D [846 0 R /XYZ 83.504 554.958 null] >> endobj 854 0 obj << /D [846 0 R /XYZ 83.504 543.003 null] >> endobj 855 0 obj << /D [846 0 R /XYZ 83.504 531.048 null] >> endobj 856 0 obj << /D [846 0 R /XYZ 83.504 519.092 null] >> endobj 857 0 obj << /D [846 0 R /XYZ 83.504 507.137 null] >> endobj 858 0 obj << /D [846 0 R /XYZ 83.504 495.182 null] >> endobj 859 0 obj << /D [846 0 R /XYZ 83.504 483.227 null] >> endobj 860 0 obj << /D [846 0 R /XYZ 83.504 471.272 null] >> endobj 861 0 obj << /D [846 0 R /XYZ 83.504 459.317 null] >> endobj 862 0 obj << /D [846 0 R /XYZ 83.504 447.361 null] >> endobj 863 0 obj << /D [846 0 R /XYZ 83.504 435.406 null] >> endobj 864 0 obj << /D [846 0 R /XYZ 83.504 423.451 null] >> endobj 865 0 obj << /D [846 0 R /XYZ 83.504 411.496 null] >> endobj 866 0 obj << /D [846 0 R /XYZ 83.504 399.541 null] >> endobj 867 0 obj << /D [846 0 R /XYZ 83.504 387.586 null] >> endobj 868 0 obj << /D [846 0 R /XYZ 83.504 375.63 null] >> endobj 869 0 obj << /D [846 0 R /XYZ 83.504 363.675 null] >> endobj 870 0 obj << /D [846 0 R /XYZ 83.504 351.72 null] >> endobj 871 0 obj << /D [846 0 R /XYZ 83.504 339.765 null] >> endobj 872 0 obj << /D [846 0 R /XYZ 83.504 327.81 null] >> endobj 873 0 obj << /D [846 0 R /XYZ 83.504 315.855 null] >> endobj 874 0 obj << /D [846 0 R /XYZ 83.504 303.899 null] >> endobj 875 0 obj << /D [846 0 R /XYZ 83.504 291.944 null] >> endobj 876 0 obj << /D [846 0 R /XYZ 83.504 279.989 null] >> endobj 877 0 obj << /D [846 0 R /XYZ 83.504 268.034 null] >> endobj 878 0 obj << /D [846 0 R /XYZ 83.504 256.079 null] >> endobj 879 0 obj << /D [846 0 R /XYZ 83.504 244.124 null] >> endobj 880 0 obj << /D [846 0 R /XYZ 83.504 232.168 null] >> endobj 881 0 obj << /D [846 0 R /XYZ 83.504 220.213 null] >> endobj 882 0 obj << /D [846 0 R /XYZ 83.504 208.258 null] >> endobj 883 0 obj << /D [846 0 R /XYZ 83.504 196.303 null] >> endobj 884 0 obj << /D [846 0 R /XYZ 83.504 184.348 null] >> endobj 885 0 obj << /D [846 0 R /XYZ 83.504 172.393 null] >> endobj 886 0 obj << /D [846 0 R /XYZ 83.504 160.437 null] >> endobj 887 0 obj << /D [846 0 R /XYZ 83.504 148.482 null] >> endobj 888 0 obj << /D [846 0 R /XYZ 83.504 136.527 null] >> endobj 889 0 obj << /D [846 0 R /XYZ 83.504 124.572 null] >> endobj 890 0 obj << /D [846 0 R /XYZ 83.504 112.617 null] >> endobj 891 0 obj << /D [846 0 R /XYZ 83.504 100.662 null] >> endobj 892 0 obj << /D [846 0 R /XYZ 83.504 88.706 null] >> endobj 893 0 obj << /D [846 0 R /XYZ 83.504 76.751 null] >> endobj 845 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F70 201 0 R /F133 573 0 R /F79 203 0 R /F80 204 0 R /F73 223 0 R /F33 572 0 R /F136 574 0 R /F18 190 0 R /F74 202 0 R /F61 150 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 900 0 obj << /Length 3700 /Filter /FlateDecode >> stream xÚí\m“Û¶þî_¡~è„n|8âhêLÒ8Î$'v§8þ@K<ŠTD]ì›Äÿ½‹R REÇδs3H‚»‹Å>‹Å¸xörϾ¸»_ ¥þûiÖÔþqçïOï\>är¦DÌž^Í„DŒ±™dŌϞ.fÏ"Æï>úUK(ž!>Cx¶Ö—¶úÝš÷è`(!ÐÀPøó†ŸÎž]F£'w/¨Š²»,†r ¢<‰R G¢Â”µ)«iH«haä­Œ¼ó#ˆš×¢ËãZJÝei„–¦2ziJ­¡$zÔÖ›§åÑ”Ñq-9‰–G¶¤;4C¹cÎŒ.0FŠ7æ(¦5ÇÑrșˇ’v9Æ¢ax­[^>Øk@Q"Û/WK°?"“h^­n×ùËë}ÄHàÁ 8+Ö¼ÃSÛ¤ÃV!&TÓ"ßQT#(W3Ûb:§RŒÎ.(ˆ¤G‘»8ŽãèËåªÈ–Y¹ÉVÂGYZÚÚÛr¾É«²neÅ1R±Âšl<» 1Ž-ÁߣS¤WWÙZW“¨Î5y{;-¶2¯–«ªÎ7™m³4 õƒ«–¡¹\­«ÅÍd¯óz£; ¯ÌÀ‘€‡L‘"ΔÝvX±áK‰D"¶báZ1Í[ÆNTÉ´¥#É™møc¨G^ 3¯Õ[VN‡ º{!>- Ýk³ÈlÿˆcRdµ½H×în™.³…»7ŸWë…µ¸ÜTî÷Ú5]¥›M¶.­8ñE TNZkÐJ¿ÔÅßþù)`%zôùÇh`„0m½ÏUµvCɱ²CéœJƒFD”/u¹îÑ<Õc ½–I=½ÎkÛ` æ_¤k}Á£z•Í½Æ¶EZÕ+×ZwUÿ.2ݪÌìkéÕöJ8CÒµACb”@xDyÈ9‡E÷CbF ¢LuÁ1‘O*Û‘ü0$xÈ8n»ØøZé!…ƒwã¶ù8D ‡9øÖ¶Õë‰(0„V•öÐÇUUÜ–Õ2Oµ•ÃP9”“(ÇL€gÓ¸5zppDwp(ç08ægK·+ aˆm§¾÷sxõ³Ã-;«­-!öbŠúÜŽÓ˜sd0°ó›õ@_ÜÚOIó"}a&HxÚc0Ñ&1JðΤ`¼Ÿž¦ë›å2]çu3çnú6>OW®*í&ŒO4÷5G«}þôÎÏ.@‚Ø‚ ˜¸‚à€ÊÙ|yçÙóx¶€g_ÍbD!txeZ.g s$„žY‹Ù“;ßh$±D Ñ€Ö¶3–D³TéøUbM‹ƒû¤¶ÓOœÿÔýX¶Ê0^b@ŸzgQ@WÇ4PPPÇd˜Äáq‚õ gøÔqêΛÍ8I ^Î>Ì™0_î÷30.‚Ç2Þ1Ž†7:œuøꢩ.k¦ÉÅ|,ëƒÚeý;»bmù÷ýÔ˪&ê§d(ÏÓéç1ŽLûd¾A³¤ð{–iëh4?‰is˜G¸[:}Ÿ­«ÃǶǰ}¾[w÷ Xx}­£`}•¯ÒÛzÀÞ§¨µw_¢xÀøÆ1ÕKŽ€ñùL?°¼“™­– çY… ‡ñi,Q$Ýû›2°»qL{ìÎãº5»ìç›´pF‡ìm IZ{ó$Áæ6Žg¹y<‡¬ídžAKaLœgmÀ “XQH j}ñÔfÀàÆñí18ŸqØâÒ&R¥åfÀþÆ Ö3¿û‚Í p¦î1Ý5Õ¦Ç6^p+J»#Šè6AùÝ€Ÿ,yЙ:3…ŸŒåa19lÆÀLJb»þu^fézÀŽÇ1î±cŸóÖŽ #ƒ^õé«E¶ÊÊ…3ëª ®~ E1ç»!í–±â;6Ä{6Ðæ?ÿ"Õ„{i6&Œ8–cÚÉ*J”@¼º]ø>ÀçH…« >}…÷e Ø1Y±—%èAú$â7H‰¿ƒt–ðS‘~úà´ÞàäÞ‘Ìãä¹oäýëÛ2B(8ÆPÝúÉç!˜¯É:ÏvX¿_ùªiúé&ÖN?‡óU§ó š¥8/]Å„ž¥É4–-1xHé2æó´ÈŒäÛcØ>ãÚ0ö×nf¦ì·óq’ôع/ Œ?½‹ýâj×ß’ßÅ›¶¦>IWSßëj8Î'ÛÏiýq¤±ýb´=Åëm}žÅ¬9ÀAÝŽÕéK7û’ÌïDg:žTKMX(·_ײש&^Û«MúSVÚª9ÐØU9C‚¶cQÍ/o´Ä2½©>-Ã{ê‰Ø6 žÅM½Y§'Ä5Ä9–eoj÷¤Ù_¯ë漉•~ûÁF¨6 eàkèþ}ï Š¿˜í÷W·`æ‹Ç.Úíæƒ€Ö  `ã?¹-7àþóyZÌÑŸ‚aRÃ5¨Aˆm¶=Çrô,ÌÏÞäuó©Ùîþwfðâ6DÿBšÒË’ŽèK æè¯÷íïZÄlý7ûó‰¾e«——úWýëqµ)ÇVª½Có×eùÆ£(à†~¹¼ç®V©¾\§KwC¿£7oìÕo-‘{MíÊÕôÜe*Ë«7o| ¥g -^½³Bt7Zç(ÓäHB¢|ã~µ~eùæZuÒ÷R{ ìÔî´†[ó †?/Ý5v;²á>°boÈnßv7‹¸ç='?ç€êµ#Wx×±pVåvP¼î ¨tÂQ¡°, ¸iö’Ã=ÐÚ‹¼t'OôƒWÐsû(µ7æ™Þänn¬×é­½ii7çNô^º­Á(ÜO£t±Èõó´°÷‹¼ÞØ'œ« "%Ô¨ïü›¨î‚ÂÔ-±1DNÜ¢š°P(àZîêë[{¾¬ãǶÌÁ%Év£ G1jË4‡yÙ΀G[Œ!ÜĬ{n8>#™L|š—}:–%î¬êÒ”•)KS֦ܘrmÊ´½£&:ý«§ G 3f0s„–Mu¼Xè>Q]MCŽE£4âN.—Ó¨€mô{ãN ërnJ;¹©WÞÓúôs´Ö:öH¦µÇ“Åp6}?`ñG;ä.,t­¼ÊA´4õ‡¦¼1¥…ôÜÔ-€óðʵ¯<œÎß—Ãécfïؘ|ðhzŒ¶ Úwz4ý˜î)Ý-ÕíÞÞÀ3Äœ–M5¨e õöСvñ…qý™çú×Gv“WªOâi¢ÙÎkÐ@Åv“Ó ç])*S¯ÃSO ±c#…Ú‘âHÓÑJéèv÷¿ Ø*ÀVÐX÷4›ÓóÜGúBE:ˆÿèuóè ~ÖíÈPã"˜‰Ø΄§]BlSöá…ÀzcôÀÔ›ƒ©n?ÖÓ)Ô‰|¼2CÏM¹g€ÜøF>‘×7îx"BVÒ´Ñ0Ô¦¬L¹œf*±D›°fãêM€µ›O5R6“8Ùä ³”òb õÞĵcl¦Q1ÖS= l5åˆq'pc0 ­ëÅ„ümÔÒD@É^@-½茀:ðiÜf¶>.‹½ãÎzÈb뙵·&*úÕ¼nûb{ÖÖ]×HôA‡ÅP[»b;ÿ†Hßù¾ý÷DÂø¹’PMv%°ÑÞ›fžÁææk3˜·¦\™R÷“™ÿO 1ðÌXÿsƒ„³¦ÊéTS5†BùTѹՂ2ZPF õ<6o\{õ¥i“yñz浩§ˆ _ÖÎMÔã²U›rÑ©À‹w"PŒOÅ~‹£ƬR…ÃØ7¦Ï¥{z,ºp]ÔCW«ºðÿÑåãðó¿ƒ.6-ºÈ©èšï Kÿ »O¼ÄâvöiÒ‹Ÿµé¬½„crx&™omØ’ó—˜`1UxÇß^¶Çišy|Óˆ$á,amwݺµÆ•-LYxNo"ÐP³®WӂƮ©¥nËF]/LYœôhÑSTìˆï(éèë6+ÜÜñS€ ÀÄ›2tæzE…7c(3cp%f^’šò¹»sº ü'šC⣓Š¤Ý qTòõ)dñT¹? ¡Æ þåÈ$n¬ó@&,šhbÿpª~ÐQý çáÖ%0ëÿdéÿ5J‘Ôhh‚0!þ×ÃöÏŸÞù/þGó endstream endobj 899 0 obj << /Type /Page /Contents 900 0 R /Resources 898 0 R /MediaBox [0 0 612 792] /Parent 830 0 R /Annots [ 894 0 R 895 0 R 896 0 R 897 0 R ] >> endobj 894 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.478 715.648 165.939 727.753] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 895 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 234.127 51.396 242.476] /A << /S /GoTo /D (noweb.NW2i2LRH-4BH61R-1) >> >> endobj 896 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [163.131 231.399 175.094 244.465] /A << /S /GoTo /D (noweb.NW2i2LRH-4BH61R-1) >> >> endobj 897 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.478 171.86 165.939 183.965] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 901 0 obj << /D [899 0 R /XYZ 125.795 791.866 null] >> endobj 902 0 obj << /D [899 0 R /XYZ 83.504 755.001 null] >> endobj 903 0 obj << /D [899 0 R /XYZ 83.504 743.045 null] >> endobj 904 0 obj << /D [899 0 R /XYZ 83.504 731.09 null] >> endobj 114 0 obj << /D [899 0 R /XYZ 72 704.854 null] >> endobj 118 0 obj << /D [899 0 R /XYZ 72 395.03 null] >> endobj 360 0 obj << /D [899 0 R /XYZ 81.963 235.123 null] >> endobj 905 0 obj << /D [899 0 R /XYZ 72 232.396 null] >> endobj 906 0 obj << /D [899 0 R /XYZ 83.504 235.123 null] >> endobj 907 0 obj << /D [899 0 R /XYZ 83.504 223.168 null] >> endobj 908 0 obj << /D [899 0 R /XYZ 83.504 211.213 null] >> endobj 909 0 obj << /D [899 0 R /XYZ 83.504 199.257 null] >> endobj 910 0 obj << /D [899 0 R /XYZ 83.504 187.302 null] >> endobj 911 0 obj << /D [899 0 R /XYZ 83.504 175.347 null] >> endobj 912 0 obj << /D [899 0 R /XYZ 83.504 163.392 null] >> endobj 913 0 obj << /D [899 0 R /XYZ 83.504 151.437 null] >> endobj 914 0 obj << /D [899 0 R /XYZ 83.504 139.482 null] >> endobj 915 0 obj << /D [899 0 R /XYZ 83.504 127.526 null] >> endobj 916 0 obj << /D [899 0 R /XYZ 83.504 115.571 null] >> endobj 917 0 obj << /D [899 0 R /XYZ 83.504 103.616 null] >> endobj 918 0 obj << /D [899 0 R /XYZ 83.504 91.661 null] >> endobj 919 0 obj << /D [899 0 R /XYZ 83.504 79.706 null] >> endobj 898 0 obj << /Font << /F57 146 0 R /F55 149 0 R /F73 223 0 R /F61 150 0 R /F59 148 0 R /F70 201 0 R /F79 203 0 R /F74 202 0 R /F18 190 0 R /F77 551 0 R /F33 572 0 R /F80 204 0 R /F82 552 0 R >> /ProcSet [ /PDF /Text ] >> endobj 922 0 obj << /Length 1545 /Filter /FlateDecode >> stream xÚíZ[OãF~çWø¥ª#5éÜ=.Bª*µ+µª´jy£<¤!ìF$$%]´â¿wæ›ñ-qƒcÝ< ¤3¶çræ;ß¹ŒYò!aÉ»¹“þï&)®~;ùéüäû_t–ä£Ü“œ_'&)¥’LÉS:9¿J.R®—ç¿–±d¤“Oîü¿áòw¾]뇛‘Â=€¾éÐÁþ»ÉÅÐ(™®C™§ÓÁP±tCÒ©æé½ëT§+’î K'nvYº„\@®œ´±e 9ƒÜ }J3°Á`Êéò–¦ë,½Æ|ï!o!'µÌj+¾«¬Ü½-Ü #`†œr]Ï4ºÙ…˜Ò%Ä@ÆÈpsó`9s?"O¿ “°¢°¾ò:.P¤ß6Ú÷¬‚š-ÔœAÍÅšÿ,õbcûrÚµo»=×–±ô»ÁPdnCñJòÁ0SéVü„Ÿúv7ÖG÷ŸJ”@‚FTÐH¸;xSoãEŽ^$K/1àiœV/üH- dÕzÜ:ƸžCÒX‘6éßXñ#IwÐrKð¸nZBÖÏî_Ñ$,ÁÀ öÂÀ ,Á€-<)uE?Û~-üxJÌ3÷í±6—YßbD´y›…p!ë¢j"ƒ¦l¡)g!&…h…ÛXØŒÒ-·'ÕmÓr{NeRn!ŒÔ©¨8e-hàj…ë s¸Û‚«À؞ʷ³—€šLßCÞA.!¯üFs§å’5n&nJ-Qâ†Xq['r7è6J 6½«xT2 Ä~E8fÕD0Ð “3P¥*µWeiÇÈ <ò~ðX~6{_jĤŸ^Äfà(‡›/Îÿð›É·9;Ë+îê€ÊeÍ»<Êv†ËØ˧…Ê©xÇ"x”qQ³Â¼`Á$Q«®´ãä't=ÃP X°çPœÙŠË | yØ{Æ‘Yx¶ ±¿L¼f@Á%®÷íôÊ 6èYírË””IÌ o!Çw4TŒ··uG”1M¬®%4óZ¢sUKnBˆÿjMcìèö>§Œ\óèlj’cdx‘òn¢ˆCtèÈ?ʽޥ¦b„ÝÜÜ‚#š¹ë´¼¶=ý³ûÀ&2?Ôß-€Œñ«ç±f'5éïµ–5Zn2XÓÈ`ÇkHÇ„9˜.”0êÞÞ*ÿ–#“ Ÿv¤±ŒiI{ãšcl ýWEÜX:ÆÞÎ@<‹'š¯Èže5N[Vs g؈¿˜æ4´a#pVµjRz ƒr’¡ò¡ðÅ_£ðev _¦VøÊâj òèUòâ[ PÇ õzåD‡‚3*à6aÝÉéÿ "uæ©lFä5ϨÔßV×è¦A»¥ÁÞ0×o0?`ç›éa·­Ò`Ã3š³–¬8Ôˆ;*ª–¿Óæ Ócº¨û½aúL¯Z*¯ˆé"sÉ­q4AWÕ8DÌiVÐøÑþé Ø ×TlE™ò´–²úGìeRV˜«=mñ½Õ(¶Nx÷œ„¦£ŽOÊ®QûA}nHûì„Ù 0»¤ ‚>b£!WÕÖ-ãZK¸^ÔžÜ4Þ"ÀŒ†£Ÿût±F*³ð5 >*±Ñÿ—Ǽ-ÔöëZ¨^‘¸ÜGâyk©ü¿I‰¨ÄwRT$î†úŒÃœðÖ ´ÞÂeÍ_¨³QÎrî• ¸ÿÖX'Ž˜ÄHê¼µ2øóùÉ¿Dt) endstream endobj 921 0 obj << /Type /Page /Contents 922 0 R /Resources 920 0 R /MediaBox [0 0 612 792] /Parent 946 0 R >> endobj 923 0 obj << /D [921 0 R /XYZ 71 791.866 null] >> endobj 924 0 obj << /D [921 0 R /XYZ 83.504 755.001 null] >> endobj 925 0 obj << /D [921 0 R /XYZ 83.504 743.045 null] >> endobj 926 0 obj << /D [921 0 R /XYZ 83.504 731.09 null] >> endobj 927 0 obj << /D [921 0 R /XYZ 83.504 719.135 null] >> endobj 928 0 obj << /D [921 0 R /XYZ 83.504 707.18 null] >> endobj 929 0 obj << /D [921 0 R /XYZ 83.504 695.225 null] >> endobj 930 0 obj << /D [921 0 R /XYZ 83.504 683.27 null] >> endobj 931 0 obj << /D [921 0 R /XYZ 83.504 671.314 null] >> endobj 932 0 obj << /D [921 0 R /XYZ 83.504 659.359 null] >> endobj 933 0 obj << /D [921 0 R /XYZ 83.504 647.404 null] >> endobj 934 0 obj << /D [921 0 R /XYZ 83.504 635.449 null] >> endobj 935 0 obj << /D [921 0 R /XYZ 83.504 623.494 null] >> endobj 936 0 obj << /D [921 0 R /XYZ 83.504 611.539 null] >> endobj 937 0 obj << /D [921 0 R /XYZ 83.504 599.583 null] >> endobj 938 0 obj << /D [921 0 R /XYZ 83.504 587.628 null] >> endobj 939 0 obj << /D [921 0 R /XYZ 83.504 575.673 null] >> endobj 940 0 obj << /D [921 0 R /XYZ 83.504 563.718 null] >> endobj 941 0 obj << /D [921 0 R /XYZ 83.504 551.763 null] >> endobj 942 0 obj << /D [921 0 R /XYZ 83.504 539.808 null] >> endobj 943 0 obj << /D [921 0 R /XYZ 83.504 527.852 null] >> endobj 944 0 obj << /D [921 0 R /XYZ 83.504 515.897 null] >> endobj 945 0 obj << /D [921 0 R /XYZ 83.504 503.942 null] >> endobj 920 0 obj << /Font << /F57 146 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 951 0 obj << /Length 4344 /Filter /FlateDecode >> stream xÚíÛnÛFöÝ_¡ °€²¨™¹Y @³M¶H/A›dw ¤y`$9*Q %Ùñ~ýg†£C‘²iùЛ¼Œiz<žûmÉà〠~8!þç?ßž<ù—LTD\(6x{6Ðl ˆƒ·ãÁ»¡z|J !Ãïi>M³Ñäñ)Óñð_›l´ž.²Õã÷o4‹è%QBZ,B§œDœÄngO¹N†#°7Kœù%Ü šÀ(I¢DQ³’]áO7‡r>ÐQ¢‰£"¢äà”F±`nÚD¿ZRYͼ£0ó[7-®wÐ>¿ZNòeš§óÉz’¯JªêF Hbð>%øª‰ *bðêž°8sxLݯ?üÚ„H‘D± ÿ|Ö¼¼¦ZÓ•[-uk¯Fé,ÍÝ-H/ƒâ¯‘–ÜàŸGR$î¿Ç“'ÙdlˆTÒáâb’»«õùÄ]\ž/fþr¼˜§Ó¬BØÅ Ùõ–øt36à2nkz3)¥ÍVìõyºv-æËÍz²ÚÙà‘&®àI$¨ÞáŠ*­\¡ IøU&‰ˆ‰@韛¶^Yæ›æ5b¹³„ÖöL$jäLQØBŽ¼ ó´‘)âHÊ þÓ´¤ÁŒ<ñ® ŒFp¾C’zçÝdLÛQìBàZ\Ð&HWpñM#HÓ÷ËA:ÆÞ(KšQƨÙ\€Ó‹f”1vön»½+ž·² 7 Í>m¤#i¦ÚÌôzƒ9¶;ªLl!¯~o^JÄñM:PNB thrmd,/l=±këåÓ ƒŒ‹I#ÝË‚Èø.TEÕ„3úÄ€£XÚŸÍ/À‰€YOÖ2vIlfp$Þéù®˜uòâíɧ“‚» ®ŒE/AHʼ5«æ'ïÞ“ÁØüñlj„1Î/íÔù€j'ÏÏoN~sŽZõåZDáíYä;€O1f²yA'cb㌙WåqRyÕ]l s!ÍæXl;RÍH$JZ0WÃˉ%}.ÃQ¾X­Ü-ÃnóF,Q³}­ã]‹V7Q­âAîÞ¬ Xáßl'>Ý#p9ã72ŠjÍ(²-£ˆºC±Ã(úF‹›^/‚Û‚ÊàXÑAeâ>ÎLˆ¨é†*D…Ç„¾=÷mΰGœ$íÉ‘·EB¶5$Öƒ<>[䎡—ùd<.¸ 2"2®Šå4÷&I>ù´™šÿ°*>„œÊŸö‰E`A «ò¹*Úø4±æŠó©USèÄÏô›;wSŠ×Úƒ#£‰C°*š‡·Ý>[í“ÍÙüPceŸþA˜…šìdðç ¼úɽX¼bVôž&$ŠYAƒÖ¾q+ÐDrÑA^üê._[I°ŽÙ‹Þú6o±Àõ¼;UE¼áñ©ˆ‡ ;^˜QsÛq îdvÙ;”Çkc/Ùå6vÌìè6´¶×îñ‹ð×x¸Âyï¤x€QK ”åÌî>ØåpÀ"Yž˜—5†µ(õ{¼Â‚à6•Ý¯öãÔÞI=êP¨… —vÑÜŽUHK-Ê€¨|¼B¢MoIšºÕƒÍÒZ #,ª{k,Œ,ÆÜÂF ‘Œ"¤XÔP#* @2¼´‹V°8õr°"NÙшÓ1 Ž³0Æ;ÚbGûëÌ ¦bcýˆŸZ"ÈìõØŽ+,¬-¬ô8â¬Qà§wb wc wJÞCÒ;Ìce—¹5`îø@æþ‡¹c+Τ݊Îí¸´ãÌŽH,Îlž%)ÞÒŒcœEE»s{½òrl‹uk{¢/ß㠌C·!ŽCn‰›èŒØ4iamÆÁf(­Md®‡*=®qýí­M²ôÞ˜|xz¸ñeiL$àoåñø}0ʶ"³¸³î~G\šçœâpߣ˜tD»PLFržYãhaÅqŽ³(|'‡œÒT_TþmÖøÖs¼GÁ–@nR°‡ TÕr‘Þ‡¹°ãÏv|iÇWÝïB¨$ÁbâÕ„³U2¯2Š1µcŽ©>Š7½úP5£QšÞJ@ jÖý fU‘/NülÇ—þ¯@ À™÷ãî÷ÅÕOúŽ¯wŠƒqÐõÌêÐ×VŸ>GáÇø¯ÄÐT›|åǯüØ)?*V„ŸýÒÔ!Ù|I?¸RjJ<5ý(È&$|œ0ñöLf¯§ÞuÿÛýNÇâS³Ð/€+×€OsϹ÷V¨'±ÐìBº‡O¯£v Ǯ7… Jó6íBløÔ̇¶$|øÄ^³Ç§Zº\6Ɔ¸UäO¬"gXzÎαvrï[‰5e *± Jk D<\0蕽~U‹:Lú "Ûb$WꃣHÉ› d1и Îd™ïü³+a¹Ï‰Ö] Ë’*vCwp¢!•¥´ï|´æÞª|Ä=U9zC'¶I†Ä_×ÿŠ™"ÕüaK]ؽ[ ’„¶›gKæ B[Ìä­êÜšÔ5[ÕD\†šîóYkÌšçor‹)á°jý€Âb"¡èîî²õC¶©ßçWÀèPõ,€˜z>¼‡=R‰g5¬@Î~Fç!9Ô‰O©O‚P ÀJ”–òþÒDhâDhá*Ðvì­h«è¶5Ap0Ñu¹¡àÌ|û¤t粘†ãHIÞ))|xok*+%Ë;3;^ÝCn2†gû-AÒnfÇ+à]À!PaZ}ú¡ÜzΰswGô×Ýù5|ÕÝYôËÝažä§ òs*?/²\á*È%à©™]ú hcGgóZ²|†õø/,{+˃§æs;.C•¼ò%!k¬0Ï*鳬Ŕ֭·Žç²Èþʧ×öú·ruå ¸ò)‡3šÑö¯3,Þü ®S_²-Ik'+F÷'“,ÜÙ·ÜP\ÔTÿöµMv<ÇIs¨¿†<èµ½2eù r`;/»ˆ'ä µ©ªa+-P‘‚(® ¢ê)‘~¨/×/6“¬èuû½~ÑwEO½<úP_Úšy»(¤C?šùXîç -bPཛÒ)uoÒ…»ßC.;òÊ»«ûR*Ä ©ðUÝî“6‹#…+ŽhM%-ÄQÜ8ê•÷q0{'}JHVK¤ßØñE-8çæü»W%ÒÜÝçïøwjsÜ츰ð ÜV¬ké_¬±¾]:W!ûº ‚mN(ú*®Ùžb× 1°§”¯6Ù1µ˜˜5`âPfÈe‡ïãʶ@‹ ‚çX·°Ý‰5Ü÷\6‚cnM¿‰-¯wf`ÚúÑòЕ†Bvfb_c‡øÀ®-­ÞûȲ÷c¯zÝÐJõÊÄQ¡ÙpëÔ~i†’ꯡô Hk¬nÞ§¡Äµöe‹†büë$[nt ’S¯Á‰¡²Ä¦xV;}=cÔð"¨*­“¸‡t….%b›B„ó«~à˜q@‰#f,{rWú¤É¼Ö¬æcg¹3`ÿìf7õ­ƒ}¨Íl¹ï2;ó#RÀóÔâá-ïph*õÝ!ƒ¢Ä7v ¥yè(.B^æõµŽŽ|kðb¼hý_Oº\_Ëá?¬„Zœ·žù¨õÊ­šî<ÃÊ?­AÄ$¶Þ ¸³?âè‘ä8ä7'­‰ÕÉuâ:‹Ö$ø‹«æ ù¿S7bßwU¶”_…CïÜ’…ûÒów–á´=oU,T˜Ï”!uº¶ÔVt¶ìvÞ¶ÀÇ–­Ø ±ˆõNÞ:®U7#³x¸•'Ï& ¶&­µÜFk)þ”²®Áuá+`{)PŸ¡*g³¶t#Ñm°Ššzï*B9é¡]XšµÎ†÷K„òP• ½/ÄÏY?¯@T¢Z+Ik 4.žƒ|( žl›n),?ضTD+3æÛTY䡦=Q×Ý1ÝN_ýþ?pŒµ!Ì=ñ)ß»óE“Òîà#ËZ¿é4˜T>;âðÑö+ŽŒ¦ÇQbÇÙ‘„Kxñ!+ì2Ñï´ÐÝÂÈíò•´¥ ž‚F…k ¸Ñú¹¦°ú2çõ¦t±‚ ˜_Ä«”)N,¬GA–– sÖE“§æ ¢®U¢_N@·ë>jwhŒ˜ó£zÔ6ï[~ÆìÈ7ÁCª\û4ç9дµ/§m<Ï" …1 þeEꔵ®õ œÚyƒ¥#Iéýcø bZ<¯}&pŽågˆ0VõS†ŒÖµO¼`…ii$Ý¢ê} [Ùy˽aë+—XîѼX¹ÂJPÛzµ+;n?Þg> endobj 947 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 576.188 51.396 584.628] /A << /S /GoTo /D (noweb.NW2i2LRH-4Mxm99-1) >> >> endobj 948 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [152.844 573.551 164.807 586.447] /A << /S /GoTo /D (noweb.NW2i2LRH-4Mxm99-1) >> >> endobj 952 0 obj << /D [950 0 R /XYZ 125.795 791.866 null] >> endobj 122 0 obj << /D [950 0 R /XYZ 72 754.004 null] >> endobj 126 0 obj << /D [950 0 R /XYZ 72 667.24 null] >> endobj 361 0 obj << /D [950 0 R /XYZ 81.963 577.274 null] >> endobj 953 0 obj << /D [950 0 R /XYZ 72 574.547 null] >> endobj 954 0 obj << /D [950 0 R /XYZ 83.504 577.274 null] >> endobj 955 0 obj << /D [950 0 R /XYZ 83.504 565.319 null] >> endobj 956 0 obj << /D [950 0 R /XYZ 83.504 553.364 null] >> endobj 957 0 obj << /D [950 0 R /XYZ 83.504 541.409 null] >> endobj 958 0 obj << /D [950 0 R /XYZ 83.504 529.454 null] >> endobj 959 0 obj << /D [950 0 R /XYZ 83.504 517.498 null] >> endobj 960 0 obj << /D [950 0 R /XYZ 83.504 505.543 null] >> endobj 961 0 obj << /D [950 0 R /XYZ 83.504 493.588 null] >> endobj 962 0 obj << /D [950 0 R /XYZ 83.504 481.633 null] >> endobj 963 0 obj << /D [950 0 R /XYZ 83.504 469.678 null] >> endobj 964 0 obj << /D [950 0 R /XYZ 83.504 457.723 null] >> endobj 965 0 obj << /D [950 0 R /XYZ 83.504 445.767 null] >> endobj 966 0 obj << /D [950 0 R /XYZ 83.504 433.812 null] >> endobj 967 0 obj << /D [950 0 R /XYZ 83.504 421.857 null] >> endobj 968 0 obj << /D [950 0 R /XYZ 83.504 409.902 null] >> endobj 969 0 obj << /D [950 0 R /XYZ 83.504 397.947 null] >> endobj 970 0 obj << /D [950 0 R /XYZ 83.504 385.992 null] >> endobj 971 0 obj << /D [950 0 R /XYZ 83.504 374.036 null] >> endobj 972 0 obj << /D [950 0 R /XYZ 83.504 362.081 null] >> endobj 973 0 obj << /D [950 0 R /XYZ 83.504 350.126 null] >> endobj 974 0 obj << /D [950 0 R /XYZ 83.504 338.171 null] >> endobj 975 0 obj << /D [950 0 R /XYZ 83.504 326.216 null] >> endobj 976 0 obj << /D [950 0 R /XYZ 83.504 314.261 null] >> endobj 977 0 obj << /D [950 0 R /XYZ 83.504 302.305 null] >> endobj 978 0 obj << /D [950 0 R /XYZ 83.504 290.35 null] >> endobj 979 0 obj << /D [950 0 R /XYZ 83.504 278.395 null] >> endobj 980 0 obj << /D [950 0 R /XYZ 83.504 266.44 null] >> endobj 981 0 obj << /D [950 0 R /XYZ 83.504 254.485 null] >> endobj 982 0 obj << /D [950 0 R /XYZ 83.504 242.53 null] >> endobj 983 0 obj << /D [950 0 R /XYZ 83.504 230.574 null] >> endobj 984 0 obj << /D [950 0 R /XYZ 83.504 218.619 null] >> endobj 985 0 obj << /D [950 0 R /XYZ 83.504 206.664 null] >> endobj 986 0 obj << /D [950 0 R /XYZ 83.504 194.709 null] >> endobj 987 0 obj << /D [950 0 R /XYZ 83.504 182.754 null] >> endobj 988 0 obj << /D [950 0 R /XYZ 83.504 170.799 null] >> endobj 989 0 obj << /D [950 0 R /XYZ 83.504 158.843 null] >> endobj 990 0 obj << /D [950 0 R /XYZ 83.504 146.888 null] >> endobj 991 0 obj << /D [950 0 R /XYZ 83.504 134.933 null] >> endobj 992 0 obj << /D [950 0 R /XYZ 83.504 122.978 null] >> endobj 993 0 obj << /D [950 0 R /XYZ 83.504 111.023 null] >> endobj 994 0 obj << /D [950 0 R /XYZ 83.504 99.068 null] >> endobj 995 0 obj << /D [950 0 R /XYZ 83.504 87.112 null] >> endobj 996 0 obj << /D [950 0 R /XYZ 83.504 75.157 null] >> endobj 949 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F70 201 0 R /F133 573 0 R /F79 203 0 R /F80 204 0 R /F73 223 0 R /F33 572 0 R /F136 574 0 R /F77 551 0 R /F74 202 0 R /F61 150 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 1000 0 obj << /Length 3315 /Filter /FlateDecode >> stream xÚí]íoÛ6ÿÞ¿Â8àPm\Qßnp»u°b»­ pzý&Nb´qRÛI“;ì?ê!%S¶“£•Ÿº0°žb“õ¼þž惓A>øùIîÿev¬þû4¨?½yòý'/_ 50C# 9Ø?H5,Ër J>ÌK1Ø?¼ÏÊòÙ‡ý_š‰òÁP †l0­þ×}|ûs5æarÈ‹Â~føkÄwÿuð~O–<{ýl›ìüÙ^™gSȤBe#;Ë®i< ñŒ¸ ÏŸiE,f§c¹Î^`Öd‡veIw$é.êqLã„ÆÔb‰{@Å鶽™]Ñç·4þÑÜ®²ï¦"ç,¢Vx—9-,éÊlNã%Õuá÷lYLòjK§D,#̤ÌÞ]u§sÝ]›{ ÷­DöôÙ^!³ç"{I‚g¯â¾i7æ¹ÝvTÕìÙž[Ñ^¶&Y'”–$Òž%{Œ ¨E’ÀŠ¤Î÷!Ó– «Ýn·!¼Tfÿ²WõD…%uÈÔš¦Ë³¯^šƒ8§ºSN|R’ì.‰÷+6…,`Z²í„ÆjMê©EË+z>%h*ùWfÃP£J„+ôãsÅÿÝõ‚Ë"{ƒá¾<û«ôÚUf¿Ñø¿ÒøS´^Ý@¾«J‘ñ"»±#³üõ+“ƒÌÂX,V­¹Î„j™a¤›%ê%=Œ…½·ixvÜKéäac¡)’ÛµA9ÁØ7¬e¥:ù|Üļ‘Ø0s°¬ÔƒÝº mÝfÒâNƒVb Ú«åSzŽj[IGžÓèü…ú<¥q\™Ðè¾Y^åôò-à¬ê²eU_cžRV¶Œú”ÆQ³ Çùî­ÜÐxr}ÉN×¼ð®"5Êü|QÖƒ{&i:z ¥!‰bèmÓº+Ö =®¸Ïgõ7í_ç+¿š¡n/ì6bX„«[ó°VBT[‚íh¢ô˜7¨]{…²,—¡èA R¹öñQ€Î!„ÌèóÈŽä S+2žÓ–(ÉY$‚%ä š³¼(ú襔 #táÉko¨Î–LT2ÇrJ°98ãí6L¥Da*%ío=õ‡ÑvSåãæKˆâý\´õË=Z4jP!g¿«À~¯ÿz„t}î6¸µ‡²ÄàîIÏ@­‰õj«³$ç‰ØÀ’Ø‹y ÷?4ŽPS§j0M¢q$L,B”i ®B$‹à[‹pQA!ŠóQmŠê`"`²HÒ‘<™X¹‹€í"`ÝÅ(åoÖØ.eËg¨} È&h@p­‚­¨ J&—bb\ÆZUGDÄoˆˆÇ("††…ÄcTD¹°£°P‘rX•âÊó¨×.Â’2",©v¡]Ö·X„U&ŠõÀª.ZÕ eÀÆ7šØ¡nñ1 Èkܦ]:Ä*pˆõŠC\×ÉÍa ¼ñTA"Â9f#`öc©½»>j6§Þ–ãà !«eÍj sd§1a¼¶< ñÆ ã,½rßC€#¶NBËIQg¨ÌPø8¶×'±³¬,’‘•¡Vœ8 œw¨_(z¨¼îk6:§ ðZ[ÿ5Fb·àÁ*@ °D ºÀ²g/ƒDŠ0Ä|‚ÊD +—NÞNÒþŒ2&‚ܯi_róð ^ûšM×S “Q<÷ѹ‰×`{8‘Ÿ½+†Ä„ºU"¡nµ ¡n¦ ÉPþ+s“ uÛ;zé“/PÎi¼¢ñ-*uJcdnÆ uh^.êLrG|ñ¿W>v/ð„Qp" “ýIâDjGòˆm1ý÷ÀÄsfÝ¢y¿&òP]š÷뢢 œØÐi$ƒ¨tªƒ›Ȧg³ûºÆóÆQA~FüQ%ÊšÙ?IJM›ù˜âÝ×Á aùèyHÀrQð9ï’µ2€Ç+q„L>ªÕÒÔ;ü ©/ƒ‚¼(ó>°x„ ëÖþ_70ŠZœJ¬SŸA¿v æE«8ú¢ ô·\‘ìCˆû×°Sˆûßô„3¶R­ŒÁsY„ æØ`Ða°ÌUÐ"ñ!š&Z³ý˜8ô²é¶É½€˜7bÓÍÕmËP-c<ÞQ»GH¸–¸• P[ó\^i|¶€ÓìWAÎÀ´u@Dº¤Uj餋ñ$?r\ó‡M‹³:`Ü`ÚîWfƒ®%Cø)q•– y’øŸN¤ðNó-ðæK©KzOì)¢ÁžUeïèóa ©?GKjÒ¬{DKëòLìVk˜uç±–¬ž©$ ¢¿ƒÀºÜþÉ\€rö™®èó;/éÊYìÙ"b_EkÙ=ˆ=`àûÞŠ“%?-Ç,>¢ÞÊý0ìß‚+ã–g×A0Œ¨uŸØ‰€‡gè æußï4Në§Ü„R«…÷du˜Ýž»ûùi´1wwÌeA¯¬ßuÖsñqÁZ^éûOÄøQw îÄ=^Ô4è kX±­¡W®©sÕrÝ{gyZ$sŒâqpŒâ$>‹®‰ª)~Žb£ÓÀÐøŠ*× ó?&À„LiîjE ò»Zˆ|m+à©ê¦é‡š>ÀVËê–~Öj¥Ÿµê[ËõÝ粩3ò¾Ðxlì´Õ{9ax*Àù LB¤¬éØ. ÏÀd…žˆ=¼ Áq¦sǵ¿ ºŸŽ®¡Üàd±±S[•Y°ÊÏ+«Ä¡…ûkÁiDUF&¢ñYãe`s¨5.Õ¶÷””ÅJ†òŽ ]—1ÜW%º#…Ÿ×mÕãáyÍ–Ïi“m’kf?4U\yfÚÙSw³aôœ#èœQ [îb+˜íÙ€'ÛêOßâñé%C¬N`zšDÚ+ó<ã(%qØ SRîóY`զѧ}xœ÷Å7Ë}´w¿Lò uÔŒÖ>Ì«6†z7ø>Ø‚EG:ò’`/œç¸#CgkO¡ØP¹ò–Ú» )íþ°ZwéTðÚÎ)AMM"‡;›tÊ·ß•GÀ¦lMj¼Õ…GÐáa´íü2ÒSæmš5u4õ=Îú1Àœ®¢çŒ7|Ï:¾1„X°¡–¬Mˆ,ïœôuæˆõ7ÒV-/_+Þ^RS/INîË×vŽÅ7øP+YáäâÌ_¡tvx~q3ŸœÎÝOìhûYV?‘z˜›²þ8p_i­k†¥l–]ötÚÿÒä,šÜ0âW.‡…Tƒ=nÓNJSm+ó§ý'ÿ³Zp endstream endobj 999 0 obj << /Type /Page /Contents 1000 0 R /Resources 998 0 R /MediaBox [0 0 612 792] /Parent 946 0 R /Annots [ 997 0 R ] >> endobj 997 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.478 58.114 165.939 70.219] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 1001 0 obj << /D [999 0 R /XYZ 71 791.866 null] >> endobj 1002 0 obj << /D [999 0 R /XYZ 83.504 755.001 null] >> endobj 1003 0 obj << /D [999 0 R /XYZ 83.504 743.045 null] >> endobj 1004 0 obj << /D [999 0 R /XYZ 83.504 731.09 null] >> endobj 1005 0 obj << /D [999 0 R /XYZ 83.504 719.135 null] >> endobj 1006 0 obj << /D [999 0 R /XYZ 83.504 707.18 null] >> endobj 1007 0 obj << /D [999 0 R /XYZ 83.504 695.225 null] >> endobj 1008 0 obj << /D [999 0 R /XYZ 83.504 683.27 null] >> endobj 1009 0 obj << /D [999 0 R /XYZ 83.504 671.314 null] >> endobj 1010 0 obj << /D [999 0 R /XYZ 83.504 659.359 null] >> endobj 1011 0 obj << /D [999 0 R /XYZ 83.504 647.404 null] >> endobj 1012 0 obj << /D [999 0 R /XYZ 83.504 635.449 null] >> endobj 1013 0 obj << /D [999 0 R /XYZ 83.504 623.494 null] >> endobj 1014 0 obj << /D [999 0 R /XYZ 83.504 611.539 null] >> endobj 1015 0 obj << /D [999 0 R /XYZ 83.504 599.583 null] >> endobj 1016 0 obj << /D [999 0 R /XYZ 83.504 587.628 null] >> endobj 1017 0 obj << /D [999 0 R /XYZ 83.504 575.673 null] >> endobj 1018 0 obj << /D [999 0 R /XYZ 83.504 563.718 null] >> endobj 1019 0 obj << /D [999 0 R /XYZ 83.504 551.763 null] >> endobj 1020 0 obj << /D [999 0 R /XYZ 83.504 539.808 null] >> endobj 1021 0 obj << /D [999 0 R /XYZ 83.504 527.852 null] >> endobj 1022 0 obj << /D [999 0 R /XYZ 83.504 515.897 null] >> endobj 1023 0 obj << /D [999 0 R /XYZ 83.504 503.942 null] >> endobj 1024 0 obj << /D [999 0 R /XYZ 83.504 491.987 null] >> endobj 1025 0 obj << /D [999 0 R /XYZ 83.504 480.032 null] >> endobj 1026 0 obj << /D [999 0 R /XYZ 83.504 468.077 null] >> endobj 1027 0 obj << /D [999 0 R /XYZ 83.504 456.121 null] >> endobj 1028 0 obj << /D [999 0 R /XYZ 83.504 444.166 null] >> endobj 1029 0 obj << /D [999 0 R /XYZ 83.504 432.211 null] >> endobj 1030 0 obj << /D [999 0 R /XYZ 83.504 420.256 null] >> endobj 1031 0 obj << /D [999 0 R /XYZ 83.504 408.301 null] >> endobj 1032 0 obj << /D [999 0 R /XYZ 83.504 396.346 null] >> endobj 1033 0 obj << /D [999 0 R /XYZ 83.504 384.39 null] >> endobj 1034 0 obj << /D [999 0 R /XYZ 83.504 372.435 null] >> endobj 1035 0 obj << /D [999 0 R /XYZ 83.504 360.48 null] >> endobj 1036 0 obj << /D [999 0 R /XYZ 83.504 348.525 null] >> endobj 1037 0 obj << /D [999 0 R /XYZ 83.504 336.57 null] >> endobj 1038 0 obj << /D [999 0 R /XYZ 83.504 324.615 null] >> endobj 1039 0 obj << /D [999 0 R /XYZ 83.504 312.659 null] >> endobj 1040 0 obj << /D [999 0 R /XYZ 83.504 300.704 null] >> endobj 1041 0 obj << /D [999 0 R /XYZ 83.504 288.749 null] >> endobj 1042 0 obj << /D [999 0 R /XYZ 83.504 276.794 null] >> endobj 1043 0 obj << /D [999 0 R /XYZ 83.504 264.839 null] >> endobj 1044 0 obj << /D [999 0 R /XYZ 83.504 252.884 null] >> endobj 1045 0 obj << /D [999 0 R /XYZ 83.504 240.928 null] >> endobj 1046 0 obj << /D [999 0 R /XYZ 83.504 228.973 null] >> endobj 1047 0 obj << /D [999 0 R /XYZ 83.504 217.018 null] >> endobj 1048 0 obj << /D [999 0 R /XYZ 83.504 205.063 null] >> endobj 1049 0 obj << /D [999 0 R /XYZ 83.504 193.108 null] >> endobj 1050 0 obj << /D [999 0 R /XYZ 83.504 181.153 null] >> endobj 1051 0 obj << /D [999 0 R /XYZ 83.504 169.197 null] >> endobj 1052 0 obj << /D [999 0 R /XYZ 83.504 157.242 null] >> endobj 1053 0 obj << /D [999 0 R /XYZ 83.504 145.287 null] >> endobj 1054 0 obj << /D [999 0 R /XYZ 83.504 133.332 null] >> endobj 1055 0 obj << /D [999 0 R /XYZ 83.504 121.377 null] >> endobj 1056 0 obj << /D [999 0 R /XYZ 83.504 109.422 null] >> endobj 1057 0 obj << /D [999 0 R /XYZ 83.504 97.466 null] >> endobj 1058 0 obj << /D [999 0 R /XYZ 83.504 85.511 null] >> endobj 1059 0 obj << /D [999 0 R /XYZ 83.504 73.556 null] >> endobj 998 0 obj << /Font << /F57 146 0 R /F55 149 0 R /F73 223 0 R /F61 150 0 R >> /ProcSet [ /PDF /Text ] >> endobj 1065 0 obj << /Length 6555 /Filter /FlateDecode >> stream xÚí]ë“Û6’ÿî¿Bùpµr­ÅàI»çÔíÆIÊ{—”{«®ÎNUh Çf¢ÇDÒÌØw›ÿý HTƒ¤(züH>̈’ 4Ðèǯ/2{5#³oîûú÷g÷>ÿZ¦3J£TJ6{v9Sl¦ˆ³g³çó8b÷”2¼¾Zåë|sÈ/î/˜Jæ_no²]‘m–yùþëëÍòPl7ûû?>ûÔ«f”D)I©®—Ì,Ž„¤eµO‹u±Êv«·ðÓ”Ï[ýÊæ‡×yùÁ:Ï6åÓeUëý>žßÚW»íMq‘—¿ËÊÏn_oWöëlõ*¹Ëʯ·—å‡K·Åðù¥ßbEÜ ϱ„–›ÿbˤn™4â1­Jü¥,‘xµÈ(–¼*ñ߶î–€G¼·E^°z`ˆMº+¢mŠ|Vá~1Rwél "fQ b° <’"-‹Þ‡×å(—CûlmŸ^å›|W,Ë7›úã«ìpÈw›’4Ìm‡"«› Ãò9üýûwûö+¶ùÑkY¢"™ÖÜÎö%•—ùåv—Gú bRE‰¢Ðö4b@Ò”ü{¾Aâ<æZRôËj»yeŸŠý¡|ÒR¢_÷…–ôòÙ•ýÞ•FžH#¦ÚÕ~[Õr™ïéw-Åq¯O"ÄôHŠ>?*Ø$ÏS? š‘²^åÙâ±Mëü×ãï:tcÑ€n¸dWÅ&Ïv÷)¥U<à'»8$S}Y³ÄÍš1†æ» ¢F~@dAç‹RÇ!<êÐÆ‘üHQmtù{*|5 ûsx*ŠiìzªÁ>WÖq=  R·‡]¨ôd¢¨zP"ÏS1 Íá´·¶{UŒc4*ËvZ6Žn@Ë\ÊVËŒ¯iò·Eöj»ÉVöó¼xõúPl^¡zHE)‘L¥ˆÉTŠ8’e¸"º,;A[)€#M|Aˆõ= wjK!Rs•qQgL<…MjŽ2+§,PïÇ1Ô–Zï]†^b !8¾FéÃü8“Ÿ$”ÑóQ’X3C1rÀ,ŦЃû «y™‡8Ð(!áÇ-ð„V·€{Xâ¦G¸hÕ”zÕaO Ô”~f”ND“û^YÃí¦+JE6` ]¸1|Yd{Ôò©$JhòÁ!QÜQ¸Ýs¹sJr÷.QÊq"AB{œšþŒ)¼Ÿ ¡Ô¶›‘ è °Ûû‡?¡Á8HR'%º#¡ D»¶±žh° ?uœSÛ…Ú "γ72I @<ü‚– j:þd»z67#©âöÆ£{t7ÛuQ­&€#-ÊiÉÍàŒäž€ðøÓ„æ-è;-cHëÉl¡"S_¨ôÈásP:"ãضnñë&'³n±nGŠíX·%^§±n.ËÎ7´.‚†k¤àèЂiÐá«E³€5“Žñ*“‘Š“1)£vÛ‰±”w:O‹ Í‹Ï›C—Š›æv7C€U½VW]¡b?ýª3æI6`v]—×+kp÷¿^g»j©Zþæ `çææ7¿³Á?8ó;ŽOóëòiZóÛžï×9Çþ÷-өǪ$R*ÅU13 tèÀÂ)‰…Í)DYLÛµ4q_‹³ ´tJ&`EÙÑj#A}ÅYïh­:•:´ªJÁT³¯ 7Ç1SÀ‹pû½Ma3`­ºÊÉÈw˜A¼¿ßôu¾±­1e“}… OLŽ´¿íYÙ»Éê IêÜaN§µ¤°Óéj*A‡¢Œ“%ERž—Z—2hïÒNnÉ{QF bFâ eûŽeF#ÉP†KØÏ¡Ÿ†4H.ùð½q¼ —Wç"!@c(Ìàƒa{ @å 4P㌴gÄ ‰´ f@àO’tΰCd˜,ÌKøèlU¢ šþ½1Ñi"gz… ?ZÕÓ^ !´lHý) •ö–·ûÇB´”¡BŒ3 áš…i ÄO84`ü„Ðÿî½òÉLE<*Œ áçye¡ççû›ÁíõÊll¯üÏ·<ŽnÀ-»”OsÄ„‚*|p‹L&áNíˆ]îÔîÑ7Ì‹òVù£NO9Hû‚k¥ù¥­pD&=^ñ¤ËëUuY~tz=zl0§':fÔ¥Ù^Wúâóg»¦ÉŸ¶uÉ©mµ+9gN0Þ0ÔVÑøL{ÇÀ“¾ºß?ÃÄY$”ʾïÎuŽ#0w.á]¦÷¨UQؾ ýɵv)<òÆŽß­±Ç›@ÔáòfÒ¨ÃGPeÔ¡ Øq(Ê'Øx£›ˆm ÝÊÔµVETÕ3É]àÉ2…¢Ö]E?"2NLô6C¥ýˆÂY¸ðÃ÷ÝyÙqdˆÈ%ü ¢q¼ "—7ÓÎø!‘¸H¤Ó¢7µ0†¦ïžë…ú¤IÊ÷B"·“ƒ šˆ…® (ÿG&vH&6þøÍY™XÏÄCð †gÆY˜žq-ÌGŸÒ q÷áäAB;¡çíÉ…¡ËÞ“¹ÑK§TïÄ­‰á6Ÿôm¦Ï…êÄ#)ã¡Eû—t·AA †íwµß)iƒ {¹Ä˜ìyç´\"îR€â£Á€Bv9¥cÛ±0þ€W%kÛ:ˆåh˜´-©9JÑ.‚07+Z‡u‘ãã=ä慎ï”´!²î—µÚÔ6YN›¦ù>ùØbÇÇPóé´bpº 3^ ðA«äh= Óš¤F†IÊ”ÙB¶ÿ&r& ðãõ=.²M`ÀY°#š€…ÔŒSÇððd¤YÆÃ×,O¼€tm¸Da»•›ÚÛ_àÛ† óîÑCäf‡ÒS­EfCéÑÐ[5Ê:u¢ïÈÒ:ˆ-Ôv6]Û=^¡F ¸iøÜL#p¦N¡‡èr!€z?3;žL¹_B*ÅeiÕ¹À'Õ“Š­t»Ÿ(S"¢ suŠ$¸C/ÅKX2DëAoí$YV%RÐ.™<í‰S¹AEÐUE? @VɽîÔs]¬#Tånuݱª[ÝÇË4zéðèÑu¿ãõ@§ó Ô¤8o~\Ä ô5#SÕ»å[@•¢::ø»îß#É"=—ð&¿ÞU¹àM~@> ¸à}n»œ.à;‚¥n>h¬:Þ&Ýäо ç(­{²UC·dÞñŽ“!§qô8q hWÜïn¡õ¾Ø ¼¤Ü¬àQ,Óܤ“ :,iiÔë~&ôi³Ö–7œÒ‹¢pªîÍU‚1Ö9Öº˜'0MuJï8qŠ-Ì$ÁßÄÇÓPÓØ‹yÔ°–Ksj˜Sˆ`ËÌÓ(åT‡^.ÚµM@M¢… =ÖTCPÉÃBÜvªÁ#”KE'åÀXwѺ®8mÂâK4 ÎnÐ<¾6&ø Z‰³B>°/š»LÙtpk\MJ‰™_¸è;0— W/ò£Q\£ožiƒ εA,i¸Û% !ÐÉûAçÉ2€F`õ™‹2í@¢z&:Rˆµ{— É"*lÈþ$ßÛ‹bÙ;ÇQàN—ö•¥Ï0èuýïè`¡â×Ñâp;úm"×óIKd³Òh¯S]êNõTƒ“¥Å° ø[03B‚ ˆÎÕ–€HWö§®J€\Ê21 g[Â¥»Œ½Æjí=lÜì•eÒ»šÐ)õ5Šúú‡_Ÿ!©+ãšñË.g¬ëºôa°*¼ƒSÃoS5w~b—©Òá·©žn“³ÆŒ>Ç4ò45™° ®ß„GŒ¥ÃnSI7Ô»”{oSI:>º¤¯·©NÂѹ}AhÇeª§ÓEä?1·½Ÿ¥C‰ÁRSèOt*ÒfOŸ.³UÇ ª‘dqòï aÿÎ:÷žò€^k®W^{ô¶ªÆq·ÜH‹ ’>VXƒ+Ö,­눥dk.¶À3yúÞ®¸c²VÏ“[ª–> é,õTž¤ÿ4ÔÞ³Î9Ôȹ ž^¯;tsÍ€jºT³ {ÎÙõÕ€ # ¨£’ƒê)oÌ?Áv¿è;ýÌr-N”ŸfieUª•‡4ëÚžC±H½ŽaÆÕ,Õ‚mÙøäpË0j ñû<븱ïdš¨*JuÞÕå\ê4ÚÚ“(©.,{²Ûvæ5Šf@›]ªëëÕ¡¸Z½¯ÊcÐä85ÛKLþïÑädZMæƒ4™ššÚt¾HRUh/*J› ”,°2ÉÉsaüœ å8Æ:s`zf¥76ÜbD÷œñù×µr^gfË ó[Ëé‹ü!lÓcѬ$#î%²ä°g ®#Ë2p înM´ÚÈîùç¤ø±ÆBHmdÄRú‘…rˆEæYätEžÄ Ë»6ÈrBƒ¬·Öl¨ípjç àõ&_¶»ûRÎ7ÙêÚÜ‘–ê ÒÕu5-ѲE†©1³'©G2‚éDZ¸–êývë½míeùê\Ýf‹ÿk /3Û£üR·eÅÌ´)|ôÒ¬Iš-@þ àW¥w’Í¿ëƒé·Œç×û’Cðx0Í€‡ïòÛÃv³ø¦ØéëœÌG—ÛÝúz•åÚØòdþ4[_­ò½ý®6ŸæíÅ.»ÝØoJ› OYùòÍ“òõ¶8¼¶_4Þ-·f_ŽÎϸþÔfólgøek ÿkßÛú,J,n6{=Ý®d(èá…yÒ*ÑŸš²_òMù…‘¶òD£¤Êæžá‹íòókÝnhtc­ÑèÔ”ÕáO±Z]ï»ì`[S2h¶å×{Ûº}¡W~˜m.Ê—ΦÓÍ‘…UQ®½€hHdÁƒykèéÌöU¾,.ßz rÁëXµŠj=JMN¸dçÛÍA ý2[¥HY•>Ϻ[ LË ±,ƒ–hþ[ßd×ä#‹}•¡/Á˜•—o±ú&U¶ýVð–0` Ïÿò°|ý“náŸÊç•/ÿ¡?*?ÿ\¿ÆfHô@— GÛ<–{uþßòæ7§Æ>Ð?¾y`ß]eúí.[Ûôïéü·ßÊwÿª+yP=]Ú'“ÑËËß~s[¨ùÀB –D\ľ€Q™Á‹çÅÁ¾îË×lkiîJ³«$±;kè´‰(6õ{kâù&3ºÅKU¶¿Ž[bÑØû^©«­±©¶ÒmYSG ¨¡7ÒEÒùvg_7ž}I«ö¦Æ%å¶TÑ,0xYl²¦ëhà)+_–ùjU»]öÖ±uûíLÌ£Ñg]—­½4¦&ôG:³*öß6^e 79pioÔ󙽂}S¿š¡×‹¯õ 0ò`¯¢F^%³ó ,Á"A[ÒjÚë²HLƒùÏ‚!pŒ`CBÛ¸¹' ÑoC·¨ÏÔ(“WdöˬzúϲwP>f&‡·HY”轟¥£q746|‰ä,¢³~[>þ`¸$z 1ª™Óø·t;{¾ˆ%›ÃP p6æÿÚüßšÿóoþÌÿùŸÕŸ¤ó|’FÄéüªã†˜0Äü•ù? \Ð'.Ëëùí%s0•BYFÝÔlQ–E…èù|i™<y M¥×-ÊŒŠ²ä·æ¹üvï"³VØ<™eÓÊìØfðCu@^Ÿ‚P¥FuÈT’ZŠ ›¯Ìÿ½ù¿Fj+…[£º¥Ô~mžAJm™C-Ç©cÒ– ‰)1†•”|¾X’µê@L-?ŽOóRˆÊ.â¯Nù*}Ký›¯žÝûôkÓ£ endstream endobj 1064 0 obj << /Type /Page /Contents 1065 0 R /Resources 1063 0 R /MediaBox [0 0 612 792] /Parent 946 0 R /Annots [ 1060 0 R 1061 0 R ] >> endobj 1060 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [39.433 111.738 51.396 120.177] /A << /S /GoTo /D (noweb.NW2i2LRH-3dLroe-1) >> >> endobj 1061 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.862 109.1 166.825 122.166] /A << /S /GoTo /D (noweb.NW2i2LRH-3dLroe-1) >> >> endobj 1066 0 obj << /D [1064 0 R /XYZ 125.795 791.866 null] >> endobj 130 0 obj << /D [1064 0 R /XYZ 72 754.004 null] >> endobj 134 0 obj << /D [1064 0 R /XYZ 72 290.162 null] >> endobj 362 0 obj << /D [1064 0 R /XYZ 81.963 112.824 null] >> endobj 1069 0 obj << /D [1064 0 R /XYZ 72 110.097 null] >> endobj 1070 0 obj << /D [1064 0 R /XYZ 83.504 112.824 null] >> endobj 1071 0 obj << /D [1064 0 R /XYZ 83.504 100.869 null] >> endobj 1072 0 obj << /D [1064 0 R /XYZ 83.504 88.913 null] >> endobj 1073 0 obj << /D [1064 0 R /XYZ 83.504 76.958 null] >> endobj 1063 0 obj << /Font << /F59 148 0 R /F57 146 0 R /F70 201 0 R /F79 203 0 R /F80 204 0 R /F73 223 0 R /F33 572 0 R /F18 190 0 R /F77 551 0 R /F136 574 0 R /F74 202 0 R /F144 1067 0 R /F83 207 0 R /F81 206 0 R /F148 1068 0 R /F133 573 0 R /F82 552 0 R /F85 678 0 R /F61 150 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 1076 0 obj << /Length 3403 /Filter /FlateDecode >> stream xÚí]Io9¾çWèÒ˜20RŠdqKÐ@÷Lt'Ý“N2'OjI¶…È–Z’AþûdéÕ"©Êc 4] ·zË÷Òéàr^>JýobJûóqj¯ýãý£Ç/¸è‘T Þ_ $i• dÆFiÆ遼ó„Ÿ}xÿkÞÎã’¡H:âÔÜ€'¯ì“_‚`#%Exàru½8R©’Érõe=¿¼ÚºWÌ ”yEdö¡F©ÎÂ;|ì)t«G™Ðá‰9 °n‚¥Ù 5i®Cbûé‰ÂôÒÁˆ FÙ`mÿtÕ·vÒæé¤5Ï'59fYb敱dõ1”ë ·‚Sû£¶ FL‰_3è4ù»¿¶¯‰LF±„úÊYÃQˆ†Â(tiMÖÖ/Jame‰tÊ«(UN;7gCÁÒäGûKš2ž<µèäYá÷>‡[G×gfû2õ)”7åå–iòŸ”§þæP³úCóžAÝÜ"ù2e`þظ¡†‚a0O »a°ša? CÌ6ÃPÕaø.¹íÊu Õ·/ñG¯¹kú ”ˆä‡2tiIdÚS  >à0Mžlý•HÝŒã4ÄaáõL¹õõMsméÁç øiÛ…“,c8ImÅ…7tƒÉäÈH„Á¬Fç%&É-ÿ*Ns „%7šÄ–×@È+¨/"¯‘,–GÈg•ÜAiéÏI'݇+7PwÏGêž&Ðôm© ä/¡{ ƒ w7íiWŒ-Ñ.Iª1ÊR”©ù¬¦wM?:ñþÞw“qSËë~n$ù[¡CÏf~Òv¢¾‚L~GWÜ2l >kÚª¢¥˜kœ%ßvºštssºxc\A€CžÚY3a`”F_ \AißÉÄqý¼ö^ByY:ŬW>§ï˜6a~ 7^tÄhT¸•Õ°²VÖI% 쮌ÑW¿FOn ouàKëŽ;„çÒ–;&eîJ–üT»“/ÒSö?k˜Ú³¼¡Ùœ>‰J©§‰ ­<Ü¢ùÔ@Ô*iNÎ@™6#MCêIó¢9iÚ÷ÍPˆ¥[.€÷wbRTĤÈW4’b¸‚FÇPº.§¹þ@o¢${À“ÁŠ¾êG¾A¸Ó¶ä»¨’/òuË,`æÉk¨ÿ¥ÒŠìHV‘Öä"ÈJ ºs ·^¾&!óx40¼@x⦉×q¾$Ð"˜Qšã°‚ÌIkT¿Î[LêMæa5¡É°Ù£B÷QèžB7´$ÿ;S*ë`é›vvþ‰7­W$˜YKtT7À+´5^)Jõ×Pþ’cŒ#¥±P‹‘ª¯šØ²q@,¦œVIíŒkcÉJi­Zó ‘ñY°jÍŸ¶k3ï+ ²/9þ±³½£Á!cÝÛYÒ„Ô½H#°`%š±(O~†} Œð¬ñ[°º]8™— Šì˜Î©çÙåY‘K‡2Å£*v™ã0¦u <†šÑ“õáÍPG‹]Gã›;:`|a N=egn„²e›îª”Íš»q¹Ròï–ÊDùâeŽ|>Wqš¦ÉŸ¨i'7qˆ·X« À2b÷§€ÖM¹„òÊ7èÊ"€)B…Ì Êo¹TÎÀüÈݱìƒÌñ¯Ac«8æaTï åÒ2£0šêiÉ`È,ÔËr³A÷$q©cYÞzp:-2aÞ'=CÜ»ýn›zçm ¢[ÌÒ"­DAf¢­<¿ô¬^žKäá òü”Ï¡Ä>·iÙ&uÒù2DÚJçZbæeb~ ä)`13`QÚ˜3’ÊX&rš¼Ì2Êä6›£»7±œ7§’ÝÆ ÞgZKíàÕ‘ÝHz~vjCB ‰wpå9òÿnü3}LÌP”e{jî”´ƒÌ|°!=‚Ÿ˜¹(3LÍÿ4G µDž§5²Wè“MbÅ^æ^(%ˆ«ƒãk\ãøêŒmT7.¹=»_s+9r¢ºgþÍÜæÂãƒ/À0˜UnOh8Vq\Ã2šŽbYa‰ b~‰2nQFE$oøy/IÚ͹ÑMP‚Á¢&¤3«ê¶¬º®Åh§VhoáÊ%œ'zúÄ°­ïœmΑ"yÞ:ÀÅ뮲Âó²b™º»’×h)¬öîÙ)¾®d…8ƒéÍþFŽ$I’Q¬¿FÞ"›}ÚØM=ŠÂV4íÆVó¦Ï[(ÿ(a<ÐpØôáŠSD¸™¬Û#µÆv¼äe÷”QKõDLÇÅnF]ÉûWWª»u&£ÖAäëS¦X„¡´Otù-¨»%Q(öîܪa+ÝÂaU­ °`¦é²ªsR]·qRõ1CÄ©V”¦ñÄõo5I|ÎÛT\7²ëbY ÌÙ˜8Ë°Ø_ e:±Ì4#¯ƒÒ¶’ý(áæžÓï!¶^Hé›úD÷Y5Áp2qcDòyÔ0·µB? h‹Èxܶ‹z€±—óÙØçDá6ÜÀÔί’.VþG'±¶ °˜x%Ò§!┑*äa.ƒ8¿2飒Êـˠ†ë’­êx&X³û†b»‘äFËj8êƒaÑxXk•ÓrÀW87kŠ\Ñ‘PVë¸óÖdrÏ"ý \yש›C*Z†Tu¤: ÎGªR¼ÔñÔ¾4^Üb’œ²š„'¢1Ê?ÛÆ’µ›<<ºYUdí6VôïXp®GÔ|ŠÀß%RM³˜;;¬ô‹Â¤¼õV *:ì•°‰îäûÛ+!RäaÀCÊû_—¥XÌí¼Hቲc^ysrÜ¿ãÜœ”§Ø=¡PZ°DiÁª²{BvÝ=AJÄ)ã¤-n¼*ºß´Åºhã9gým i÷G©€D±‹™Ò|+©AUA ðdH™®ósmÊ˨ó#Kv(²C"M1ÑŠöHÐêAeimnßbèÃøfGq<­£Äòië(åæäQJŠ ÷¹òï ¼õ{ØšrTz˜£8|o Ñ*ÃKn/¬»2 W\:Vîœuœöíp¾ã&ÄÕœýïÙÍÂÁúíO*¤ŸÕoAÑûo/ Ï8/1NyTMïh„éÒ×êŒ|t7ƒ{]Ž%˜Õ<œ.¸†tí¥˜˜Aho|=ó¤Y ^¶†Œl+ Q„ïµx×;*1<‡–¿3Ád kkn‘Hƒ2’ÒxfÉ YK´ýçv¯YÒ•HØÑ€“TH® ɬ¶üˆ6×Ø\£’4äñ›ºQnRøzšo½9nâ™Æú¯åqe¨Ò òt‹™ÆcdÒD³™êÖ.’Ãܹª>#º@ûz§:zkNãBÂHD•¯½ÜÆÚµ…æ®"î5r„6Ž#¸ÌŽéý\î#ÏÐÏn’Âð}QR¶êäú䬲EP@˜*8p65þënàG‚.#vNŒûس $¨ÅN¯ð·³7,§@5ó„É·J p¨.r]ÜØ´G‡åY“28Z;«ú0\<õ†à´â³Ý¢ú§ñ( %Aƒ3Ÿ5I¨-ýîúHͽƒ1>Ï·ø¼Œå<ÄùmstFÎeÜnPrÚÚSw$9F!ÔiÌ !}ŽÂ ŸJEÈxN¦9J‡¨¦I¸ &(‰bî“%voÅq;±ÖÑ»±_ûû5z†ò™/{Èý<¾ñ5„ÄLIμ>Èãü•#ó’iÏIv)Ç¢”SÔtŽ¤&)KΰN*Ä»EXÖÚÑÎøÃÐ:f\à³²%µçBÅ$…§V“tgJÄw’ÔlX—عÂÃ9?'8HHT\á¢ä "Zïs.Ù»ÂFÜ¥:iåÜó†Â—àÉ_Q5Ž«ë€úc0ÙžÔƒa ™dã†TûÙ)0BnB–I”F‰÷!Üæ'‡qïa˜Å=éc铘"zU‘cd^ÎPâ*NŒ‰-t·ˆ–Îܼ@˜ý®Øw˜cn«¹l˜;8Jæ+øež -Î997åi-ŠJ1K[ótF OÓÈ<­½"¼C8ûªÆÞ±ѸsÔ¤Q}?¦ÑòÉy§>Šë¨6$³þ¼›Ñ‡À»2ó(ýIÈEr(]€Ó„PåP!dõBYÝù;ÌŸµçïìÁð7û?'üÝÃÍ=ièf}#9Äh~Nr¼³4âm¤Þ;‹ƒp1Op1‡ÞÀBd²Ñ"Ø`UÚøÐ4™¢SÙj¾öY-½¤~@5ÇΠAÂꎩþbxbxÄ°Á¯òó²‰u¥4<¼ûÙwh[î!Êvà^îx±dFv:Ãtò9ÿ-ƒ‡T´G‰U-*Úkù`´HÝu‚ƒ:»¬ÜþÞÖäˆõÓ±­ÀÌG]±«v^:ä;lÃõYUÕCŒ}j&.Ì“£†O6>â±U›Û¨m6¢Ù˜ Ãç3ê á0ƒñUsi?/üo†Òáûþá…øŸ tC„s0 j¤Xñ7,˜ýß©&vú”ÈQ–q³bZXVŸzõüý£ÿßÊ3 endstream endobj 1075 0 obj << /Type /Page /Contents 1076 0 R /Resources 1074 0 R /MediaBox [0 0 612 792] /Parent 946 0 R /Annots [ 1062 0 R ] >> endobj 1062 0 obj << /Type /Annot /Subtype /Link /Border[0 0 1]/H/I/C[1 0 0] /Rect [154.478 739.559 165.939 751.663] /A << /S /GoTo /D (noweb.NW2i2LRH-3Gkbo6-1) >> >> endobj 1077 0 obj << /D [1075 0 R /XYZ 71 791.866 null] >> endobj 1078 0 obj << /D [1075 0 R /XYZ 83.504 755.001 null] >> endobj 1079 0 obj << /D [1075 0 R /XYZ 83.504 743.045 null] >> endobj 1080 0 obj << /D [1075 0 R /XYZ 83.504 731.09 null] >> endobj 1081 0 obj << /D [1075 0 R /XYZ 83.504 719.135 null] >> endobj 1082 0 obj << /D [1075 0 R /XYZ 83.504 707.18 null] >> endobj 1083 0 obj << /D [1075 0 R /XYZ 83.504 695.225 null] >> endobj 1084 0 obj << /D [1075 0 R /XYZ 83.504 683.27 null] >> endobj 1085 0 obj << /D [1075 0 R /XYZ 83.504 671.314 null] >> endobj 1086 0 obj << /D [1075 0 R /XYZ 83.504 659.359 null] >> endobj 1087 0 obj << /D [1075 0 R /XYZ 83.504 647.404 null] >> endobj 1088 0 obj << /D [1075 0 R /XYZ 83.504 635.449 null] >> endobj 1089 0 obj << /D [1075 0 R /XYZ 83.504 623.494 null] >> endobj 1090 0 obj << /D [1075 0 R /XYZ 83.504 611.539 null] >> endobj 1091 0 obj << /D [1075 0 R /XYZ 83.504 599.583 null] >> endobj 1092 0 obj << /D [1075 0 R /XYZ 83.504 587.628 null] >> endobj 1093 0 obj << /D [1075 0 R /XYZ 83.504 575.673 null] >> endobj 1094 0 obj << /D [1075 0 R /XYZ 83.504 563.718 null] >> endobj 1095 0 obj << /D [1075 0 R /XYZ 83.504 551.763 null] >> endobj 1096 0 obj << /D [1075 0 R /XYZ 83.504 539.808 null] >> endobj 1097 0 obj << /D [1075 0 R /XYZ 83.504 527.852 null] >> endobj 1098 0 obj << /D [1075 0 R /XYZ 83.504 515.897 null] >> endobj 1099 0 obj << /D [1075 0 R /XYZ 83.504 503.942 null] >> endobj 1100 0 obj << /D [1075 0 R /XYZ 83.504 491.987 null] >> endobj 1101 0 obj << /D [1075 0 R /XYZ 83.504 480.032 null] >> endobj 1102 0 obj << /D [1075 0 R /XYZ 83.504 468.077 null] >> endobj 1103 0 obj << /D [1075 0 R /XYZ 83.504 456.121 null] >> endobj 1104 0 obj << /D [1075 0 R /XYZ 83.504 444.166 null] >> endobj 1105 0 obj << /D [1075 0 R /XYZ 83.504 432.211 null] >> endobj 1106 0 obj << /D [1075 0 R /XYZ 83.504 420.256 null] >> endobj 1107 0 obj << /D [1075 0 R /XYZ 83.504 408.301 null] >> endobj 1108 0 obj << /D [1075 0 R /XYZ 83.504 396.346 null] >> endobj 1109 0 obj << /D [1075 0 R /XYZ 83.504 384.39 null] >> endobj 1110 0 obj << /D [1075 0 R /XYZ 83.504 372.435 null] >> endobj 1111 0 obj << /D [1075 0 R /XYZ 83.504 360.48 null] >> endobj 1112 0 obj << /D [1075 0 R /XYZ 83.504 348.525 null] >> endobj 1113 0 obj << /D [1075 0 R /XYZ 83.504 336.57 null] >> endobj 1114 0 obj << /D [1075 0 R /XYZ 83.504 324.615 null] >> endobj 1115 0 obj << /D [1075 0 R /XYZ 83.504 312.659 null] >> endobj 1116 0 obj << /D [1075 0 R /XYZ 83.504 300.704 null] >> endobj 1117 0 obj << /D [1075 0 R /XYZ 83.504 288.749 null] >> endobj 1118 0 obj << /D [1075 0 R /XYZ 83.504 276.794 null] >> endobj 1119 0 obj << /D [1075 0 R /XYZ 83.504 264.839 null] >> endobj 1120 0 obj << /D [1075 0 R /XYZ 83.504 252.884 null] >> endobj 1121 0 obj << /D [1075 0 R /XYZ 83.504 240.928 null] >> endobj 1074 0 obj << /Font << /F57 146 0 R /F73 223 0 R /F61 150 0 R /F55 149 0 R >> /ProcSet [ /PDF /Text ] >> endobj 1123 0 obj [531.1] endobj 1124 0 obj [516.7] endobj 1125 0 obj [999.5 748.3 794.7 744 1060.5 547.8 632.9 990.9 869.4 1171 935.7 860.6 807.5 809.6 877.7 673.5 724.2 867.9 812.1 1178.9 871 734.3 803.2 0 0 0 936.9 936.9 0 0 0 0 0 0 422.7 422.7 0 0 302.1] endobj 1126 0 obj [847.7 652.5 845 712.6 794.4 973.6 781.7 959.9 714.7 700.1 646 584.1 587.7 523.7 651.6 648.9 354.1 583.2 596.3 646.3 638.5 588 664.4 610.9 700.1 580.8 621.9 873.6 587.7 821.8] endobj 1127 0 obj [990.3 777.2 987.4 842.8 932.2 1127.7 918.2 1112.7 845.1 829.2 770.1 702.5 706.4 636.6 776.3 773.3 451.5 701.6 715.8 770.4 762 706.8 790.2 731.8 829.2 699 743.8 1018.6 706.4 962] endobj 1128 0 obj [666.7 611.1 777.8 777.8 388.9 500 777.8 666.7 944.4 722.2 777.8 611.1 777.8 722.2 555.6 666.7 722.2 722.2] endobj 1129 0 obj [777.8 777.8 777.8 777.8 777.8 777.8 777.8 777.8 944.4 500 722.2 777.8 777.8 500 500 722.2 722.2 722.2 777.8 777.8 777.8 777.8 777.8 750 1000 1000 833.3] endobj 1130 0 obj [805.6 805.6 0 0 0 0 0 0 0 0 1000 1000 1000 1000 0 0 0 0 1000 1000 500 500 1000 1000 1000 0 1000 1000 666.7 666.7 1000 1000 1000 0 0 1000 0 0 0 0 0 0 888.9 888.9 888.9 888.9 888.9 888.9 888.9 0 0 0 0 0 0 0 0 0 527.8 444.4 0 0 0 0 0 0 1055.6 1055.6 527.8 0 0 0 0 0 1444.4 1411.1 444.4 0 0 0 0 0 1055.6 1411.1 0 0 0 0 0 0 0 0 0 0 500 666.7 0 0 0 0 0 0 0 0 0 0 0 0 450 450 450 450] endobj 1131 0 obj [585.3 892.9 585.3 892.9 892.9 892.9 892.9 892.9 892.9 892.9 1138.9 585.3 585.3 892.9 892.9 892.9 892.9 892.9 892.9 892.9 892.9 892.9 892.9 892.9 892.9 1138.9 1138.9 892.9 892.9 1138.9 1138.9 585.3 585.3 1138.9 1138.9 1138.9 892.9 1138.9 1138.9 708.3 708.3 1138.9 1138.9 1138.9 892.9 329.4 1138.9 769.8 769.8 1015.9 1015.9 0 0 646.8 646.8 769.8 585.3 831.4 831.4 892.9 892.9 708.3 917.6 753.4 620.2 889.5 616.1 818.4 688.5 978.6 646.5 782.1 871.7 791.7 1342.7 935.6 905.8 809.2 935.9 981 702.2 647.8 717.8 719.9 1135.1 818.9 764.4 823.1 769.8 769.8 769.8 769.8 769.8 708.3 708.3 523.8 523.8 523.8 523.8 585.3 585.3 462.3 462.3 339.3 585.3 585.3 708.3 585.3 339.3 938.5] endobj 1132 0 obj [277.8 777.8 500 777.8 500 777.8 777.8 777.8 777.8 777.8 777.8 777.8 1000 500 500 777.8 777.8 777.8 777.8 777.8 777.8 777.8 777.8 777.8 777.8 777.8 777.8 1000 1000 777.8 777.8 1000 1000 500 500 1000 1000 1000 777.8 1000 1000 611.1 611.1 1000 1000 1000 777.8 275 1000 666.7 666.7 888.9 888.9 0 0 555.6 555.6 666.7 500 722.2 722.2 777.8 777.8 611.1 798.5 656.8 526.5 771.4 527.8 718.7 594.9 844.5 544.5 677.8 762 689.7 1200.9 820.5 796.1 695.6 816.7 847.5 605.6 544.6 625.8 612.8 987.8 713.3 668.3 724.7 666.7 666.7 666.7 666.7 666.7 611.1 611.1 444.4 444.4 444.4 444.4 500 500 388.9 388.9 277.8 500 500 611.1 500 277.8 833.3] endobj 1134 0 obj [507.5 507.5 507.5 507.5 507.5 507.5 507.5 507.5 507.5 507.5] endobj 1135 0 obj [478.7 478.7 358.7 876.3 358.7 876.3 358.7 601.1 601.1 601.1 601.1 601.1 601.1 601.1 601.1 601.1 601.1 601.1 292.1 292.1 0 876.3] endobj 1136 0 obj [458.3 458.3 416.7 416.7 472.2 472.2 472.2 472.2 583.3 583.3 472.2 472.2 333.3 555.6 577.8 577.8 597.2 597.2 736.1 736.1 527.8 527.8 583.3 583.3 583.3 583.3 750 750 750 750 1044.4 1044.4 791.7 791.7 583.3 583.3 638.9 638.9 638.9 638.9 805.6 805.6 805.6 805.6 1277.8 1277.8 811.1 811.1 875 875 666.7 666.7 666.7 666.7 666.7 666.7 888.9 888.9 888.9 888.9 888.9 888.9 888.9 666.7 875 875 875 875 611.1 611.1 833.3 1111.1 472.2 555.6 1111.1 1511.1 1111.1 1511.1 1111.1 1511.1 1055.6 944.4 472.2 833.3 833.3 833.3 833.3 833.3 1444.4 1277.8 555.6 1111.1 1111.1 1111.1 1111.1 1111.1 944.4 1277.8 555.6 1000 1444.4 555.6 1000 1444.4 472.2 472.2 527.8 527.8 527.8 527.8 666.7 666.7 1000 1000] endobj 1137 0 obj [830.5 608.2 649.3 604.4 884.5 430.8 506.1 822.9 715.4 982.3 774.1 707.6 660.6 662.5 722.7 542.1 586.9 714.1 664.7 989.3 716.8 595.8 656.8 0 0 0 775.1 775.1 0 0 0 0 0 0 320 320 0 0 213.4] endobj 1138 0 obj [388.6 388.6 277.6 756.2 277.6 756.2 277.6 501.8 501.8 501.8 501.8 501.8 501.8 501.8 501.8 501.8 501.8 501.8 216.1 216.1 0 756.2 0 362.4 0 717.6 884 612.5 831.6 662.8 611.2 785.4 720.3 554.5 552.3 668.4 666 1049.5 832.1 827 827.5 827 828.1 828.6 669 645.8 831.3 1046 719.2 832.9 602 277.6 0 277.6] endobj 1139 0 obj [545.3 867 917.1 932 734.6 877.7 791.6 870.4 1002.5 855.7 1050.4 805.1 809.6 748.6 627.1 682.7 639.6 694 687.3 439.5 688.5 673.2 759 738.2 686 749.2 680.2 744.9 641.8 724.5 883.2 712.6 913.4 1022.6 610.7 684.8 1050.7 0 0 923.5 0 0 0 0 0 0 0 0 625.6 625.6 625.6 625.6 625.6 625.6 625.6 625.6 625.6 625.6 374.8 374.8 915.5 628.1 915.5 0 694.3 931.7 801.7 868.5 997 744.6 625.3 925.9 946 506.6 515.4 816.7 692.7 1240.8 998.5 968.6 712 997.6 749.2 690.6 617.9 936 791.2 1175.7 813.6 688.3 814.2 0 0 0 0 0 499.9 749.2 726 611.3 743.4 625.6 535.6 703.7 763.5 467.8 436.1 689.1 473.9 1095.9 811.8 697.6 727.2 745.2 549.3 576.2 531.9 787 621 978.7 655.5 732.1 593] endobj 1140 0 obj [428.6 713.3 757.6 770.8 596.1 722.7 646.6 716.3 833.2 703.3 875.6 658.5 662.5 608.5 501 550.2 512.1 560.1 554.2 334.9 555.3 541.8 617.7 599.3 553.1 609 548 605.3 514 587.2 727.6 576.6 754.3 851 486.4 552 875.9 0 0 763.2 0 0 0 0 0 0 0 0 499.6 499.6 499.6 499.6 499.6 499.6 499.6 499.6 499.6 499.6 277.6 277.6 756.2 501.8 756.2 0 560.4 770.5 655.5 714.6 828.3 605 499.4 765.4 783.2 394.3 402.1 668.7 559.1 1044.1 829.7 803.2 576.1 828.9 609 557.2 492.9 774.3 646.2 986.5 666 555.1 666.6 0 0 0 0 0 388.4 609 588.5 487 603.9 499.6 420 568.8 621.7 360 331.9 555.8 365.4 915.8 664.4 563.4 589.6 605.5 432.1 455.9 416.7 642.5 495.6 812.1 526.1 593.9 470.7] endobj 1141 0 obj [516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7 516.7] endobj 1143 0 obj [556 556 167 278 667 278 278 278 0 278 556 0 722 444 278 278 0 0 0 0 0 0 0 0 0 0 0 0 278 278 278 333 556 556 556 1000 944 278 333 333 556 556 278 333 278 278 556 556 556 556 556 556 556 556 556 556 278 278 556 556 556 444 800 722 667 722 833 667 556 833 833 389 389 722 667 944 833 889 611 889 722 500 667 778 722 1000 778 667 722 333 278 333 556 500 278 500 556 444 556 444 333 500 556 278 278 556 278 833 556 556 556 556 389 389 333 556 500 778 556 500 444] endobj 1144 0 obj [524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9 524.9] endobj 1145 0 obj [278 333 333 556 556 278 333 278 389 556 556 556 556 556 556 556 556 556 556 278 278 556 556 556 444 800 722 667 722 778 611 556 778 833 333 389 722 611 1000 833 834 556 834 722 500 722 778 722 1000 778 722 722 333 389 333 556 500 278 500 556 444 556 444 333 500 556 278 278 556 278 833 556 556 556 556 389 389 333 556 500 778 556 500] endobj 1146 0 obj [501.7 501.7 501.7 501.7 501.7 501.7 501.7 501.7 501.7 501.7] endobj 1147 0 obj [556 556 167 278 611 278 278 278 0 278 556 0 722 444 278 278 0 0 0 0 0 0 0 0 0 0 0 0 278 278 278 333 556 556 556 1000 944 278 333 333 556 556 278 333 278 278 556 556 556 556 556 556 556 556 556 556 278 278 556 556 556 444 800 721 611 722 778 611 557 778 834 332 389 722 611 1000 833 833 556 833 722 500 666 722 723 1000 778 722 722 333 278 333 556 500 278 500 556 444 556 444 333 500 556 278 278 556 278 832 556 556 556 556 389 389 333 556 500 778 556 500 444 333 222 333 556 0 0 0 278 556 556 1000 556 556 278 1000 500 278 1000 0 0 0 0 0 0 556 556 500 500 1000 278 990 389 278 778 0 0 722 0 333 556 556 556 556 222 556 278 800 325 556 556 333 800 278 400 556 361 361 278 556 556 278 278 361 361 556 834 834 834 444 721 721 721 721 721 721 1000 722 611 611 611 611 332 332 332 332 778 833 833 833 833 833 833 556 833 722 722 722 722 722 556 556 500 500 500 500 500 500 722 444 444 444] endobj 1148 0 obj << /Length1 2160 /Length2 9102 /Length3 0 /Length 10266 /Filter /FlateDecode >> stream xÚwPÛÝ-nÅ] „Å%¸S\Š»—‚ îîR´‚[ñâ-PÜ -.Å¥¸wyôÓöû¿™÷&3IîžÝsÏÞݽ¿„žFM“MÂj–…:ÀØ€ìœB)e= '€““›““ ƒž^ ³ÿmÇ ×;»@ B¿xH9ƒA°›4öਠu(ºÚ€Ü Ÿ_ˆ“ÀÅÉ)ø·#ÔY rƒX”ÙŠP° ½ÔÑÓbe {Øçï¯Fs&PPŸõp€„=Øbr(ƒ`Ö`û‡ÍAvM¨9 óü‚QÄsâàpwwgÙ»°C­Ä˜Xî˜5@ìvv[~¦ PÙƒÿJƒ e qùЄZÂÜAÎ`ÀƒÁbvpyqu°;vh*(TÁ:+ýéÀ øëp@và?tEÿ$‚8ü 27‡Ú;‚ŒˆÃ]uÁN´2Øâjÿ_Tz +»â" ñ[¨A`æÖöÑßux ·ƒ8€Õ .Ÿw€ ÈÉùìaìÌmî—‡jý¦ê÷-eÌ¡?Ç‹—rvyb<ÿaÅ ð>Ì©Øãöp°;@a!€‡ô|–PgŒŸ5rò8ÌœAæ¶à‡ËÑ& ±úéðÆý?пço˜Ù¯/×ïì_äwÆiÿ ¿óÁÿA€œ¼ÿ@´ÓoÿûE#ðwðw‘ÜÿQùÒ_eþ‡ò©YÚA¡ÎÇ¿!Ùÿ–ðob|LJ‹ÙáÎëwà_„óà7*ß>ž_‘säûuó_äù•é×*ðÿeÿßpÿùEÖ¯Èozÿƒü"˜÷7èÅ¿)øUòod¿j~xps8ƒ,~^,¿*¹þ±ÿ­ê·¡2wu~ ƒýqñ=LÜßë?p`°Øczj.bSòé¢R‚ÂmmPtŒ~M÷›÷´s³ë6j Óûô Eç3‰”¾ܹÆÓg3Ô·Þ; µ¨áIêM×>7/^jŒ¬5aL ÷ åïHÔtS¢?fÓz¶îsëä£h‹ØߪHŸíä*€­–KpáÞ%çQÓ]2û5lrM}ý=ßóG7%£l±Ú1FïÆésÌ2&HŸ ÀØ(јñF {j¤¯åMëV†eßø„"ÃËß(ç·˜ƒÉÄmØ]ú†?†»;ƒXè©«JŽ±ä*º"ïQþàÛu#þ÷~E8lYUd[&)„BÛÀ$¯Ò×Ìüz®nîoº°bÞM‡]X=Œ.Ó¾Ù'd‘.ו(è^Z¯¡Ú£\íPËðΣ7vÎÀ{–¿z—…—#‡šR±°·ægy¥È²#0uéí–,«|’{áñýÓ·~eȉÛîoÜʾ÷Ž‡ßëvîºvàà0íõlÉÑm‘â'¶ºyDÒ™± ¤=†_t¡Žv+&Œ{?„‰F݃–EtoUõæpÑ‹ˆôßBPf‰Ýsz¿-?KB4O+fP ÙK¾jÅÏõüXÕÍ´`ꪇ'‘2‚jÝ¥G=uH]‹§¥¿!eRR{_ÁEñÑìÝ‹]Rã4Öñ«s‚ÇÁÍ6†ãDòÕã5:¡Ê‰6fªþS~°Sɵ7Œ›åËt¶Ou†jèrùLí¢H/0~Ä.yiŒ´¬3ôSÛHˆaÉ¢Lž|Qð:—-®ô¤KË¥sFƬ¥˜fç¨ò°à~‹‹z:»¿,ì%F¾ÞŒ#x®2’îÆ6sä–·Ät£¥•'ÍLÙ¤ ä›Uïô¸ò{Ѭ™Ç7±ó3JWlßæ„Š§[Bƒ;¢vÆT»¯G B>Ô†oçË$E—CfßíÆK˜¤ÝžaEô¦qà¬Éh >ÔØŠNÆ~ÒÃÇØ]\óœB\Šf@LØÊš©eSÔð]ÐA½˜z¾M$èÉúAX#µÈ |Dì[ –0&ß µž/á„\O‹«,¨RìÍoiW°­«ë7;×dÏÀ ×lš5H”™Ó˜!Ô«_}¼ Ì©#,=Ýby8+®;vŒÝTÃÅç£ ð‡"ûUhö+àçÜû®;CÕ*ãïM·›!å»HŒ«\_TÉà½, rÞÇH|êi(þÒÿ²ÿÞ7ÍŒ_e–©› vÊ”oR*£Àh2—¨lˆ³mþ¸£eYdزÓJñÅk~Ë;{î‘3®Ô$WE«þ±ùw©îü^ä‚î2ªr„K˵[’K·/åZûVZ’VÓ&–!ª„ùºeBq[¸ŸÏ[mÔ@\6§’Àbg—* ¡ ßR.ðË ©ÏŠ€Sšì_&sYâ?q~ŽôY>g\ÿÕ-³bÒÕà|¿ »óí{êkÑf]k}œ§wÞå4\þ³sŸ6vTœù(ûªâÍb³mó_“ά mUÞg‘´“,ï0´=‹ ö-s×yc(?ŸOxqvýbºK4¯Õ*hß5ÀiP$齚Ÿƒe1¹ÖŒÂ¥x–Z±*v à×C‹|^ƳŒôê[|ÚÈI»®ðçÊÍÝrCžÉcf¸Ç)­hN<~ÂüÙ5¿CCuÅ4Ï›ž~ãN)Y84TZÛ»,=;0aðŽ¹QcÒùZë—#ÉúÈi`L©u;!ÅE}6Æ¿àpõžÇ,ët‚jŽU/0»î*ˆ¥¿‰þýñ lI¬ÁF~F¦GaסDÏù^'ß8utªßØÍ ‘T ö™:í{þ€d¿PÍkAâß"ü®üœ¦•·’Ñe›ÔR¤_†­IÞ}Á`G̪ö]×€÷Û( ™¸]y—!i)Ò²ôÙòÛ×nSßË¢î nßÛZ¡ð´°Š7Æî ™§o¥$J2SlŒsD¯¯ŸóD ŸPîËHtŠæW }QVõŒ²èèõ®š±ÎSÚƒ¦™¢HøÓ ¢*ûó›>Eû¼7¥ê'$Ž†¹ÊnæºO :s1õ SRGRMr¬²­ ªºUñ"vãµ?óô/Ž7ÒßÃù|y ïáæ ÷ƒýKžWŸˆÜ)4ŽfÖ+iÙH~ ^¿ä—Ùß‘[I?Z¦ác„) yάBà·C‰]êýbô©Ùt¦¥³ÈãUg„ : ¥FBRo×Uõ FJNf« ã-uÜdžça;ÝâOºœdȷ•י5âò±*$¼—z÷Ö­)²zÒæhí µ?6=k’H¦Ñèî Ejado£ zð Îì½!B®,‘ÛšÔDpïì› €Èx…ÕЬâ¥oD11kÆ(”3¤•í¬ˆ²ÇçB•ò9S‰“vúX²©òšþûDÇŸ_’Gµ‘ ¥Èꯜ܆4ÉÕÀ{íÁEzëÀi W»êÑõžá…hÍôÕ] Y´~¬¢7(g}Û{ßÿÞM IëQ!~J«Þj@™'ÜÖÜñ¶ŽÓ$T×ËÀü»¯/Íæk-âƆ•:t4™Ä’¸‰G:Ž¯víOØVÑ—¿d§„ÇR¬’Ȅ׎ǔ &LW©s®ÄQáû²rä>I®Ò¶3û„™Ÿ…o©A&Ñ/@¦#N¤¥l$¬ý¢}M)Í¿{÷‘1t:ÔŠ2·Ëeh”9ßç¢õ“ ú«dÌ@á, £§åÁ¾µ4&ÜE`Cj#çÁ“½õ&L×¼ Ö—.j3E–LVûσäTŸ{¦`¬®·5 kæš$ŸŽ&²RkLX&ìÖßÉt½­¸ÏÀW>q+ÖNû",´öÃPýPclMpê±-œEKØ¥{[J¿ñÄ`’ì;ÿc Zóm­Å+‹™º/mâeüAy•’>i7Û©?Æû‚Ü?I¸º~óÖê–6æo¼ßé;°XÞõNº,G%[Ñ…+;„eCf rÎYŒXI° Œ>´væ6”»›Û½!ߺiû8 ‡@Kqùm|¬ª½£»ü¨±>?>szܤs°ú€FW}n,Ĺû´äÌ ~ÉBYªmÌœ¹I­PÛò8!})Êl‘;¤tK²Ñ Œ‹yö ¨a­jƒ|JŒzitþj@ÒÓÜû•qvO¿ \W76&ÿM<¡{ ¿=E;ÍÐqë«LÐcœÍÕB"ö2[ÿ^…¤Š%s+‘É­&’ÕYÉíÆKƒ£VI"ªÒP^ë  çú02™¼ì" ‚n÷’Ž(„iHÝ{@®0ê)&éñîm cÓUÄÅRT5*x£­Á÷WéLƒé³ñHŽßZ„hòù“éÅŠû†Å°¬d*õé)™x>^öobSv­½µ«Éo†ÁÕdö¥{¬²«¯(£Æ)ŸÈó“‹wØ`¦ëø¥ra¬Ûa¨ÖÚ™1Êê Èá]-VíñT²wc5±ÑઋDªª ÍžyíË— †g׌À곋¶)åõäð¥qÌo´&wv¨ÊW€%–óÑzOiq…lö×0jù´ñÕá<Ö_m6Ñp£t®ÝG®%Q¹˜ÔU™öÖz˜d¨Il NÜ/ú•­_d>Ï Fv…%jÞ¿45^’Ó0¥ÒJ±âÞ•oçYÎøÆ0~®eZ2h›Áò¡tlyÏÆg¹ºÔ$¯¢†â„™€§h5ŒÂ3ê¿·µ”ç}ù$fAíu^ñÇÌûÖá12âÉÒ0µÔðÌ“À­ê%»B†æ'–„^A\"›AíZ% –ˆº$K×äO_ ºüìÕsª,©:]ä•͆™¶5ô¹–»Ä*<¹,"E‘>’5iã¤"å€÷v¦nŽàž1‹0‹×­®ÓœÆ'°ïzf«YÇ‘¨L—Ž ¾:tþXwÖ'°Ú¯®TŽ«ÕèýοݎqòB¯ojpµ©eÇÙ•AbÑÑLÀò…ÜTUÌ)  {¤'÷‰ÏâGÄy‹n"Vë!q)6+} ü-(ùôð*]«á"·oïÇ‹ðГh¶õC‡|O"Uƒ$Þj†JÑ1ÏXÊvÐÃóÄnª†´Ð²ä{&¥ò=ÓÐ`âíÓS¢oÕÏ$ÊMè“Ž´(-g5\@¹)«¨—ÍÕÚÎÎŒ°ú¢g4\ÕÚ,”Oì O1òÞ)?¸ÿdMï&xÊ#Fצ’$,¨-PïNy|%{ïØ=^0;Î* ½¶“V¶#Û0$›ø*©M=_F÷Ó°´Lµ×¯K&-»Š4Wƒ3d>ª¬JèšnÞ—%Ø™y«\7Ù?óírÀB…¶Ykº‘±L‘› ªO§¸’-´:?Nµ"üÒ+å‘2À(é\¾$²î¼pGEgÏ ?0 x’^IÿÍÂÚ)Ø.åsnL©¸±Š@“ã³éŽê|)•üêsxr`ôFW9ù¸5x3rã € )öˆ}ñ ˆH™:ÌìH¦Šó°¼ý-Â(ª¶ø{#À²™~þXóNø¹Bp¾}Ú);~¯©ÙËAP—“’ÍJUáLW™´¹òy‰pÀûžB‰tL—ݧN7FP§¢Æ8du=¤‚Û60ŽÎáôø””°S4F†–¹CWG³9[aIM]`‡4àdåQl|É‚9ÁCjfÞ—o¹ö_{Óß2®fZ­Ë90ÂÀuƹ>97èvínÇn Ž¶{=ä¿Gy9OÙMS¡Ÿ–æõvòÉ—Û`ÙE[×õƒk¬åoLm—é£_)¸šnŒî4fÂ]I-ȱ„`sâ*wõ 5ã®®s8tõìˆU·â±+ŒÛ½‹l0Úª,ÓR"^D:ú$¯m‹¨¾íeüÁRrkº#ŽÑÈ‹å÷&}„*œßÂí3Bëd©±Î,°vbC'Þ¬å÷Õ…•®Ý½‹å·?9›öÑ–ŽZxX8êñs)=Ù¯{)2^1õû dKwÞä™ð‰nD R9ŒÓ£ gJùN¬Ë)I¹m(lÆÙ†í³{´ìFAIé°U²¤&í}”6”˜œ¢àÖ›7NÌñ2çÈyÕ›¦‘áHHël¼KÁŒê>~K¯Nw‹Cè“4WF£¯þ;kÌ5ÝÍÒËã´Ñ…ªCÍ‘±e¹,ÏÒÙ*ÈxV];+êìRGÒâ{Áoš6¥&Üt@ 2³,¼d‘•yòHgð­vñeÉh¶’Nƒ€°Ö·Vµ/ƒw¶¾79çyõϹÇz<ç¡%zîß&*uâÕyt-šDO¿0Ms‘â2€‹ ïfjY‰H_Ïû«¥¬ÂÇÉðÛº®;ó]¿ƒŠêds·ƒîo^{@·Þa ¨´e?•áàN)l¥³Ím4H½$|f_÷U#TGâõý3ñQÉ«Ž™£$r’²èòЇbü¥˜øÙPáÊ ¤@Éw‘‡¿0?Ví¾ †y;<Ê­IIÑN7gÅ]FTuôôÈy'ç…ßd8£ B6„¦ƒLÙ‡ÚMKEU†¢2si «)&æ=Ê C™»«åT!¬’¡Ñ@²Þ1˜h#þÓÎé[ t¸«T?Æöh•>æÀF,1CFŸ¥1bÿ¹ô¼=¿2ø{©åºý{p_dQ¢CUÚËq“±¿FübkxO¾ñàyöG[nËLM Æª2?#³É4^pr/ ¨¸Ò%ž¼ô›òÝ:Ÿ·h?x(2%EûÒüì •Q#§Á¥ö¾æùª¤ÊŠ_tˆÝì #Ëöó\ÝzÚ€Õ P*mLntÁb\rR¥|Î=Þ4Õa«êĘêøtGìø}!Ù~Ú”’âÅûŒ’†C c®áeÄ’/— æuA>ƒæä«ù|c'søt ëï¤ÄĤWkVkùÞ&H‡ –þ0=*òOà`ä#Âuƒ‹•*Í#“‹Ý?›$@¾®bË$·lZkÉg’dXÃ}½Èù™dc$º½Ÿ# ³PźXA"ß²€ÒÄçðôâ³¥!KlS; \GIÙ‰&¤+Óž¹êsØóÂÈÂ.Xb4#r·åùoTŠëðçLžWÛó&¼¬¤ˆÐ¯µ;*¡Ä½ÆaGRHáç›jÄ„‡Bñ_ÈkFÑÂJ+Äø¢só ƒ›au7õ£˜5ëøž§#,ª4LH=®ͲRÙ퇩]³®–ÎuJt¢ÊÇ X/­šô˜\צjT_`àESU]È¿nÜ2fëü!"ÕC(·¼×E¦Š‹?EO}Ó0äÔ?–¼Œ\HêöIÅœA„5xØ¢/i£¸=j.èñääUŽ¥õdƒÏdú–<æ:G’1‰¬Ð¯–ÅG’ééÈôe|SÝ~7­–½à‰–tó]½ õÍ’â©c³îþêÕm[ÖŠàüŽh:Áòçkfܪ%›ãŸœòöü rtPyÜ3¿“syTAÿY\~²§œø†»Y8}fÏÞ‰ O®)E‘´H«t¼ _lÝyõk¾^ûÂ+Ô<¢äºa×™«HÐ$G½Ù¹p¨îq‰öíz£ÄbZÔej“S£0€h6õ‚¯¸O\lC°û°³x4ûÂJ^=Vå¾ã¢èËZœdÔï»ÝÈãâGè3ÓÑžŽ«-ñ `“…^+eòëèœå%'%»4hŽµ½6ç]á Îä¹õؼî_44PË·†/¡ül’]ä:qÔ5BT·ªËXä7.à $ZOÌôÄÜlC¸^%–i‚îݾ䚨+‡Ùz$ºò‹É÷¢$û£?ŒªíUßlËùüN@ ÇùÂnIU^(÷;MöÅjãŧP’Ê©¸¢\ŒÂ»Ñ Ãþ†ÕQ„OÔÜø ×u*ïû—n<Lìðvìòy¯]¹&Ì´ê‹ÒNx&\S{„g˜}:(òÐñ  ¶ú¾Ú-“#÷›l¸góbaÅIn,„ä0AÚ—a:ßW>øð0b¼Ð"ÄÖßêáV/ „CצԽó-m ÷Å ”$æÌõu ¨áõµ·¥—ã`’Ë¢ü`Ñ\©Ùx“"¥˜ÿþvÔ/³K^´l]-…M„ÄÒ[y›ìE0¯øM5­3ѹÍÔ”_` "üQß~O\ŸƒØs8Á%H:믂³¨pJI÷FJ|fn\†ÔrÀ'ãy32Î8( ?N”÷Mªæ¹ëlK5ò®M$‡¯/‰÷”µi~“ŠÿÜ¢}}Ö¯¶å‹[u²œIÉ7!“О^ø}J '+iUªç˜ÿâ/^dyóÖËÛ¥|7zD—)ÑK³iTiËt1{¬ô§Ç.iªçñ~’–ííp0íóØÌ’Vmz_=tËd}7T§pÑÚ»j ›¹ÞeίËÅàJ[qÖ³ê~”È«Û.;äÁ×<›¬‚+‡btµ– cb1~Ý2hð#§¥õPï‰è„d_o}D„›á~‹²ƒ¬ƒÞü1±ºU5‚]³­£ŠiŽs.¾¯˜©=.úoÕ§i)F e‹ ýHÒ)°‡ÚU(S¹œjCŠâŠn/«O¨£½ÃöZc0]v÷ถý êj«rYr#wÎ'>ŸvÐⱶN‰³P™j¯¨èöO´0ù'õ:ŸÒr{ø¤¨³wÚ÷˜f¾]8õ£ö+šî$8ðY\¸W›,Kth=Ó/¤E.'Ï70·ÍĽdsêËVÙ7Øј¹IÀ­øÅXÒ6Íj¶Zg®Ubre–PeMG“2$‚7˜>Ž•rÖÑ;-üp Úc^›v¸EÌ@ ‹\±Ç‡í^UŽl3àÎöýfʇ¨úûæÒJ±™’µñq–‹ƒ¢ïy3cËü6Ë ”.y/‘_±GLËg‹g€œ²2¿A~È{[ã¶>"„RH0”s0%é°«~_²¸ZT6lE3Š§¶§¥²WEšÑ•Ü3&ï£ÀÙÁðäEøôêk=Zgn©²ØõU¬ìèù|ôf}£þ†5vNuö,;Ùfí‘ž ò΂¸ÅIKÏ”Èi>×úÂþ‹EÏGœTéë,½‘ýÀÄnû„°mxÈؤúÙýó0È]~ÍÚ•p´ëZ™#µð}Í=û±£¥.ÛÆ%]”9æ´ÇWiúz^µp¨µÖî½kMìÇpi"(™KÔ_|Q’ÊWªÊZŹT2’Qi2½SÅZ;:Ù%î»Áº¨ªƒPËéïæñZ*¨»tEkÒcŸK)ÚÃPÆ>|˺(–Ÿ‹IÑUâ^3"UËô:zGÊüLwé1k²f”ÇÈ\«¤¥±Å:úQšŽQ}ÄÞfÏ4³ú´pVÆ't-. m$óJI2 ËåæƦJàMžKíó®I‰9dY‡®Š*\¸b ½iŽél«Ÿ4ö®.n»ÿè@vñßü€– endstream endobj 1149 0 obj << /Type /FontDescriptor /FontName /JTQSEL+CMEX10 /Flags 4 /FontBBox [-24 -2960 1454 772] /Ascent 40 /CapHeight 0 /Descent -600 /ItalicAngle 0 /StemV 47 /XHeight 431 /CharSet (/bracketleftBig/bracketleftBigg/bracketleftbigg/bracketleftbt/bracketleftex/bracketlefttp/bracketrightBig/bracketrightBigg/bracketrightbigg/bracketrightbt/bracketrightex/bracketrighttp/floorleftbig/floorrightbig/parenleftBig/parenleftBigg/parenleftbig/parenleftbigg/parenleftbt/parenleftex/parenlefttp/parenrightBig/parenrightBigg/parenrightbig/parenrightbigg/parenrightbt/parenrightex/parenrighttp/radicalBig/radicalbig) /FontFile 1148 0 R >> endobj 1150 0 obj << /Length1 1731 /Length2 8063 /Length3 0 /Length 9200 /Filter /FlateDecode >> stream xÚ¸TÓï6N—tHã@º»;¥»¥Ù¨± ÒÝRJ·t Hw©€„4Š´(òN¿ýûÿÏyß³s¶ÏsÝõ\Ï}ÝÏvÆD¯£Ï%gµ*C!p.>n^q€‚¦¾)/€—W€›——‡‰ÉÀ þ…ã0=`NPˆø¿<<€6p¦hG8jB!5O0€OÀ',Î'"ÎË àçåûËê!P´ñr²hrÔ  ‡IêæëáäàGÔùëÀjÇàáüsz8ÙÙ@š6pG +¢¢  µsÂ}ÿ“‚UÒwçáñööæ¶q…qC=¤Ù8ÞNpG€ôðÚ~QhÙ¸ÿ¤ÆÃ0pt‚ýaЇ‚àÞ6@;Ù!0Dˆ'Äè@Tè«j´Ý€?œ5þpàüy8>n¾¿Óýý+‘äw°ÔÕÍâëq€œÀ@€¶²7ÜÎ °Øÿr´àˆx/'°-Âá÷ÖmÊrºÃ?ùÁì<œÜà0n˜øGž_iǬ±W€ºº!pίý):yíçîËógs] Poˆÿ_+Äô‹†½§!ÄÉݨªø§ÂùsÂB¼bÂÂB| ;ècçÈ󫀯ð·ñ7Œàèïu€4€N âÇfãÀ=<þÿ6üw…ÃÇ°w²ƒlNœ²#` è5¢ÿN>s^„üø¼¿^?Y f…€}ÿqÿÝbU- }9Ž?)ÿm”—‡úü¹øÅ\b¼>>>a€ˆˆ ð¿yþ>¿ØÿFulœþÜÝ¿2ªB@P€Ø$§÷¯?•ÁúçØ°þ[A ŠÐ3ÀúüŸò ñÚ!ÞøþŸ‡àwÈÿŸöeù¿Êÿw¤ì ÿ¶³þáðÿ±Û¸:}ÿô@èÙŽ˜ M(bB ÿëj üc 5öNž®ÿkU…Û fDâ€Ð9Ÿ 7¯à¸LÙÉh¯ã·süCK5Qìê@aN¿îD/ïÿØ£g炸[`ˆ–ý6“õߺJ;¨ý¯äØxxØøâ €X üù³jôù-q7 G„ ¨ίÆò!öÉcq°õ°±s"îHü—ß6¡Ûþhè_Fþ_677¨ÐÝѽ¿pîáõþ¯;bÿ<@0ð×8ÿƒ‰!0W7¸/ ø¯¢"ÐÝÓÉË „ ´ô·3¢ž« ýBxº:Afú”0‘ØÃ<ÛÓ@Õpú%Î)ñZ¾;žò›ï¶ybÎØÚÚÛüá¡,³õÁzáÚHJR?5ú‚ñÔJ²”|µ6³¼Ð•BeF2e©k$ó:ªµÈú÷ù’á®”Ï-…ÚÏ&Z=› BKRëg4V;yßxð(ÚSc9²å™ê푮Ʋ¿Ô € ÑÒ0šzßßkÎCË+®M5PÞµÆc> °Ü¡¸±NÈMáã·È‹Ž||!/¬¦s¿û!™ ¡b"ÑQEÕ¢wÞôÌÉ}ùbÈ©ÚäQäb¿ým÷Ø@Ô)”ñ3-Ù\H²IOîћÔ6U¡«"#v%ax£ú2´ÏŠ-© i>H¯°M§«WEÂLVle†!ÉH€ Ø€hÙ@c)[ ÛTOZn¯ÐNIlÇTÄGÂKÁÕ­IB kÃd1†æÃ6ÝíX Ö‹o¸|æׂ;!n–û-Z_ܾW¹Ë+ˆ XV"Ķk5ƒ™ÅüÑp£‚ñ.E‰ñlfZBù— ÜHŽ Vœ°Œú»‡%GæŒqí±DM·*n§Wƒ4k$4|ä1Ñ~ë+íküüìhÏÂØZ1T³#ÚTp£ü<²¨Ä4iq¢ÄDGéTãŽÛŠ¿ãAmG%ôI†?¹Ï%éè!K_·'e@ïæêÖXë­^®RúYÙINGÒ<ÊÅwæôílð#åû‡Oƒ4ˆPôùêè” l¿–ÕNrDñn,e¥ïX±Âi¼@‹Ð0Ð4E¾æÑÉ£¦Àk° ûz4ÃçZË ½š]‹*×ȧòpõ*íѹDzÕñ¯æ_}ߣ¸Ô#úiŸ7–Èé›øš=â)Ëú¸ Î>QN2ÿÀ°(ß?`ûòR7ÐFà“r5£÷\Ô­†óå<õ‰±\Å@í–Dý;£õ™Ÿx c¤aFЦ¬÷^Ç©2Mé$Òó¥Æ?Hs†o—W÷>e¨þÖš‘K>«w‚±RŠ–¯«èˆC‰?/kÅY0òp<Ÿ Ç…øa‚-¼'§z£c‰RՔ㽚KRf¯:úÄÏ…ÕîZ Ìç{2]ìhȱþ˜¡ŽjÈ_}ªrFûyÒ´¾»ø„çØ«?iÚ¼¸ãû ]»“v±ìQΓ°È4bÉu–DµãR9‡› Ë5e.šŽd6ÓÜØ—ÂE¼ð—É¥ÞëðÁ·UÒd_j¥ Yž dÄ|juF¨|uË¢Â:´Ù›ñjxEæwç˜ð#°ÃUp_Þ¦êù½ 9?øGÍX):~¸Ü)ñÏÉÙ’Ï?\ ¼}=¼Zã±(œû®o5XÔÍ×ÉZØщ¦5B¨DÜYÂ(w‘Õ£ºhK‘Ÿü¬–]2¤V|s®é@•âÖÉw]PrØñE°ÜQÍßÑc_߶M­ìÏ?…v…Z.3¶1Ç:?³ g$í´Á:˜ÃÃûy’·¶,Ùré >½ ¶Á;GÙ¬ù¼:Ó´ù0¤GŸ¡X"i?㇧%’~ ŽœÐYŽ°¼Ñû/Èûƒm¬Úœ¤Æén†ˆÙëJ ŒI® @ÿèÙáQÙÂùLŽ8š\'|ä»J;›)7Èèûíæù‹þUxZ•¡H_™`ߪN HË›Z/†î|ª1z·35ÖgS™kq±/³È‰‹rrf /§e; §ÜkG{ X•cèL4‰%ë,QuNèuÝ®±û“©¯;<„¯»û%J›Ì2÷û†w¼VèÄ\p}‰¶cX´ðË‘'¿ŽOçØk`Ýé0Züº«RQ UQ“ú­{ZH}Waðg©4Ñ”±öQ f4F~¡ôãùÐÔºê7§1hУˆL7,ÈEðm“WéÄ­5–ÊLÿ¶ý\nd3ª+ºéí&åJ»ÜNzGÜÛ8õä®ØÉ Ä6¥YB¡ÃÞ\V&•Êz‘¼]ÞEQÐÝ<ácÁªHrbH˜üÊÈK»ò3¤¹zÐHÆ©˜8Æï|Otx( ¯TÚþÞÙ× 7!d°ÅÈud7k“Ò½|ÂAaŒ2å<åËÊëÎ\d1?ô®OÑ ëA¹®A B™L Ke¶Iu–à õ'ëâÂf›y'—ê0Ø Î~b¢>€òÞ¨•ÅoÕì|BŽö6äóÕôƒ¦Eƒît¢Áœä÷SøjG_Î,*îÌËNÅa’Ï?´óÐ%]7Œº ÈCÊ|ŒñMÈ(ü4k{è®xýc.@ê>>ñ{f@H;@gc–+öµœ.1pçS;-‡ƒ›\0gg¿T€s5¿‘Âh¨HÞå"nê—Ìú–¬ îå jêÑŽqÞ’L¸^þfg, 77בŠ²_'¸¸³äêvöY- Ú,Ûи•rouz(7$P¾¥KvÐ'OgtñŒñTL©¤óa²³Nœ…  zÿ¹¿7z½Àu©òg‡ù¾&Áçò9û;yÇrÕÄ›=OÇ m¼^ŸlÁŒ¥‹z™<ó™¯ND±€]sι¤?T(|±ÝNSäñ²Ü‚Ïcžlà%)zouÜX¼ Æí¡N&;²/òWJ%+ŒóY¥ÜÌ!âf$3ÚÜs¤‚ç_ù¹0º¸0ZÕ“¡ÚÚ‹oë »ÐŒ3eÏÆÏ¿ <¿Êr{Ö­/JÓÝ͉A³Ý¬ ãq;Gn9/¢1HsïUžE³ú¹(G”Ú9"Ñ]À¼/qUOôgö“\ì+Àq=Û6w™˜~Ø|”ÿâ+ ûÃ;w“`dÈ\’z1ÞåŒK¦’ß•’­`|tÑQ’±¹Urãû˜u»Þšé5a9}·.]>/3lçã´'_ÍQ³è÷Eiã‹9|öƒ¶:M‚î Æ[ÆWR¤TC—óeíeìØ-ª¨2ñi‘˜Ø5ŸÐ Rå”H(Ø¿ØÌFõpë¡là#u¾+T&¢"싶‹€ÑÂ3NTX®·žfvº±¡Õ²]aüÌ·z”KTÂÖ¸ÈÚq›WÚò@S³§¥)Ï}Ø‹¹µ~ìZaã“€R«†¬ÉÛ@ÇÓ?å!¬»sºáF§ Žƒëyme±a\V£ò¥áC—úTd$Gu÷œi/P„D†·š5ã ËÉžš:W£YÂÁS¬Ÿ˜ÕT•+9l]wùYÂz‘loWnë‰xs]ѦéZ[¾´–­®'#þíÞa#z–¦³¹”øœ8EŒÊ ™OWÏŒ¨dÜÎ0}g³:HuÖ™”¥ºBF†å£0®V¯ŠUÔWt›ži5›·£©ŽYF+`<ô[ŸaïaKÎg¦ÇÿA¶D|++ñ-©±tßm®1Zêac„b¾ zZL´¥Åï*±9‡zª‰<"óÚ<ì»Ro쟯9¿. ³œ”P©yfñMܘR‚”ÈëQra6ºÉ–z¸ü£Tì˜r’$nMU ™Bzƒ—Ñîm°‡Ç¨W†;ua©R³&±?Ü¥ 5¿Ž öÖ„KûJÕG²o‡Ø* ö>³ÄëQËy4ƒYZo†Ê2~œ#?½¸^)¥ÑA9 +ÖIcà P®“ Õ’t~¼‹õ“%tïM7_-É þk‰ÀG§O5ƒžÛ-±Š¯“e®""·IÌréÔ£¢Ð*‡ö_TîÈÆçÔ¤Yl®˜†Õ¬uÕ?–¼Ý>óTŒÌO‹ÁöyüÔЪÎçÄ yKÝ$íÊ4o}ìòšÂÙÞØ{Ëh#”Û×óЙñž1„ï(¥IOF´JX£ áz}gºyô®ÞÀüÄn ÃâK(%–A%ËÊ9a£#Ü7öîå³(îi¯›òÐÓÌ׊ô•=CZÕ [Ž½~íH‹p½ÏGaï³ma“W „}ŠêtÚ˜”ýQ åot–r^5Ø#[Úy^sõ4­üÈ^¸8~·Ÿ Ú×oÇyW×Hm_å!›çùùN§.2¢ùjÞ», eÙëÉ[³EÈú'K¡ûK÷FïHߦÔïùÊîÝwnÞbýþKað‰Äˆ橤¯¼­H%˜žÙ'©ó]¶2jáK—}1ÚûÏçH“(» ÙA×,žŸÇÈè¥ß |çmN ì~a¹X—NÁ¶˜ÃåߦÔÑ ½tËO¨W1ÐLàóvÿëQ´îÖ…·¼ÃLÊV÷þ]WAÃr8ÏöôÊO¢£1[Õwñ©Ò2zÿøaŒ›G…‹&n›,†ìùu‘@F6xp,®½û†‰6ä„mŸÇMoäf}õu‹'N¶0Uè†ãÍj¬Ü쌅 ‹´±£ü2ûãžm´î+ª‹Ó}0q¡óÐ ™’ÊÓ‹‰p=xë7‰ºßS-ÒâØð½Ûו19·FüsDRšäP6R®5»£/Åç¨ WùgÛ¥Gù¢—ߦ1ôéw…iß]Φ²kû}ïä¶k?w¶K¯6eþ±ñaeeš$éøPêH„8Ý”"§ùêæíé!§œdLuC96ƒ'_æÇDlÎ&ž… Œ»Y‹}ï•Ag`V,ù~²UFÊ€nŽê#þ뀮ƹȣêÌË~MËþf‡Õ‘Sqýyé€UÅÀ¾ΤïW’ÍlkŽƒUc(š½DpuM+sÒÛj¹çªƒÖ.D7_oŸ<ÊÏaCÝB}oÊæ¹fÁ$¨ ¾ÏfßâÞè£Sš};H]åæÞ…ežHM_ÿ¸f8ˆÛÁ-QNY VPYNB{pM•›'œs•vŠ/6òl˜x¦þM¹pý€Å?uìÖ\½@Këf9–¢€$Ì\’°ù(´Ôñr:a²‡w¼´èÐ1tàÙy'ßC,¥\ôª`Ú‰œ3ï-".þ¯Þ5=5zÿŽ[C>Âd[Â#Mºô¦PìS+ C*¨s“<|‚À I#‚!¿.ËeEßm"$tÉQšÍŒêP.Pòf­sÓ=}}jê’ ùñ÷'l~3nߪsÛô0¸Ff:3cû’f/Û”>ØßòY;‹¦¦á¬ n8¢¥xÅ~#c€u ä5·ÍŸ#½Å½ †°eZŹ»\¨Y7AZ²tiX.òvþtc‰fµ§CbI{±u·Æc¾:æû™–Æ€O&ԯРŽ<`Fc¬Ç1ãaðéJH³>ž[½«ÇÖ·/¼•»#‘ÔõØÆ­Y;<'Òï Ãya{p,óî0,¦s(È¥½âi{È’j4ç‹Œê/œoLi|}‚àËÔúÚãÄìåDdwcp,ͪó w8¦[¬UKŽ ÒÀûìêâ®ÏÅÓv}}߬Lºmg.Ûä9ÍßkHéĆŠ®ºÀØîwÏWŸA D‘^<§ƒ©©}\›`Òiñn ¤Ÿ)'VÂç0 î}Ñ Æ-uF4LêÐÍ~ +;€¸Ø÷Á¶OÊ@Õ?fšvü£PÔ[[M‘-i #pbqãÞH‡ãÍ6¾Â–ó’æœsºˆ ¶>ûl…ЩõuôrY‹§Çp«ÑjÏÌêÈ|(ÁG*CÁ9ÿŠJê;¿4§`Û¤§ª¢åÒVÖ|èÔ²SÖÌe¶×±º6’{ñKt°öh‘ D¯îŽ{QHq*Àà”u¯oR©oÃ$ᛄÓÐë~6ŽO]¢ùk¶èŽUfcϦ‘ª(̘¸`HìÄ°RJ.¾Úg4ŠN_¤È¬?õ¸2¦ÛÑ+ô«©Z žÈO"qÖʃ{5]9'…wç˜ÍkÁ€É5mÍAæ¬åê ìuA‡¸ÕD»%mÔ×ïw3ÙŽ[ð>¡ö¹K²p† M·- ZW,J”úat O}­ëg_¶ø^\­@ÍBVâò>Fní>Œó¿}¼¤œ[04Ê8i¡H¹æcŒŒ)óR÷H‘šÕ1Šªha¸—åFkæI”WêÒ!O.ãË6lÙöš#]ØþÑSiSÝÛcÚ‹t5È”Ã6ñò³Ëô¡}¶%ÚáV±ãOÆðÓfgéùÉ=rWb¬üÚq«Y¼¡”dÏú ΂«ŽññbÚÓÓhFWª§}ñ¹†g²úd$\&±±BK ‡µX4ûI(WY÷wÍC( hå„šõ(€ŸŒÊÚåpêFºÓg~¯ýóŸ)úâ€ãÈ“ Jo\T [³ø·nE ww.^zòÒ«+*d¼%‚¬[¾áë qgÞ©ƒ?ìô\Úƒ¿Ç¤ëàI4•")0›¸^Ë1$ÿéúVÂ=¼ægM®ãÍ0îÀ*«J2[ìœÞ-=ãâ7¦T_ÇxÉ´ z‚A^û}Vñ—¥äŒIK%fKëªg¹3x,®Iu;9Žª£Í½(©D•)/|éªb1Ã5Ò~æ‡?ÚÁz`°HFÆúíC¿´b¶ ’ÿS§á+#a·“R—‹qæúÿb–/v†.a7I]yÄÏø˜w²D+ê0†^S®Ng¶ãÌ #5'«¬ ÅÉ´=íÃÄ1w¢¤Q°›ïš`oÕeÜï}Öx"61¥ 3Ÿ]o™rU‹(ø0Næ΋®ÕòÝžv|¸â{RméYld[7†{d7ön”Oí+Šozn£ÃäØÚfÈïH²‘—ŽL)“e3/:k¢,HØ77–6Ò‘´ú¬&ƒ8¤Ë±f«½8m‘½F6CÖQNtYÆtDO{k -vÔ k™ø=¤5¬¿-´us¬ñŠåÇä ëwp€cÞæßn<ŸxÛe,î3Ð.nú”çû»ÁÏ:™sŒƒ0ü4JEé´ÐT4æbvZä¡ ×©Ä%—oß¼ÜÇf0ÚÏ!qž©¶`µ(É;ìD"ÔÂÒ—·òΣ†=1Iýù r9Ú ?£›vb‘:;gª—NæPXÅñV£w:¤¡Ð_@‰‹Lq¼†Äб|ðxœ~Õ0ãñý ÷Xý¬Be_¹›+ÂAnë å§j¦#ûE%µBhtóƒê|Õò¯ ±ñözíž6#'Ž"Š¹ ­JJRVÏc‡c8€T´#X”HÀ½º[´×i¸ò§Õ‘å˜_»?nz©»ÈÃÐÔpX^a~Q%¯)²¿˜™<<*HgRrÆ~S“)çîùôð[@GkÑ©3$]ÍÈòÞe c )'û°ÊTçä-âÃW ¸ÂÒjÆ~~öÄ9IrvYf…þ HôÙ «T¿NDZìšÌ¥·@Žx>ÁkÂ=†©ŸNOøÞ“‚08iŠîªMtðd$wÝ£23÷>ÒË3g~?†Ò&3÷°IÙVu´b ƒÊ§R¥îŸÅcáÞ’VšŽ~µ€œº¥ÉÔ¶/¶k ª€n ˜KV€x#i†A}…4 í·M“¯u{eËsâ$|Í؇TG«î ë¶4v¼ÚðmªJžî‡m÷¡Ÿ“¥Í½V\ë A¿7¼Êš5y¤/:jä§HT†Ûä9ˆO¾—#‘ ’Ö³zËngÒ³‘@N¡†ÕUsNßœ†Lו7ÌåÊBíÌ1ÓÙ7JIJESW—£ï³_ é(u2¦¿¦ËòÞýìŠ ®ž¹ÐxtÈ´„>¤þûÖÿÈè1ˆs9ä^YVt–/ Fß•˜åð…Þ ÷>~<¬§¦Zsª\àõàš>Òÿ®N£tòïëÁ=!‰ÛÞÞ´¶óÏs††Cw¢q…aW6VâOçDBÖä˜Bs}Fm6ò ¥²&usäî¼U³%@ßr%sôÝ¿zžõ-Ðæñøz—‰•çQð1ºVX¤¼êÍ)§JÓw…]í.•ÛRúNGìõgfß9Ôrìqmæ]b&¶ï‹Øí4›áÿbù@ endstream endobj 1151 0 obj << /Type /FontDescriptor /FontName /IGNLSA+CMSY10 /Flags 4 /FontBBox [-29 -960 1116 775] /Ascent 750 /CapHeight 683 /Descent -194 /ItalicAngle -14 /StemV 40 /XHeight 431 /CharSet (/angbracketleft/angbracketright/approxequal/arrowright/element/emptyset/equivalence/mapsto/minusplus/multiply/periodcentered/plusminus/proportional/radical/reflexsubset) /FontFile 1150 0 R >> endobj 1152 0 obj << /Length1 1549 /Length2 6833 /Length3 0 /Length 7881 /Filter /FlateDecode >> stream xÚtTT_û.- (¤ƒÒÃÐ’Ò݈”à0 0Ä Ì Ý! %  Št‡„¤t#!J§twüGýÅÿûî]ëÞuÖ:çìç­ýì÷y7‹Ž>¯¬Âª„€£yA|üâyM}cQ?¿ ?¿1› íý &f3„"Q0\ü9È#¡`4S£1~š8@ÍÅ€DÄA¢âüü~þ‡;"â°+Ì  ÉPCÀ¡(b6y„“fc‹Æ”ùûÀ á€>}ð; ëE `8@Œ¶…:b*BÀ}E{üG NI[4ÚItssã;¢øHi®7Ú EA‘®P+À/Â-°#ô3>b6€- õ×GX£ÝÀH(8À P8 á·‚"˜â}U €¶þÇYãÃÀ_gñþI÷Wô¯D0øï`0‚ptÃ=`p€5Ì ÐVÒàC»£Àp«_Ž`vÃÀ–‡ß;”du` Á¿è¡ H˜Ň‚9ü¢ü•sÊŠp+y„£#ŽFÿÚŸ …`ŽÝø§³öp„Ü믅5 neý‹„•‹ð1æìUUøËÿ‹Ù@Ñaþ‡""Bb¨3ê±þJoàáýmý‚1 |¼œNk ¨ÌŠù{¡À®Péõñú߆ÿ\ƒ@+ °„ÚÀàÄÿfÇÀPë?kLó‘0w€)?F{ ÿ¯çŸ¿§yY!àÿºÿî/På‰Ác-ež?Œÿ±ÉÉ!Ü^¼ aïCa$€y‰Š |þ3Í?ð7ùߨö×æøÿͨ ·Fþá€9¼¿y¸þ% οF† ðŸ´-CœÿJߌ_˜‚yþ¿àwÈÿM÷¿²ü¿¤ÿßRrqpømæümÿ?Ì`G˜ƒÇ_)» 1c¡‰À ü¿]Ÿ@ÿŒ²&Ô æâøßVU43²pŒÄyAB|üBpJ æµÒ¡!¶„ôw+05`p¨ûuã`¢øùÿˆ™:ˆ=æVAaöÛÅ ÕÖU„CV¿¦O@XF"ÁÄü‘  ¼@˜1µ‚ºÿÖ7ÈG 1! G€5Iü«­‚ …á CÙcº`ûËøñ‹€–`¤•¥Ã?f@¨ô×4þƒ‰€`ô¯¦£1“ò«‡Úü¾JQ`Ô¿‰E@'”# î‚úÄ(è„„ašðO}&1lõ[¿Àÿ` qA"1»ø-NÌ‘ü½þ}A¡îPñø"lW\Z*ËàÆû£Ÿ`v¾ñEŒq{˜0šýÛ[/[ÉÊÃÎrϬŠîô¾Ñùž24æÍÀ½}:ànö95`&A%¥Ø¾(·Ëû®ûìà³ ”Ù Ao3(úæÅ"ùN¬.Óíg¦x¦¯žŒý|.Ú"¸R¼üá²µbòpàÈ>&cI`@Gɦ7nàD³°Î£{ŒÍÈ¢)¦‰.ª/q"`Ñ'rõ‘tg‹¦ÑI=ÓÓ]©ío÷ß«·–ˆ…PÒQyª±¤÷ÆuÒQǦªU£°Øå„åóâcîÑe9†°Ïà:–‚Y£Ûèø9Œv½²°0ì8¤ömµÁʺú¡>ö.I–ŸžoÙ¤úoÕ# âµÁ8à.QX'dÍ#§'¶q¹qï…råÙoñ®þ`cÖ´pýÎÝ;½[-œò3Fλ*·´´ÏÃçV>'Å8ÕIÁžÞÕ-V2™ÐŸÔ('ñÖ™†F~øj{°l&O÷’‚âŽü=‡=£åoÉÊ{+Øcrîò×77ž¤à—p`)®˜ÎkZœ«hî‡õ¶Šäñ@(ç-'œÒgM,«¿7ù¥áïôš#¦„–*×zÞ’D*—êÏñO™]™§¾ì+1п×HõEo„H0@>I‡‰ò-oÉY6êEâ”yÍüÈ÷WÄ.¿©OÕÆWùÏ¢guífŸ!LyH’þ~ÕZ:Ðt$Iè{]Žgƒ§óÜóU‚ø¬ªÞÀ‰LÏõ÷ù+"lËDÞ -•—´¯hzv(À92äòÜYdÓ~X?Çra8•cñ®8ÖK]÷S®er872=((ck*fG΀¬N›ÏÓPíSÈÁ°»úògöê‘ŽþR9“^æŽh‚ð9µúEì£Û\*ݧGÎàx†D™f܉oDw£c²íjpÊß±ˆ½H"o©æàôDJó£}éCáȉ_¦WÕK—ö ñ v‹{\¥ÆêÒúÐ܉Âê6fò½Î…y‚ÊÆ«÷¶f‹ž' zÄv°Ý¶|Vë6úÅâ/àG,‹ÞÎü±äìx_[\ÍŽ`:.-´ rdòeÀ¯jâUH¥^Ç5“w×Д“n,Ú9#l MК,«°p·Y„˜çj|õž± •ƒl6€ÞͶq8žEß lÿJTÌô„ø]ào©®Ó”7iÝÃ,«=ûLÖOÀs¤×äó”ÈÛ©`w­©²Îô‰¾ˆ«éŠyÈÞs¾ý{8gW ûÀ$v?’à.)Ã7 ñÙ ýù‹¢ÕŠ~’P/oÓO hë9 kÁ_{Hwa&û£Y~XÁ«ÌÄ a¡¶ ¡UJ@”WéØø'˜ Ë—)éðÝ©Ca²ÞMuHvén’¶giçÚãOŽø1µ_¹o‰:«Ûò£$nî⥓sÆòR…ˆ Šc‹HJƒÛ’ZŒï¦Ÿ]6ˆ—ÓâUl(À”Ðâ÷ñ{BÚÖ¦±7gÐË ‰w»‘dÖ†?‰Þ¢ïÀÏ;»¡(,ª1g·l9üÖÜ{&c=´Æöš¥4²t:ḵ«ðæn,Ìè­*á =ˆŽ/!7.®Qœ‚ºá~K7Ü«¬Eásʃwû¨ú‘Á Ф!Y¹Øx¸cÚl3ߌ©äÉ¢è%éD]”·ÆŒ¨¯áTFF4 ìmiwTd‰ÿLO*íñw…Û/º·2ŽWðÑ' ×Kù“„Ï×Úšú6YÆŒŠ(rH®1ǯ·8²smG«5’õë»Ñ±TnnÎ÷b,˜K¯Àë¬xòÆ°v]-á‡óƒ èÇE&­ÁÓ¦)j_tÕéˆ"Ó¹5Þˆ•,<¯ɤ{ÄUšÐE}q51isû«T©ì™ÁüûxÊ— ‰iÃ)uߢ:Bw‹î‡ßícNë^s·ldðˆ¨ ;ŽðJ¡éXŸ¸^ÁbÊžÛ³¦uQ”ÂnàKÝà=§£¸…ºÆN(}9Y³ÏH¤H±…šÝ¡¿áã“î¬Hs»Ÿ!¾ÐÃòCøÉ ìÜÕ¨~Ü!,áÒ¡½ªcìò§Iå„vª¿,é»LìgJ0Bóú²® Vkx|˜°Éþ:b”y†Û0B-4=6ˆT^6˜ØY<ð}²×3JB-h%ÙJVª £E›Y„”f]û5ó2Âcû²£·oed'û €93nì LÔk K+jœZÑöº$k‰>Ç÷eë|é-t%E¨ü:k4ÃÅŠ½ªúõÍïM­Iœ½o›"i¬øE¿,Å#œì¥A6öÍŒ¡’„çñ‚ÖbcçkVA?5õRx×ðç)Áq˜¨S±8Œ•Ý†6\¶[ ,¸£®-—˜+Œð‹ÝŠ¤d¥/Í$ä2ÇÛûÈ*cH¹uP+â'ëDŸª <ìcVÞ=ËÁëxnÄT_þËë³¥‹Õ"åú†µÿ0Í|ûs.ãi‘/,F¤KO½ {oà[z€Ì¤nǬ´8Å~¡/xt8vgY"¥X×Éð®ð›|Ù“ä¦Ö/·öÈ òàa•ÜÎ$Buìq>’„ û—¬8 þ”ÊøØÌÔ‡æ&£ –*sŽhÁ†¹—G2ΧMfÑÎ;›Pb!UÛó.cs¥ùkÛöŽ…戤¬Ø~Áïu¬"‰báJ¾62ÔbÇç^‡G.¶ßÝÙ†…b¼ý›n0ËM/‡ÌËÂ>ÛÕŒ°éŠ™¶Ÿ‡_Wõ‹’T…Ûlx G:ù¾´G+SMº†Qî híÓe­Z߉3¢ûøQóhvö‚ VQûv¡?ytnÅÍЙHIüW*Ø–ÂL_wJ>Pn’·Èj^Ýoý‚SòE[߶e>1³<®àQçÌ«d4É”øQ1Qå¿x:˜¯ùÙ¿þ­‘Å%Y¹êPy–¶Sr–pó;p.iˆÜ[Ç÷‡¹™O×MXf¨¿OH\/ýÞ¦“㕨GÑé8Þ°ÆùBÃ6™ÝîEÀÅD„ª|³à\«òºè7âDÝkס;Ý^Û'›³^“«¼¡ö<*Úd;ø¥³—Oú³¯Ã÷µûÄ•hv<"¥‹ 9ñ§gvÒ à3qÚ7âËœ‡›W#ô†·fúL¾eŒò~Ó$åû«¸[löDšUAs^ÑUNAß±î0~õ[~u«Žd”9cª?I×ö®)—X hLhî ™óPšÜ\íæ“îj¿1ìÛfÙn8غ½¢7ЬZ¨n1§fJKê1ï<Øw„⨼mNÄYÅ?ÿþÈáñ½A“ï]ö­ÁÒ¼4R 7ÞªvSª-óêÉrë<ªfqiz±[µÅÕ3Ù:æŠ÷ú*óãå%«ç…ÅóĦ˜b‰'?³…©W¡’§Ô½Š$DíâR.4Û D˪³Güñe?×äïv¤.yNt6­Œ‰îûEN´ò».êÑ›>«­dŽüy2ÑÎÅS'î}þ2΀&,g@E{ â‚r|]ˆ½õÓß½0î"„=[÷D*µzÙ)+š?Q‰È àü`ã£9<>JcÈ[^Â}zfåL5PfºÞj>Hö-Œl© 8îAKÛK:ÉaW+•4Î…Í'žõW1*çmðtµì¨º‡ž_|÷¢1’S{g–7Z6ÄHMÈ`Å8%ÿ;c“±×±32Ä®¡dH~â>’½‰~áVÎóOˆ ÕB†zÍÔC ¦›o 1öÌLÉ3Fýóœ9 9<›>uFkhEFIP דŠ¥ÄùȘNØ5wà8èѦàßàA¹¢áQ?ùNng3saÑxZ¤<¿sñY:·¡Bü§Gœ9lªú‘~ZéÚÝI÷(þALS+woå ò6™j¿—ÁæIÅ"¥zœ$óÙ^ÃûëõÇmEÚH¦Õ¼á¼§;ïÑÄ鶗OpõdDut)‚ê]&}Ôà Ò< üÐ0Î<ê‰õüNPŨ9§Ìç…Ztj0J·Ÿ%ÑÃÀúËG³KÖøi„ûßÈr‹‰7É}—c82¨¼çªãµ„ïQîø’tè¨ã>î«„PsŸ½[ŒþA$ñ^[vHµ³éí@Ò¾T 6Cœ½/îôö™g=ƒîº¢H˜iXÑ{RŠÃiï¦O$ Á“£|f?y{°Ê”û÷Hì›Y¸q…]ÉVÒ ©>7Qj}ø ÓnR®nÿÓÜ ÷‚œž—Û®f9?§ž¤"R‰¿%ÛË?Ñl)úRùüó.È¿Ø“Ž¯¥å<âÙ‘üyÆU‹¦½µîC:²ZK véø>‘;îAæØõ–¥àèîï¸CIü›†Ú*âÕæ‘vjï>‡“4‰Ú¹Å&„nP)HåÜ%Å>iQíÍ÷nn#«îþ,×æ}¶úé¢;ë5Ù+ÒÖÃì’7Ðuµ©•ZÚ´ :Ѩ@‡“wšÃY¢²BŸ ^é¤:•¦¦kä=Ök.Mh»áo½vc’§.>¶1ô#¾©“»Îm  °×ãV¿Û7ëûx9å!\ó$>ðÄâtV £ÅN?ìFãW¸ébwƒøŸ~á ‡árÑBCõ"9)µÆïS\L½oñãÙ™5NIÅ]¤…Iš–ãc%m>"uÛçjk ØE¤6M‘ûM4qrm©pÙH­~öš=ú˜QûTFÑ×À¯ªóèjoÜ(Œ‚£2ü™ÁyÓÂ|Ï›ÜÌ{­D]²èÀbþú¬ú+ó¼‹?Ô·²o¼€±,5Ÿgêz¼1¡¢"_ÃGQƹëÍ7‡û\îœsR2¡c>Ñ­gË$„Á,V§Àt%ÞáÁE7¿ÚVÑ>Ÿôš¨->R»uPÎû½bóÙ+÷2C…Ÿ…ì8IÏœvn2ònµ•-ÐÛiój žác–))7.¦5Ä7m^åćsû©O^´e†šuÚ/ñ•n‘Î39|0ªÐ‡,z#¯^š>%ëOܾÐe-Œ`™¿ï2*F¥5ì™"“Õõj-„ªÆÛõ†oze¾b¹û…Mbš†6Èg–ø¯\Op_ëLÒõâë‹ØÄe³_«ï× ZIMÑþã!]¯[¶FåQKX(ñ>ŸtŸ†6Á>.8ÁKƒ|1¡:JÜÂ%_Áv²ÜœœÛÇ©#·Uµ,<%ÄΑìä­ÕÚd6)ý퟇ÖtB+‰®I2˜å S??œ³‹£¹uþÊ&JŸ·%Ì,æO•‚”[j‹7™”ô¨¬¬[ §Ž¿²íøm«ï¯p?8ÚVTèØ-7z-=£ºÙ¤ èfAiK{X,u„ªUSeó»¹`»-ö÷/jß|Šû¬ˆqg—³Ï^°Aš<êP‰ŸÈ¼y:$õ"P¯x¸-ø¨Jö‹wŽ2ÏŒ\ÄZšÞòî· RAí±=ÙE¹‡–!¯e”Ã=Ê=Љ(`qÀ¾g˜Þ"º4ÄKq£¢U¨¤›âÿ•;…øõvÂn7¬±Mj­'hJIl“»üÕ›”fƒ#NH³iNÿ»§8¢Õ*AÀµ‘.‚Fèë"`|5$ö•â"ÇL­(ýZU NlºÂ€ÔúûØA¨ÆÍùW÷‡l“‰Ýäo·X%!¿{ŸVz¥auü~§‹˜hÜjÜH“ÑÉYq„†} NŸäãí,üñÔ¡¼£©{¿HìݵP25Õ•ä€Î¯rÎPÛ#O¦½úÞÂÝ#s$ýÑB "Üd¿²²r ~ϹöyÕf:>ŸÙRh9ÂÚÍ•C!^˜¥Û 6™Æ¬Gñcç KÁJÞh8…¸R©}]ų?êÃöaæâP¿#ë°265]òÂÉ ÆW/Y:M&±_œ‘ÒóÓ·{NÉZÐïˆÜ°nwÓ?!—½ÿÎjt¥ÌlÉ\úD¼µ>Vÿp”°P‡síjok-r›‚5í³'ÉL¬‚Ë ð¨6–¶æKØí$‹ÛËÓë\x^Hõé®]‘O>5YŒNñž"_¦vEÎè©Q ÝÛÂ^1ue*‹mþ –€1Ä¡gö“ ŸÜBOÝ[˜6ÛòW°?WmŒ :î5ˆ¼? WíÞÕö'®¸¸æK;hQë[<2‚ÞJÆ‹ôâÞ߀w¸S{nçº6'Ж ,¿êò˜,@ëçö¯ÝŸm7Ù ½\ÁOïvuV·XHk>“ä+š´¯ñ \öÏ”ðçÝhÃÍγžˆC#3—í³÷ïr¶énÓµ›QíÏ_P|Í™ˆ?xíÕéZjYSw®{hGµ˜dÚ¡^a¡Îò¸8ÍåV3Mïóã•,ÊJ6³ËÛvø›*¨nš[ÄFþó?È`©èm‡‘8ƒýíµdîòáC0è(U¹Ë‰í sÿØyÓ“ñGÈ Qê,TWô3yùvÄè ÎÛ©ïÛÃ_p®\ßÏèjòá3ä±xJhnåYçT'ìIH_ãO²tfó²Kô>$SêrÊU{Ê‘‘ŠÓxvüòâ-U"µ¨\*•1‘Û<ÆÃéÊu«5RJƒqmÍæÖ¾‰mñÎ÷™">®]9Q›-+Û“s^7Ôiš¶ÔLz˺Ї¨‡’?­ LºëbÊûIý[U1(Hm¹æÜ¿ÑêfÈ¤ß ”ÏçN˹<;>àrdÝz @ ˜nXKh¶…®bh¯Øö8Vô0ʤû~ #Õ]3Œ°ÒW!è{²Hns©“# ‹µ¸%ݡ嘃uóÅrS±ŸlptÉàv\-æOúòŒF¿5ð2èTJØ_±¦óå´c ¦O;ÝÒåë@ŒËÛYp ܤA­ÞñׄtÈË»‚^ƒñü‹:ßn —~÷¨óák  vó{t ÆopSÈÔ%¦¨¢c|ü`7¹wÓ7#-›¯Ñ¡˜WÝäÂ%·ý‚qúV¤ª“‹1§@sÞTL¡d¶Dq†KŒÊ"!` |ø•û¡1{ÌHð¨ÌƒŸØCgâض~6N“6¬˜¾„o&¯¯¨ç¬}ËÂW1Ž£fæ·®Ñ=rûR¢; M©•|™¿8ßüΉ¬‰éuTqX]ÜW×!pf¬Èï §=ju¬®†Â;Ÿ°³ÔÅ2?z2hÇî6~Þ2}KÞú’ÁÃÇÙ'ÄCVÚ#ŸàÃw1&^Â+S1^Ä- gçbŒÌo•öUµ÷¡IëÌ;‹ Æ™9ËÈíüÔH˜“,Åi½˜£È¿`Ž‹‘dýÂð‰ïztI•)ÝU~:ªüöZ…Q­eP1¬´?6Ÿb!o±Ç¤&ü±@ç ·$˜‚_§˜Ñ’ñ€ ”Þº¦$˜Þ+§ÿöÆüRFá¾ô„užözF£ÕŽâƒ"I`šótÉ„n*RtàéÏ ¨oÁAþ§½;‰…õ?Fë 2¾êã¹Eöt•úºÍx9ûr¡Ù,2¼½Œ, ,Íç¦_>`$˜¨û’Æ•¤wvmµ¦×Žóô 6D†<»E8ôŠ¸ûe9`èÊ+ÿÝu´UEž‘’Y_;ÜÞ[y;ŒfàtJ{ t©Pòš¯½ O ? å#xØZ§Å˜ïØx7£a‹ £÷ÈWÖ$¼§|øè~7A~MªÙw;š^Ç€L¼#Óä¹ýBl%WUÍ+x u*r|ÁP:œ ƒv¿Žáa dN×ã‚W|VHwk*Öõ˜3H?%É{†Þy¯ïÞu§“®•È†ðŠäÇFÿˆ3av endstream endobj 1153 0 obj << /Type /FontDescriptor /FontName /HWTUNG+CMSY7 /Flags 4 /FontBBox [-15 -951 1251 782] /Ascent 750 /CapHeight 683 /Descent -194 /ItalicAngle -14 /StemV 49 /XHeight 431 /CharSet (/asteriskmath/bardbl/element/latticetop/negationslash/plusminus/prime/radical) /FontFile 1152 0 R >> endobj 1154 0 obj << /Length1 1671 /Length2 7621 /Length3 0 /Length 8677 /Filter /FlateDecode >> stream xÚ´Tl-,-]Ò H× ÝÝÝ :ÀCÌC‡Ò]Š Ò%)Hwwƒt#H‰è¾õ½ïý׺wÍZ3óœ½Ï9ωý0ÐjêpHYÁ- òp‚Ä ÈéÉ‚€ ‡äÆb`Ð…" Ù±ô!.®P8Lø2.0âÞ& FÜÕà0€²›Äñ ƒ„@7(ôî" »C­jœe8 âŠÅ wòrÚØ"„ÿ0[²@BBì¿ÝRŽ¨%P#l!Ž÷-Á¸%‚ðúWfQ[ÂI˜‹ËÃÃìèÊ w±gax@¶mˆ+ÄÅbøU2@ìù³4N,€®-Ôõ@nð»@÷¨%æzïⳂ¸î³t”TNØdÕ?ì€?›q‚þ÷§÷¯@PØog°¥%ÜÑ ó‚ÂlÖP@C^•á‰`€aV¿ˆ`Wø½?Ø u[Ü~_ —Ò€ï+ü³>WK¨•Óêð«F®_aîÛ,³’;:B`W¬_÷“…º@,ïûîÅõçpíap˜Ï_'k(ÌÊúWVnN\z0¨³DIöOν ë› à ñó ‚gÄÓÒ–ëW]/'Èoð·ù¾?'¸Àú¾ ˆÔrÿƒåã v‡.n?Ÿÿþ}ÂVPKÀb…aýýÞ ±þã|?¨'Àx¿~ ð×çïf÷f‡9xýCÿ=b.eum5¶?Kþ”–†{|8¸yÜB|B÷KË€~ÿôw þ*ÿ·U ýózÿR f ýQÅ}ûþªÄýÏÕ`þS7,€gP‡ß/4ÀüÏþ›ù€–÷_ ÿgüvùÿ[þ_Qþ¯ûÿßÉ»98üÆ™ÿ ü8Øêàõ'ÃÍâò_‚ä«A¬ nŽÿE•à{iHÁlþnÔUê ±Ò„",mÿØž¿º¯;( ¢ w…þzi ð?ؽØ,íï_×ûý† ÷ZúwJ9˜%Üê—è¸ùø`°ð~³¸ùø> {uZA</5€‹GÜ»œÜ~k¸ Ö¯I‚¸¹\.`Kˆ-êdu¯ˆ5âçO˜ç?ðsüçýÜÍé?Î|ÿÿãú—ç/i¨Í߯€ »—0ô~ü…€¿Œˆ ØÁ êêäþäþC@<ÿÉq¿Ô\N.p+7KÄ¿í‚.W7GÇßOüÿ ø?àßnÿ꺥›‹ËýSô[÷#ùëüû݃@]G ™Ä¹˜¼uÓUÒ½CŽ‚Æ^I·DSÜxm˜Ä¦ ,/÷¥÷t'óœ½Sé¬à/%¦xä­L›9øª—‚$!]¹Þõ£4ߥLñkž“©ö÷™»òï%JÔŒSÅjôÔu}?÷¦«Ùª ÒdÈ>¢ô¶ f›óû ç3<Ëk:ØÅg´Ëµo~þá~y!²T³]È¥uÄ™Q.“xS&%‰A©«À_û)>`vj(Ï&Ç—q;]Æ  çuôaAÓ O]<8mYa#wSŽž#OïT¤_ 6q*'ª ýÖ‘ñ)ZàNÁ‹–•&œ ›žŒÖ§ÿÇäf"ªWþÚ ûØ…Ràc7¥#áA$0öô¢ò qïT£B"‡Š”u\›Ÿí<­ëP Ýhf••utù†µxSs”²\=ËX®éƒh+MbüëÅtí”~üz³ìâ¸èÝi –¾IúLëâÊeÎ3dD(†Ö7Ú%U/* H8™Ï¦yÁ¼„§¨;º¿f¢òîq˜¯¢2êJ¢û»ð©®Õî6ŒL÷ƒ¸?òtµ nQÕ#i{áÖ°‚é^k=Q"˘²eÆ+žŸ*`W‘}Ù=(fßÆMM™Ã•Zð‰©¼Ã }à,YþQ0ž:œ•Þ-e‰û⧼éh!~"ÃR™1í©œñÛ‰L¾HaòëÆüyˆB§Ø ÌÂ>uÕçýÛ‹¸?JÎã%‘-n…‹>õ/Ñ; qš®çÇWa´KBÔÉ/È¥È>°¼g~2¿›Ïl¥ø šRrnež3èʉ™âSÈE…ÈÒH’X`Ê> ¢ùÒ ~âËø~: 9ͺAÐcLÈþ‹ÝÆ8¤8ˆ §mót-Â}?ŠT©;-‰­V¹ÈJ]ýØ_^ŒÈdµVaÖžw±‘«î9ÿ€aì·Þx(šºïI“º’ ^õœMøË(v1ˆáa˜oáb”{Ø[—÷-a ‚¬ /Lèds—¥OÈ…XiEãNQ0á;_®Æ¶Ç´Uï®”K°bÆùx)Ì>}À5ô©xBéèA2÷æÊ àÊ¢IFk@êÕx·R.=K2…+®Lë|רæ1ØDGpu$&„ª>ŸÜ…M~§ÔÏ̘Mœü2UÐyÚøĽC´\yHUã9;ikœ‡ÂB€vðèXó)ð± ýúƒ‚oŽež¨¹…ðÄ"ŒÍ0 ${&Mƒ‹;z¹ª *E¿lh£”§Ü‘üj—§ˆ°!ŸÙaì)®ú°@¡Ñ-í+ Xˆô Q-0jÔ”é_1ÃQñ›÷7¯¸'¶µ€k¶£r1ÌåÈ.…$’/ßTt$~ öHO7åVx¨3Ôz¿ r8‹f¾LßgÓà¾óU]ǬК½Íœox#,¾üêØr¿< Ãy’“~ž¬’7wñ8t}õ§Çœ»|øI“„«ª>‹ÿü%{ß ‡7Ù8²•0²7߄醭ò7·ï"œÊpÂøˆ'K!óç _§]ëoy˜Ì"Ñ%oAZ~ÙAZÍ×:EzÛ=×ܾMœ&áJ…å;¡%µ®⋲„º ‰ù§lã×ÎËg”Üש¤ØјŒoL½¨+?Ñ–]s|yï¡2ÚÉЃ1Ås®1O†ÿéQ]Wƨœà#äÅ”ùVß÷}‹ÃÑC’³uÆߪSAKükÄûÜcÔ¯ÅÍ5b©¾¼®4¼j‹$z `Ü·~<üT}ÅMW¬ÈŽ®Ûåƒ'sýt¼$ÈÞÇëQÓ‚¦öÁ®h/KÂDnÉh@oKw¿ôíÑ–«M¤ÁñW±Óó5;5µÂ‡þ;ÖQã·>¶±záIºo8Ê—Æý­„€ DýgÊÈù¦-¿ì¯Êd{G‘ADt§ä¼ÂÒå%èStÂùÊâñý¦‚)ýÓ.2Ï‚‰aÿoîM¤­ÜFEú¨­¹’m©®UÍç‚J<¯ZÓv•» k }ÍúS…ymG Ö'»z:/sâ9Üõ[¬[vi0ây5Ež'K%3»k9Q¬¨)…Žˆ†YMÈ×> ÉÄ™šèMqÝ‹~.Ϊ𷠩¯*]9Ã^)e’O¹vpE¾NlzoÏnZ±GIAT%,ìCíá {Ô(“t‹M2¨X_UÆ>iܸ‡Y› TígŽ°å=ºHÍ+Ä=ÊI““ýt€Ð{§5Ü· ƒugª˜ÛÉî²\ŘIÊþsûÚWÒÞ± »ª‚Þ¨¸lUkŠ”<à òµµÜ„Wá/½KÚ‹U¯$§ Ör™Û”úÃg×Ñåx_^)–nȾÝhÌÙÝá<¥÷O£7`Fy. úu…ØÄ5—~Óãoº¡ƒ1ˆÚ‘{AŠÕˆï;xê1D°×–,Z#žüì’Óî°V¢ŒwŽ ­šÀ§:c/?¥y]62S€ÇÁ<îñ¶áÖà-À)RW\ã!o5K=Ïbà„=ƒÅcO£U¦«ƒ/¢Bò³ËI6ërQ¶OÉ㉹ç_Ïr1ßÀß8m·'y>åÙ¬@»qŠý÷K­kM·†ªdÙ½K׉Ói” PÂ_kËFmÇh¨FæâÛôüCeÐu8 ucôJ–;TFYh„9Ïݵ×>Ö›Yz}¥ÄLš„ñÖ®îçÞtÂ%ó-rÇÛ£šÄµ}4ku–‘r°­:ë)Õ±GÛ²×1°£­³ŸùH`Ò}«¿sxò.Ot}z*Ÿo"'&< õB*ç¿z‰ªÒ"ãÿ•fäR’Âe–’-ÚаUw2Õä ìZzþ=ñÚe{¿~!Úç<Ü4Äúý•[2̧ò¾äE;YƾùîyþA|ÁáYùlÚÝøAÉG¬ˆÌ*«©·Áx0؆` Êp8š+í0û£ Uæä¹âb8 ³Ýð€+Km¨J.® æ^0™0q%2Q|-s^ùÇõ}‡… ƒd/Ù,êá+@ýK-ó’Å­u…Œk¨ô|šù%|§2­xññN~5ÛÁ†å`6E2ô 3³E•n–°‚E?M¡Î&57¤Õ=u÷˜Hµ8 â4c´NU&*,§aŸ„ߨ.f“±în [–ÏiÔ†b˜Ëî£ç/ MrúÜëI‡G×)œÇnŒÜ²ËxJÖïfzã1œÖ9«]ÞÞ˜˜@ÍŒFƒbÃ#k4ˆ& ¤Ð|u"‰rHwRÁg¢3ÏTÖЋ°›R¹[ðÈ ð3I⤦†mùf±¼üÆ®_S­iÎÃ{'³º7 C——Õ×_9hŠ<Ù¾ô°·c*¨¥ËÔ]E”cq¤Ñ3…ï­¢¦Î]QhPdÉ|§rïÔÐ!–èuí`œ–™{RšK ¦ÍÖéSG{91ŽÉÚdQ¡Öõ"mÉê5Z¦¯™¥-Ÿ(½ê&ƒ-|’aïa`}AŸ=õ’îOº¦Ã#‰’“ð>‚d¯|‰ QC](âìæ´÷aШìC‡o%ù=Ž@bµÈÈiOy@4‚Ûadƒ¶EØéX±S×Íú»hhd¸²=!`fŽ#dÝò¼0Ó {µdÖ÷®¸ªïqûÚ=™Ô®#3×hMá;SÇj»˜UO?1aJQFÄÛE=Ã.ªYÙkD…tnykÌ—3!sój2صbæ\wÀ*¢}Šr«8å€û“Ir—M“Žð#uê}[I:e!‰!9?¦ê 7› ·,<×ÝPmÍJ|N–Q¸8!3GIº[ic›bhÍø趟ËSuáÅ}G¢›ê!—¼]žãÈ-±õœò(à•ëÌâ´‹%Ÿ‚•íÏY8ÔÔWøkï¢`õ\ef¢Ób¸‹÷c4–ª?Rïíb3B,/¡/…I˜(1tÞQs{oË/Yä-Ô¦’EÓ“…ø5XG-b JøÇv(!…Ãkè©7¾é =EaOÈ°]ŠK›/ÃP÷Åkõ_a{jH\¼yr8» -ëT z?\I¸m$£b¾sÁ@ƒPV™ÅU÷Ø÷oÏú~Id›ET#Ãâ½IC¼=n—ÄÑXWRFê“ÕzŠ›ÃKtN.(Å[%#”Eáš«¯~ÓÙ™f²¬úvŽÒ¡Ö} Š­é¤2cÜj‰õŒØålŽñ\xÊ÷aC9È=Ç©z"z\O,‰>Clj=üñ1áÏ/‡ŽÑòë¦æGY #5]É%A’a cÜf«˜ÐÈ ’lS•Eæ6zˆ°ð¼ÃeÅ'ÈÒPµ˜Yjâ³èfŒÉ¶¢Å8š¶48C-u-0åÍ€˜™]«Ó`í»Ó~u]¡r¹Ÿ.—µÆÀVaÝèoåK=ífò—³A“[Z­¥·<ÆêÓD³ã¯ø†o‡ ç£2:σó]_9;užtÄëA²záò ,¬Iï¢Mw½¿á¼¥a82~Ó²\@d¡öÅõ”Q^+$o®Wt8Æù8™çÇKvçy3Ôx»Àê‘âLŠaßb¼7݆Û÷lµ$˜«5mWÞÖjç§fÎýàº>r/ûÁ«fûQ+t**…wÊ«~v ¦<;¦ì®Ë¤¬øŽYj’ºA˜ ÌoBûL+‘Ö†äW„ñ RK`Rè|ï1õóvC wchû†Ž‚úšÔ%Û³eö j¤ Âã]?,²HëB÷ïtØÄÕ”šYð¸rWÍxë%¡Ã{XÝ­Â2\Ø3Ý£ÒsJ¨­M T ¥­ÿ²XÞÂ1ƒ[ÿÔµýl¿¥KEèxPoñP½ó:àìöÂÍXçÍ@ŒÊ”¿úî\Càd)(ç´‡QõG¥Ö8ƒ {=·’׶ÇØþЫ¸‹·rMÍ»çÐlÄŸ7¶=‹9CåcZª‡’LøB"ƒËX—Õj¥h,ÎU€HÍ^¿éÕ¥QÉL⬳+5³Ê>rLѪ7ﾈ=W8ÊGÇAiáÌ—$zóxM»ÙÈokêª>±å“VhеŸUú_féZwtZËàZ›AÀNz˜õ†AH¢ÜŸÏeÄÅQghÉzwsm[×qÕJK™°B\ YIS\=(’Л '64èWÅxÉV$¼{Ó"ºx—ÎÉ7fQuóZ?!ÐèÅÆªÌ @~ÁsmLwzج<ýl·‡´]~Zð`óʦÕ;ýYš@.…=JP°Ù‰{Q¸,ãHË÷UYo;*|K”‚Fz#céšÐ¨×yè<qNîŒGÀ]&ñ’ñOÓêuç(ã>Òµ×ÀÎʇ¬ÎÓ¤iíziŽÌƼ)SUGiáê þÝêŠ]ƒ˜¥<£† v¤ð¢qM´ÛWÈìoà-ýº´öÏŽ{û±SiЉ݋ƒñëdâ šRFû˜RC =‹wºEU\:s¹žác0Jã‚åS. AZì±vØ€à@Q^¸éPX è{_¨~9²pgÿ¤ÛâNÄcsìaëæ\ï÷Ⳝ²Ì¤NVÜB¢ÊÀDÁàd\‰8¡”OBTÁ @øÙi`E8¾¢VL s;µqh Ö ªž/xÅÁ2צ¡Þ¯ù¶¤Æ²º•Ýh©è_däõå„uÖÅÙ;®¥Z»] í'`´Öùàët[µpxRÛe,ó•ô›<ß5GôßX÷«Ÿtdî}ý,þ©Í Iv‰9 ÷˜&’¡Y{K¯8V™X0XÚ´z©õçÎŽ>wõ!§èPrÕ‹j6ö)¦fϯzý–j7Kñ>ÕïŸú'  ÛóŒ4¢i^^&-óŒÝ/ÖßÆ?û‘&y—LfHø÷¼Žl dIÉq'Û5«%çïaD íoŽѨú；àÔ$ž‘L,û¦ô˜4³þækn—Å7»ˆ}EFWÐih [z1ÐœLmwl³´lÉ0ùB® wÇô%çí¢NEºÂ© >zmU'ߣÒú’½€'2qo‚ 8¸¥|Ç‘ß;n«sºšÅ‘Ȥx¬‹§1bÙ£ór¶TyÙ¡÷*Ô_ÑbnQ†‹™çóÎê YoÈѬ­øß;Öeô½öè˜Äï}Øb7½beÜßeVKÇ"F\¯Õpé^­8e8ÿ¼Ž̇&€¢Î¼]/ðôUq8Ékçª\‘~&6Äã'œ7IÕAQ±D!:a©TùÌÁ•ùI»ßŨ­¢üóÊ V1G0«ožDh—ŠµŽ¡åÛI£X£§¹>²ƒîFå%9ç”úœ:§X ˜µd$ßûyÚ~ZOáC¯ô}h7;uýyÜPçû³–¶D†9$¶ÌEµ&²–»Qßti ñÜVÇÝ6XŒ™¦,ðK÷…Ç­ŽÒ1 o¶17GË*¹Nç­B—ô¬@¹‡ínÏÞÙ­q$ip’Á-=+‹ãÙ½6d˜‡;«>2EÂåT,ºCI%lÐGLøÜD'MRÆæ…1âÓ)‹K^°C”°j1#}2tóóHúúBl¯æOvTš=ŸåD¾ó{…tÕ¢à ÈzÆ6ôLa–ZUgJ‘l>{V¾Ô‡S¨Ââ ÉC$pV6'h÷ À_´!ÜD—ådxR'b-ÏÌL]ï"ø|u@$¯Dâ‰ã~€& |íÞ?ŸµMÆ/© ÆOMµµ|½Yï}L@\þ³G3F«’ƒ¼|ÈßëjÅìf<*˜Ì‚®6ž›bn —› "ÈÞ¿ýêB¯¤j^Ì[ÕÂå«cë`ø(‘)¼E€+û$Q>tý.ౕƒã†Ãų¢ãÜ'šr½ÞiK焵:"[_Ÿóœ»ÇPhWž³j¡ÏbÕ£œ…í* &ülˆž_ SQ"#§kݧ•Ú÷ ïÇÓÏø•¹,&3Iíz9}E|LÉ®M³W­‚¹70û§ Ä®[¦JaV+¨eʆ(âà»4)4±:ŸÍÅõ±-ôØé¯ÁŒÑüèRœì…].¯¢Þ “Ïñ©+Õ^ù²0%ÆJ!Û:ûçLù_Ƽ­~Da>œH‡üv±ê€f~ûá‚Cl•ã E1ªÔg‘}¬Ï/)èâ!5œÄ¹Â*”vºÝßÞSË-õî„„¾ :Z¯éÎ䥧ûpêfYK[#¬*{ú¤vª ¸14’¿òI®Ž~™h\RõËëØhƒ¼Íättüt¨@ ½.¬%×LÈ´šÎ ‹që#Äk$Ìë¼Àå;§•˜ŒŠPìâ꼬z›Nö=å@P­—”¶:GN´uAÁ#¡,ùq„wgÖö{áŒ^Ì¡\ªŽyZàÑ]ËÛK&pÝû|-LÐcN"Íë¼µ;±âºœc¸Y—DIeMFNKÒÛ‹ÒEÍV¾ÅýtøÞƒ×Xm·Òh=+ÚJÈ:Z¢xyøÛY{ví!vlõÕuÛ¦·Ë±D3t€TùPNŒÕqHP„i™â2EÓ4¼OÒ¦ß}9âÝ=©÷H4yIçã†3z½ ÅâMÿÒ„‚ϯ¦œÄä]`RGì—úWüÎ~ ÌÕ—™ÉSî¶ù»Ò îUùˆP–økÒ±÷ÝÜŒˆÑÆÛ×{̦ø¹'ò©'#‡%ä7cÌ”´2ÊN†v1ž-Qù*'¶•3a–µß…ĬôÌP±£ÜŸÐ0ÝÖÿø*å¶Èñp±qØ®²hƃ7í™3Žy‹gÚâ‚#mõæ¼Az->¨lYlüîÈD½„ª«Sg‘ýFJýÙ¬ ¤—g ˜XPI)‹,»èys$TS_Mw˜`Û¢k gØ·i.â0ìiëtª@öá*V•x)Þ7¼scE¸‚CO…uá•oÓTã.HÔ·%µpÉ |šX˜Ït|h<_ÎxÒ»âN‘í}è´¹°ãsèwQt›,Tížë£—¹XÜ–âÆB‚áLò½êà+ï‡/k2 ‡Ì7ÆßÈçK3#`%ÞtÀ’DWT©ÂQé/c*§Œ¼ÉêLósé-ÆÊÏA[†ÌC ”J¤ÛÝÂ[rÚjâw•´ïi+o‚~LG$}i˜öx*P°¬ö0¢ì\†¤lªhÏ-˜DÁG8ÓÚ' O4Çʃœ•jÔï_Ðahúv ã‹÷i0&PôâU‡ßS6xåM ·A«ÐY–ˆÇ‡<ÚÂ}EƨÕt ]:éýë™N¤9"Üļ›bØ!ÌŒÄ;xaaÏœJù¤9mÞº?ÔÓaØzA®€ â«Ø×wÍq1•y5ýD$·¬¶ ¥=Ð…Iµmϧ§zÆÊ>õ‡Ùco®ÉòÒèE‰X׫ÖëÄÎœñÐÓ{æ¡M£½ˆwý6aÞ×4‚NÐõ0’Åó—ž çÍÚ‘…Ýb”2e"õ~Ú¬7óY.3˜~òmŒ‘›"1ñ¹3}Ì¿Losvf-í̸^rýp3 cÛ[‘ã§ýìÑó¨xÈY©ç¢L¿¨êÇÿV?î endstream endobj 1155 0 obj << /Type /FontDescriptor /FontName /KJNCRM+EUEX10 /Flags 4 /FontBBox [-24 -2959 1361 771] /Ascent 0 /CapHeight 0 /Descent 0 /ItalicAngle 0 /StemV 40 /XHeight 431 /CharSet (/bracehtipdownleft/bracehtipdownright/bracehtipupleft/bracehtipupright/braceleftBigg/infinity/integraldisplay/integraltext/producttext/summationdisplay/summationtext) /FontFile 1154 0 R >> endobj 1156 0 obj << /Length1 1534 /Length2 6784 /Length3 0 /Length 7800 /Filter /FlateDecode >> stream xÚvT”k×6Ý-¨¤ƒ43¤€tw#%20 1ÌÐRÒÒRÒ -!!”€´Ò©ßçœ÷¼ÿ¿Ö÷­Ykæy®}í½ï½÷µï5lÌ:ú¼²„T Gñ‚ø€âEC%M òDll0”#ô/œˆÍꊄ!àâÿÁw…‚QhLŒB5p€š›#$‰ˆƒDÅ@€(öá*P»Ã M>€E±É#œ½\a¶v(qÀßNk.HLL”ç—;@Ö ê ³Ãš`”Ô ÑìÐGXà(¯…à|`‡B9‹óó{xxð|W[).€ eЃ"¡®îPàgÉ-°ôOi|Dl;ò·Aaƒò»BhÀf …#Ñ.npÔ€ÎÐWÕh;Cá¿É¿ <€?Í€ø@‡ûãý3 þËlmprý`p[€ Ì ÐVÒàCy¢x`8ä'ìˆD ýÁî`˜#Ø Møut0@IVFWø§>¤µ+Ì…äCÂÖÈÿ3 ºÍŠpˆ<ÂÉ G!‰~žOæ µF÷Ý‹ÿÏpà¸Ï_o608Äæg7g~C8ÌÅ ªªð‡ƒ†ˆþÁl¡(€0P$v_u@=­íø&0ðr†þ2‚~Âè|}œÎtP_˜ ýC䃻C(W7¨¯ÏþýF 0kÀ j ƒý Cm~¿£çï ó˜Ñò€???=B+ ‚€;zýCÿ5b~mYE}ý{JþÛ('‡ðøð Üð ‹@@aa€¨àûï8wà¯ê¡:`ØŸÓýGDU¸  ö»t÷þ*Äý28ÿ¬ àß´h=CœÿÈß( ´FþÏKðËåÿ§ýŸQþWùÿ÷‰”ÜÙ9þ;Ø æèõ‡áæuýoÂCèï5Ö„B`nNÿmUEÑ›! ·uü»}0¤Ì Ñ¡¬í~‹ç¯î£×Î‡ê °Ÿ €þ— ½kÖèˉžÑ/½JÿN©·F@~°ìê ö"¢…%€ž¹½œ¨ç/MøùàÚàì†òØ \‰~NR à·r[;@Ñ¢ ê§ñ7.ø7þ{r„ïø­Žèóþ…ˆ C@]ÜÐ#û ð;ÁànÈ €ß}óÀÿ•­Éßè¿S¡s;;þñþWµÖn®h6ê—Ñ­øëý×u…zB­‰f&ÖÁöÕÁÍg•²ô¼kCx‹Ë­á±&ÝaÂ(ö©; ü4å19KHùí÷É:Ó%¡£“Oè¹÷Ά=Í;2?%©d£0»Wå¾òæöŸvØB™ :ÛÑäðUÊý8]FrK33Ž„‡“W!¢‚ë×­È S±m€ è„‚9‰%›ÕºiÂ̺Œ01§ˆ¦r5Kq3P5¹ÆŠ„EŸÊ5?§=÷Z5‹NXXèËèéN<ÈS÷Z¤<ô-µ·sÖûø^Zš¸ µz$»œð‰|I‚àþÄ;L½€@æé”ÄRÖÊb'Q|P“}h5jôÞ§Û=ž1,“Ò¶^Ÿ çä‰ö’´×åcéî¸uÌšiÒî >fªÙ<ÿIk£‡"ÆãÙiþ"ô‚o–µVnf|V”‰TyÏ`|øÄH¥øBÞÔHùi{PrʞŲ­_±JzpP3oh&¾釂 £ŠšF_ýŸsš1-ð@X|É(^¼ñ½s¾°ùt«‚îŠÖòéšÙJ\SÛê(/øªÒÜm3ÖäÏ"ºudkV¼KW¹4t'gïõú ž èoÆ CW˽sgD }ú+UÁO>‹”ò$oh}èB ÎJ3ó¨ ¿¹Wj"Ö (ï0ò ûËrùè Òmål¼âT³iïu)!ôòvQ¤Ü§•Ž!ar.Æñùè šõmÃõn|mDj»·´?–ju÷ÍóP{{"Ï•øqõÇ3û}"àq ~‰Ë‡Õ–tÛGùOù‚è¬ •_™'=”ŠÝ"è´þT¡ß DóV68Úpm·9‚U;Þ[AUfÐËÆ{Þß¹ëƒãó3¿¦L©Î,&&Ȇ±ÙþÜî$χݸ¼(Ï|ûœJñ°hEQû€à ÀL¢9ŽŸ€Øó\aRÞÄ ŽÓ¤tÛÏ·Â0ù½dÃñÒVtmˆ`ÞïÖ—Ò6¾£âã6ÑýMÐbîͧar‚ˆ#"[½¡¡~ñsäTóø~gëÇYík,-Õ¢ÍñüÞŒqníšn:¬i¬4 Ÿtí¶£SaÌÒïL"ᢑ¨ÌON‰\=Ý™CFv¥k^ö„® ‡Qñ_û:(èêÑG2ª”Èq6zÇÙÊJ2n]ö·ó|-'†,:䜄òÇÁ?¼¹_4ÂÂß»‡ìfD„u`YØᑶöþ(Pp*¦äÌtªg#²B›xŠ.½I‰#ë}tù’øið–µ\„-M[/3?®ïú$X;'hâíØysÕÞËA©©¯ê3ªioÜÖfã|Ϧ]«Ú©ÝKŠBGËyË‹m+w‰¢·É†ò¸òì¯Ù×4–·Éô´9€ñŸ t)5è¤ ‡ôy²ïÕìõ¿W¹~­³#ÀD=¼cö¤ä Pb“XÚ—b(T™IÀ}7ÉŒÀ42eléoâ}šed¥_Ê“-5Oƒ­d‡tÓÙ¢L=üŠI¯eî‰D/"îžR. >é¦xSëÙÚ¥=¿uÉOphÈ2L”/—ÇåuÖ«ÎHt—î*³Œ †ZûÉ|ç©Øún4îù#žòÊKÝcõJ&ÞÝÐn·`²ã ûÊ€²xÇi™‹1xB0쇂³Ê9aºGðÁBÐØ*¢m¤iPV=…ù8k;OÛ%wF+¯ˆBXw¦AdîrSwÕ~¾ˆãߪ³¡ÄÁY˜Ã¡¥®ì‹¾möì *2ûe‚ÂÆ÷Ïç£7g¯Š[Vwr7…•4Ôåî~Ú"“¶%ØÎñ¨R[Œo80EKtù?¬VAjfÛNãe›ŽãË÷VŠcd¦cw Š’=õ ¾†ÞØS.‘Lä<*ÛÎzÍX Db‘§F¥Ž‹|⵪xO6©U[ÉGƒA7?@ùqˈvÒãVzBKÅy¶6¡É¸-N̽U Œ‘„ÓJê¦|*²áwwÇú«ŽÂfp  Øþ–‡ÈDçáÄ÷b’6ÂRn bDå#Öú/¾É2Î=hèD±”Xµ®Õ¼î%°û’–*a€¸A®ËÆZ2¶òòŸªî^Y»ñ¬´øâú1Ó—³—1o%Ô>Øí‰w)Q&gˆgte]Ow¿Jx€í±›G1koÛ³‚3²²P±µ„0­éûääNó +4HØJK“bs'Bäƒd¿^Ž~."Ž. 6iíÖ·2ˆ _ÔÇþÀf)‘mpxf=1œpƒT>\7¬ñ€sù‚nÐY}ßw¡1Š1<ÍpƒÔ÷´§JÕdÎä…GÊÝ!£œ~1®†Ç[r QS[ˆ‘”~ɺrÆŒQ7È@h©­Ã*ñF¢—¸2¥Lòȉ1sˆýf'˜²(!4¢¹U3õˆ« P3º;æ·¼¨W}tí4Eø$±_qÇwöéŸ>Áè +Ûuýc|£cáÏCKQrNñ¥Ž¬*#½CÅnÎs(»"‰€C6Â3xªé± ”Ö4í̾|¥u¯Ö˜±ûþÍÇÆ¿’yA«’çÞë”ë,AÔeÃÓçsíGAûc¤Ÿ><ðôÐ[ny! £ª˜Úg”èiñÑÖç={¸wò™uU{ZUЕ¸yøå dü>AÚ¶(¬°"ÖPŠ ðH«Û\ØáÄD‰®iÄÖe€VýƒŠ}fëh.î ×Üœû&»cGé£A{‘!> åË?5±ZÓÜ-ÂÂ~h’oDX™–{Tͧ(MÖB³§²ÔÄ0(jýïDž„¼[õk("5†Î,ª¥Y:ª xÈ÷úƒ‡P®nÓ½ŠýU§±¸ buêµÔçCBÎþ‚´/áx-ˆ9¨à á0¹í«†ùNbbB—go¹²$†¦š}l{Y¯a$Ø­˜÷`x)¹‹AÔ¤ã"ª‘tFG¨¹8}Lºìv/FÂùÁyÙ£‰ Ç¼×2diYÙ3ºÅ¯yêmL|—0•«¬Gê×÷g5wM´PN‘J¹Ôd2ÆÇq>ÒÌ›­-Š{®CKо7Ÿ!ûcÃ+u®N’+ˆ«äm-å·[_b»ª×¢)Ö_r¾*6”9vig=[0^žzê¿ùd×¢†Fyû„«¡}™ØOtúæWF£|Hƒ!O[i‰YWËüõ&ÃÒ>†™º÷Ä¥rNBÅ"éò–ÊU öÀƒ\eÈLJ]q©Ïà.`~¯µp˜4w¿’ü­Tü½ÛTÚ¹taN–NigZJ~œ•[îväŸY“5ÍM8êÐí ñsëÀ¿l®Ç{¬UàÍõFkz`;j–_ì%ö[ƒgàä€BÞ\ JHg)D0§Ô,¾c÷pzVèÂé®76ÛâîfM¤õýž‡MÌCƼí2‚äVñÒ~Zóps2(ÆA‰[q‰Y)°+W¨ ô¦L1oãdsè*TVHCÛÒRUáqt”œKÇJæé{‰¼ˆÍXI)MsœJ§Ïj²{šÙ½°Eì²ØçB x0¦ÓêÇèÄ+`‹c–b×£÷r Ý$“·2B„ÖÌ{É[bö7¿}¼‰I¹usµÙŸµîX ›e¦YÆåÂ*›µÎIü’}¥–K-Íý³‚1/˳Å\í]0ùr’¦ÜÜM}k_Z}LÖpN~¦Œãfü oã¡åŒ¾¨!>ÿÏ{×E8~q),}Õûz¨#‹P&(Á eqilŘŒxuÛ šN |;=Ü)™Xø^æZŠn²ú0qp“è‘b7‹{zBûË.µÛµ‚;–ª­g\ù4æ×.Ã[ƒb{sNÇ)I¬„çzšea7KßìJ²ž€ D¬Î’cz$K8ÉOË+³ ½Æëà$a­»SQ%÷ÅðÅ ¦*×90HÔÝ»žŒìů҂ô Õ™„º ¨p2Ô ²×M1µ7—ô‰êŠ7Ê]…ˆ&†Ù\Œ/š˜õ·8™ƒ±ðŽgï/ÐlÜ œ}{Q|4[ƒêݧ,<§­—æŠO#P‹Î÷© KJ¨ŸrÛ}ܸk@|–ç~Öä¤ÈhÌ·–ÉÿUÚ!yaéq‰¤è#^¢²˜•”3óvÛ*ƒ@•)b{lƒ^\³pZšïÑz ¶>Ùü6·ü€ÙÒA§ o·² ßÕîÀ-°kRb>²s[D3öÆq\YU_x©Ö”ƒTB£VîM¼qŽ.]/ß=²Þz c¥\Ž‰r8Œ£ràÀ¨‘0ùÀ'ý,ñI§©õÕèÒ‰rYÕ¸E†q &ã‡W9$BÛ»­¶~¸RÉ5â¨ÄE]OnüYücˆ£I¢÷ Í sc4A«÷-‡Â¦ K©û'—DlžÊ šXGœ@XˆS¼C¬û²Tl¸2üúÛCsç„ë²ìðwóx[Šx‡ùÍÙ”œU$õÆ‹ä„éŠizâN Â@Š±Fº˜ jû¸F_liºªõÄÒ:Š“”±›#µNú|hXDßׯH†è„lØróv6†WL¢l8üC\Aù`ÃÕmj Í·bne÷ó²ù›OqúrÔ¦5¤pryY˜/ŒéÌl Þö²ëîth_›ÆñŠd`žµÏyá&TÅÒÅŽ¼LærÈÖ}¥%Ò ºG™k*Üø&ãp`ÔýNMÏN»ïTq#é纑üMÖ„i'Ìkc´ 4g09·|°‹ju·ÑòíCŒì`0d6ÉÆÔÝ1N´)ð>„÷+¹tIH;?úø€tƒúžuŸýë"cí&•Åð×Õï©Ê³»ŒCÌ[¶y¹I ïÚ¦Bãæ%,úظ‚7«}ù²>=ŸJâÎ]U¨ú.×¾eUäùD9¢ —œüáª+ç¯G¾„0ûÍYzÆnélghýöò–1`ÏñûÌÏP­ö‹ï{´jì1¬¬™ OÓò-§@ï4;Ct˜’ìã\b\­ŸñpÃòNL×Õ\IŠoOd*Ȩë8y—‹¿Ìªø­ulüªéY‰‰?f¯ ÷Ë8y##æô3_õXoœSwær\Uá/AÕU¢ã£]Í^|AÂuÓ‡EºnPå4á{[^Aº—ø?ÒQ”QJjbÜïQí°õ¹‹óÒŸïýfÍ;Ù#*‡æ"&éû˜ç·#DÝâÕ¬ÆÉ+“je8E†Û9D¨4Y«Xõ-üa»=wÌ ¥O{€ Œ4þSJ¯7ÑŒj5FoŽg\s’o)ØbÓL –v>(µUªV•žŸ'_—^ËÛ)Œ:D³gòt5Â7^ÚÆLœô*/ù°ó¤îñ H+Ó„èkXVêU™‘wdn=màúŠƒQÏ!›Îâ•œ1ßVbS®+]‚}“ä!ƒÄ·F–D™*¦QºûAOœóQœ$4¡rÞ›nyïäžii*06ö Ýæ fˆÃK+ö¯ÎxÔÛ¿CŸâ¾–ý`謉eFìX( š ~‰5hz×€i\l,¤øä!°6}ºÒ4Îp¯\eØHªÉsKæ,óá%wîuL>P+GÚͧá(e7åspˆ+ŸçŒUÎcqNù€àÑ‘Ø2 é«©Ù. &¬÷7˜RO/ÑrTÉ4OüJûJãKv~¥rnl8š#¶”sÕ¼þœå$!0ŒU~üêfÔðê.¿tWÀí.€®>×Ôqw¨<^„ø¨ÿQ¹°QïÚþWšƒ^­)±;2;/F;¬ß|Z«‰£tZ™ŽW«ÏhrËÉX@øJ¿8{:À¯üB´v©Õ%âBáiÏ%yn­I¦Œdi æ¢øËÎf?›WöÅ‚å—p‹e)ÅÁ'+—ß­³3~¡j瀤úÙ%¥_¸ˆìƒÃ[‘åO‘‹ô|å[í» kÍv<“÷ˆ(\5óŒÈ'•!åw¬–>.^Ü #>dL(ÿ ß܉,$NËÂmÐ]ŠMµ›eËÚé)Å—?ÐSN~ÞžÁÙuæ¹’¦ãüKb8ñ;Ñà æ‡=æìb!­'_ye#‰³2lâØdÙMõ~UUšb¾ÍkÑ˺‡¯õœ—FéTÆÃ…ÇwDZµ5ÉÃI=ùPÏó¿?Ð5IeÌðe¡ü_[¿ÑÓ³'í"â¢ÁÉÍæüÝä«7(csŸIžCKëáM8(o§/D˜4Ÿ®à݃¤gÇÌÄyl[ø] —ÌìîɶŸ½“<ƒ³êöI †Î³·;+Gc¥z¼©Ù úÆG¾]K'¥L™>|ŽŠžS ‹~ÇÛí š-ñ¢h,x+ÿøëý” R)ý«H!¾±[È ¸þÀ«½#ì**œé¦#˜ ÃÊ^˜IÖŠ©&}p濵Âjú`¹—Hò•Âåëâ·¹ˆÚ…¨²Lý^ýaLYa-ï­= œƒãÌ­ÀH~H©¤ç;‚ÎØï;um‹—}·ïg¤ï=›3,µ {—’çIÖGžÖRþ,œ†Óô “èI8ðë"îWÄ¢G¸;a5æn†/·Ì.ŒÍ•Þv+‡ë›>}Ö'ÛÐkú~F%—°¢¨á·vdÀ7Û•aÌG«ØxÂXq8»Úp£zîS3­×+ëùJWÿXIxOŽN+ÙáÍ‘`Ïô¹~–T¥¨KÖ¨‹ÃèÀ OsŽ B´\›Aߪ>vº½¡í»ÈDQOR·œ#4ÂÿræÇ M ¡ŠÃ…³eÀâ©B¯í÷U¹¨[/± ˆ±ºV˜y›ü4íûé€mo“©Q…&·.â»ì¹Pie¼O± Zhëã¡g;ªÉ V> endobj 1158 0 obj << /Length1 1458 /Length2 6558 /Length3 0 /Length 7539 /Filter /FlateDecode >> stream xÚtTTmÛ.ÒÒ"‚8ˆt5¤t·Ò!Œ0ÀÀ030Cw—t§¤„ÒÝÝ¡¤t· ˆñ£¯ßûþïwÎZ笽ÖÞû¹î뾟çºãabÐzÆ%m‰xQ@ÀÑ\@n^Q€¼®‚:ÀËËÏÍËËGÀĤEà `&=ˆ# Š€‹þ/‚¬#Œ¾ÁäÀèž:Pq‚€ü ($ÊË àãåùá( ;C-êÜ‚"`’E Ý¡Ö6hQÀß¿V 6PDÄùÛ mq„Z€áu0Úb³£x†°€BÐnÿ Á*nƒF#Eyx\\\¸Áö(n„£õ6N€ mx AA!–€_‚`{È_ʸ ˜:6PÔ_ø3„ÚìÜ0¨Žºñp‚[B7›ž)«4‘ø_dµ¿œ€?¹¹‡ûãý+þÛla°G‚ánP¸5À ƒ4ԸѮhNnù‹†¡7þ`g0~qCø}r0@AZ¾øGÊŠD£¸QPØ/‰<¿ÂÜdYn)‹°·‡ÀÑ(‚_瓃:B,nÒîÆóWeíà¸ÇŸ…niõK„¥’Gup‚(Ëý¡Ü@ÿ`Ö4@—ÿ¦"|ˆâjaÃó+¼ŽòÛüß(ðò@"«/¨äæCà;ChG'ˆ—Çÿ6ü{E,¡hÀ ˆ5NðOôbõ×ú¦øŽPW€1ïM＿ž¿ÿLoÚˇ¹ýCÿ]_¥g:²:)þÛ&#ƒpxp\|‚ (( ^ÿŽò·þÿhÿj¡ÎÆûO@e¸ ò—„›ÜýG†óŸ®`ý31l€ï ¸ie€õŸÎ7á䵸yÿ¿ûÿ·Ëÿ­íEùuþHÁ ûmfýmÿ?Ì`{(ÌíÁ qüo‚>ä¯ùU‡XBìÿÛªŒßÌ„4Üöwò (¨+ÄR Š¶°ù«qþ“û›ƒAá- úë†pyyÿËv3ev7·ê¦B¿M›!ú÷–òp „å¯iã€Án¼7MÅ'(ðÞŒ¥%Äõw?x¸áô é„öX! ~ÕQà88Ýdûþxì¡p'Ô?/€ysaÀa+ô?(ðúW]ÿ†ùo`ØïÕÂÉñ†þÝB7:þ³þ}K@ ® ‚3 ±@ÛÊÀ¦ËwÒ´.\øË+-a1†Ý¡‚hæÙl5¼4Å sËÒ{ƒÉZŠƒÇg3zúB“¯Ò|„ìhíø߯h¬Rðg¬ËKE( Òœyºó’Ë'Èé)v·$§>_±ö.RJàRÏÝ$ø׌òFC=ðZÖ5DWÅôC ‹F‰q_½w¢Ó5vÙ©Ï°š–yŽ¹¤dßMþ´øuZˆŠÓx‚Ç`D’EˆéTÚÏ*½øj¥Ãƒxÿ%ù+!?;‰as‹§}¼1?ÄÇ•GîeokøË¡¸ ¤).Ø°Oõ\Gv'œ3?†òˆá'p )w‰|_3m¥&”!1ÞLEÆ‹¥=téD­ö$§dÞ:æv1Å]¶xe¦.]+Ytò(5ê}ÇÜ>$$…~¼ÞS­ýC~‘Á µÉ…çÃxý󯾕*Òv•ÑºâhÂB©h¶±ë/¬+l:vO>~o+XÏNªæ\ˆR±ÕS{–6DÎÒÞ’½Ú‘-–Aµ›¢Ý#j[SN‹%vs¦À´ºbIýڞпäióü‡OaO–váeîC\ã!žá;V«u]wsùëú_Úhþ|î3Ìÿ–ÍЈð1¢!½*—#`ÛÐ ÙþÁ&1fØXšƒ\1 ë}£É)#ñ2µÔŠÇ³Tòú»¥ }xO'ÂDv8Gðma+{ s©þ‰ßÇ™3éû‹HG±ŽÌ¼wÑN¾yr•©zê¾ÕIâu‡úݬOcÚ½ÞçÐG}4ùÆtä´¸²˜Åœ3E ÛÚ ü\=z§ÍE·' Ýxå•œ38ÃwU#|E=P·²>5W³ýððqÕ`ù.WN¯%«ôøZ n¨ ,¿øxt tµüÏ‚A“à‡8¦MÕØa! dž®#Z !¹¸»Yïlp0…'&:—I{«ùé½-Q"q¯[n³ý}„ôæìz>ü_0wµIêø-U=L:³6Œáë²Ïñ²î>T®ˆŽÍOÎ%–ç…ûa†>PµßŒ;s/œˆŠ“#ô\P2;œx/0h;¿¸.â nÆÎ|¼¼Z(#’¡")@N½ÿ¼?=1Ç©}ås@´ É9A†®{éfËm…BF_ߺS€_Â].€ùŽpÞÊUó•–-p…ÕøÙ¿5Š _+œø¾Få·Ö:IÊ í .ÇRšÉ!{²xbÊËm{¥G_•É”v›5Ï/ ó¦²¬v6Ç‹šƒg®ï½è¨úÝ‹«z™V/¸úê.¾ðÊ‚5 Óœè5uªõ)DzExloи‡ô6Xˆßz¯é62ù´Z[Ù-FŠÃ}@éKñD%Í~D6f1wMg¾¢­h/#K,ŒÛkµ9ìÀT² RgÿüFùí®©mÄ« ”¿j6WÆtŠäqöÛî}¥Þw3›Pû;<X Ø>éuþ¡1™EóŸ>~:º ªwm¬Qàp»šý"kñ³®Ü`ÝfŸqK™rùÐÆC›7Ü#«Ÿ9ð–9ɼÍy¡RLJñ8dË…]'® å$"^8ÃE\?‚›6×BŸi²Þ~š>;i8I¾Ì†o—ÄŸ1¦q ™‡ãïka À®OutÍ)± 2e~£û¼D»)Æ×óî@g^Ý×̬¡Ê±6R%ß5ì „¬¾æð>‹íÃÊ/É£”öZ±ùélßÞfzƒ—á$_Þíå=Ð$rôÄ}$8¿Îç ¡ßh3@dH>Ñén4CªÌf Žq­·ºEà“MXÈ;XÇ]œÄÿ´2¯|%;Á.ÂzVV4‘'±RKž;V4=p«OtÊÛy$f¡ZxchG´{lÙû)™Àwf¨Å¤|_ÍìÐGÖgÙMÇðrrÇí¸¦ÁŠžS¨ælT¨W{8Aµ$¾þÞ´f^ÝpááAº»½}œ0‹¤o¼Å$˜'r=¿Ì0iFÓÏvÅMÒÆÀnÈ¡¹)Wk3dÇÖÒEÇ`zdì¾Í?þ°AµM6\ãî›ïHX]¨¥'‘ ƒýE'è–·!F¼“M^o~ÿŸçD×?ëÌ¢y‰“‹JQ&?qÆ¿06X#ÀsÈVáœÃ‡’ßYÆžë‰)ªî âhó:Ù_†b6kf¥·Å³ð sÆ\ïj»º…ïÓlzr‰zaë…h¨AÚˆpjãŒÇ‘6Áë‹p!åbaÖ͹Â𯈕û,»â˜=ÜZŸänÃQ æÁ¯ ùäa6™Ô¾¢x¥5ÉáòÓyP±=vì¢Øc¬ù*A³Ñž;²°WÔ;9çn?í»Æ•hiÇ%¡`Ùßì\vÑ·9×GÆB‡“í¾趿}£±¡‚H‡ãÚ ßY°Ûhé— lxlK6'è„Ïx}pK+ý(M¡°»¾Å Ýв|q¼Ìp¡t4ý¶TίÚbŠœÆÒi’I³l;)ù’ÜšJ‚´caîb¸Eë§Â݈Vêæò}JmSb[ÏÖ<§—}Jô[WÉݤƇ.är…\–Ò¤(æcžæ€¼͹~§\ø‰Ÿ­ÿyÍmæQžð-®Õ>"u°zm¿¿D¯Q‰0-n”þëüŸÉ¨Ž{È“k•ÌO®ùñÅžÕzñ Áß•—XÃø^‡c¹a¹áY48x1ºñ‰¾æI_/ØkËÇl7VºL”a¼·iìóÍa`²OýÚs¼lÀgí™GÅ¥­Z#$WÍÔ¿•-Sñ>ŽS±„m S•h7ö»ŽÄPI‰º†:'¿Î{è@Q/ C­æ[ÉØê³Æ¼¤óO¿ߥ /.ix ¬¼aäN@Z«Ó}ý¾¡ä©oãD«áFÎ:a†!–¦«Ê6?ÆNЧí»ã>ßÅÆì¾yò½ ¬£À`Æä¬t[Âœe3! :S˜¸:YËRGXû?Ê]šóJpç¾à(¾ûùÉXEGŸvÙ…Û®„†HövtÍ„ˆ×2é ÌcÙ.AÎ) plOX^íªðÔYã˜æ÷iêïÆÉâ§×+|X¾Ì·OM@zô©@„E—3È•I.^LZú’x‘m¾ñ{.CõRù:'3«Q¨½Lì9‰À¤µ©–Ÿ´*š–œ¤×è Wnà™§ ñ§ƒUV†¶¦_vÕ<¥‰ÂtSØd¼$™ýgfŽ.­Æ—<¯jXŒ…â/Ç=¦IÒ@­äü䙸ÈÔ.+‘GÙ˜¤¸.eóSpw”¢Ÿ&]oj#Q•Ðg_‰¡3ôg$2¶™ÕRf×OyHKNÝb;,‹^>ƈÙVN /úŽKÆ9EY«Á{ÝN\1l艽Ÿy~+š÷î¡ÕB­Ÿ>} —cÜ©u{¼&uÃþ ÖÂœÜÏ ú2TW¤U X¸]B52d߸šݤ0C :’í¥À«]W9ÅP®ç" [ }kÏí|}iv¾ÈMS¦&ð$tž÷5ÙRÍ÷’ûLr2ݹGªœâî ¾½:òÞÿ1„ù|sšàÔ¼ÚJF«¸ž<± Óþ^M„^’Ý&ß| ³V…}7wç•ÕlsêçÛÆmm€–X´óÀzU:FRäÇ9iüO‚Ъå­Â³9¹ÅZ¹& ì ŒGgI¼~O8~LPãÑŠ=\ô 0æçÏ])IÏytµJ¥ÓGì-žÊ2ÂOv„ÕÇ ù±Aý•ÚË„5í¡ÖìüÌóÛîI¾ÚÓ©fopé&N·_=Ž³>“^þq¹P ÒMތ٘ÒhSFÏÕk® ¬¾.ÉäÓùºÑSŒ¹O.{¾×‹ú{wf~ô²Â÷«’M [Q¬:êH|Sò=ýRÖii˜i<–õÑ(uŒzK¬PW׌ê†÷›ñ7¬YÖD—«O~ôaÚ#5ý.Éß\¾®"éJì šYíS0ƒé­SN’Ò2LN‰¼vÛÍ>&.nWZ@jœ4~V­…zÎ\­«Ñ›»ˆ¦¡öd„éw²sÏšÔè½¨Ý µedž{}5 ¶T“±M¥£¿P³d‡VxϱՆ½ò[ˆ½¿4üšVÊ%¢ÝW?'Íäpô€ñ.#‚‚xÜ€û}O€ç5²ò1ï$1Oa¿Ôù~{$BßÏÕØ9ÕLë !ë+¬9q%ïqÃcwÌ6“RÐ~²³ýág|ÿÛͺ©yN|O£È¨>$kÁ×ê)Îת^ÜÆgrÛ"Ô¶ ±¹oªñ8,ÔXÁ§òù ÈBm ÿ7{§Î§dÚàŸ4ܯ‡|1ôÉi‚b\ŠúëkSðcŒúvPŒ©Ðâ,ÂX|§QIÞ Á(ÛêÞx÷ŽîH ÌÝù¦‘OC &2’–¹5ñçݤC(•1­AÀbäÎmÜì©ø †oä4aìØÚQì·ä°»1ìcN¿FSâ ˆ÷êºÖ:Öݺͫ è‡M¶—h©3ÄE‚ µÙÖPµ.¸º¸V0ÔúñC¼*™pÅ2ßþÅPÒØäNÓE¥DÄ7ª,W†{«i£ñ / ¯¨X^›ØŠUPÄ/P šx½k~øD^ó`Œ6æ²—v»=íò}\ Çüë K·Ê«øcK<°´®ê€Oc›'þ!ˆ¯ýжú¦˜’—`2PžˆñBÐÅ nv²CÕ½H¥* «K¼¾jr·ëûÌeßïè~§&{ýˆÑX¹[ß—mò‰c™œ7‚«ðƒ`Úä›Ä/”8él)oeçAÀÐ呸5¯dŽí6Ê­ÍÛaïé’1~‹ª„€6&é1|le•(ã:gÅ0f¸µ€gsbJ絓¾ ¾`µ”•™Dè /e3kWg3óó2kRP¾ú-»J¾mʼn® ƒÏåj‰2\…»ª¯AÐ4Êd°Íõ¶÷ÓËí×Åb·YÎë%ÜŽ¶‰æh7L¿2°Ù†Mz()&ÙVfŒ7šGMåjE Fqk¼œºš¨i›ÕuJ-åÝ/£OYçŒk÷G*Ûœs&#€ó1õ$M7/I¡þì‘Æ°8G2¥aj^!i—C}%y—œ58ðAßgÞ_œ0„º¶ÞJC·t9뻶JQZA¶F …þ$QOƒ»œ³RA*EäGV¨ZÄ·ôã( ²«Ã7B¹£4' Å5øc«ç?y`gŠç¨ouÆã‹JAëD>a¸!e“Ò¯Ø}á—g…³XIgò ܨ{_%Jù-càú‰‰…âeåŽ?‚Fs|ÜÄ®ú"jE]è°0ô£·ÏÂOÆҕƇ¥ý³@•,¤òV3|w2(§O®¡@6I8¾Ë«¼ƒÄZž«®Ž\Þ ZÝ©0¯p’Te:''"ˆ¡‹?ðôL4ü0¿å]¡žµçqͦÐ÷í‰4ï`é²ãkÍÈ-ÙøØ 2€>éÀkØ·1³ŽîÉID=´;–ñ¶@Êdçå§Ì^²æ¯³Œ,N TL¹]¸QÉþxò„Õû>Ç"}»©Ê5È%Û÷‘èµ/òœAÅOÎNÑ´ó*ëÌ3EÍŠæžlÂÆH×ã0¡·±Ê×wùã¹’¬×8ÔöRÛRcÙ†YÄkS]3Í ÉŠ÷;(ƒ¨ùE†\ëqý ŽžO»O\µ 4¾>:Þ3;j§>•ÑVJNuËö­ŒÞ;ò¨:2Uè!î½­Xö·I{G†MDôç.ïšWÒ#"öóØ0vʨ>¼ßðñþ‘Ö£‡OÖZË”‡œQi©X<º2g 0*¹»ö_É”Xãúþ$8ôL ­Äf¿!åRîü2ÿv¯>M(€;ÕM,ã#±¤}àÄѳáJš®\¸·/M£Þ]m‘7Bóþ•ôÞUgz…8öñYêIý±²dp’‰f;qaøú:qg‘·_«[XŠ-\ÞNkE\æ[=ù8eĪìŸN-9oüV3‰¨‡°ÆÍ\*}d`ç^Ec(ô[µJY–×Ùþ\k=ŦÓumÔ…ß'ŸÌ†MVÿO–ÃzRR¥%µ£™P”²iqêÌY÷ìçú‚ Sþ… òdyõÓ7 HL»É¶·ª‹ÒA)øÍH—iÛðAÍ79°"鈴½n]Ñíš©ÒA÷øâÕÚŸï,ˆ‘çE{jÈtØekpŽ—AŸÔdªv·Íð2ÓÃ5ø^nï·²š”\ÆåVèÔxxg„º©—¿z\ ¦ŸÞ.²ÛYȨۯ0ŠÒ+Ðéð“¸Ð5µ*ý§:ŽQ’ Ë™©¯WØ'ÙÆvÒŸT¸¨.ëæZɼnƒpJivfUOs­…7›É€Í÷}n (õ'ÍÌÙî#õ<“vhùL†_ YÎŽ±Xþ!÷´T·ÚùÛýèjÿΦ{B¹|æ–=A¤~‰Ú¬¼¹g“Šfš2v¢9>³Õv€‡v90Ím@ï=4…[ù´%žço¡e>®©òùµôˆŽþ|mO'>Ö/Ö¸+óŽ!Ùšàô4Ë´^Êg‰˜d×VÓÛ‘XnäÙªSWIo<ÛéÝÏŠv|²3·æælvnÈÇOÛ¢>Øý†™8¦®öºŒåâ KÝ)æÏ+…‹Už±õ¬©Q<ít:ôq±•’'ow`eêÇXk¡v<¬Mÿtûtä+öT´˜$¾w&]2Àm³€/µ}çtíçÊ .5^æjs²àf?¹yJ4ƒéƒŸ9úšBñ ƒ–•ðIòü@VáËO™Þ,áRL‚B~ÌÉχdóÍ)ê}ÈŠ=ö%±|ú5Éã¢Þ©¦SÈý™óEè‘”]ç:Î"ª–mÎPl›*M{éÐàŽV>$Ù<ÌF\·•ÔD`}þ­ØžÆƒŽf.µ^AJ~ùÔ `UOÞ\>S† nëˆ5{î''XÃ%’ÏÞ¬o¬ÆØú`"ߺ;Xdá´….ô3ÿÚíÔ,Ìa^zw4ïºôÖ¾?Úå$dd´¦1kA¢®±ªîÁÝm0¨l\Yra–J‹àwcž0jV£©Á4&ó¾N©ût0IöÈ{ROIÚhè‡Hã²â·˜OÞ½â¯Ë:ŒÉ¥Ñ› Ñôä ìÂæé Y|KÚzl°JÑ6TÎÝ#„ü|@Æ·ÔÚL r›åæÌ0j ͳ–¢ t;‰·wȺœEzFØÕýZbÏhˆõÍ©x‹QrbM¿}%é¶ÖéŒÕðÉ~ébc+P’i_öÌ–ÍB*J«TD­iY°ª£A’f—9 =/°ªÙ‡Ì!G£?¶Œ…†]IÓ[䈲 àOÖ’)*PÉ‘w²È ‡\Y¯Ž©ú¿x¾hGé×Æô4&=”Ûª½°›x§—x·ìg‰Ì– %&(Y³_à>³qŒJ¼¯?è™Hb]Ï)X6½ ÚòIß¾u½{ð?.ˆâ endstream endobj 1159 0 obj << /Type /FontDescriptor /FontName /HSNTCT+EUFM7 /Flags 4 /FontBBox [-1 -257 1158 741] /Ascent 692 /CapHeight 692 /Descent -189 /ItalicAngle 0 /StemV 82 /XHeight 476 /CharSet (/equal/minus/parenleft/parenright/plus) /FontFile 1158 0 R >> endobj 1160 0 obj << /Length1 1433 /Length2 6802 /Length3 0 /Length 7767 /Filter /FlateDecode >> stream xÚvP“mÓ5 Ò¤(Mb^C¨Ò¥…Þ¥£H€P$¡DªôŽ€ôÞ»4¥ƒ‚4)ÒAŽt¤}èÓÞçýÿ™ï›Ì$÷îžÝkÏ^gï +“¶¯,ÔÉvB yA|üâE}]9?€Ÿ_Ÿ_€ˆ•õ1íûËOÄjC¢àNñÿ@È#aôµO‚¾j8!ª® $"ççðó‹ýtBŠ np(@ƒ ê„€¡ˆXåœ=‘p[´8àïG‡'$&&Êó; ëC­ €m s¼>Ñ âÐs²‚ÃОÿ*Á!i‹F;‹îîî|GŸÒFš“àGÛta(Ò ü¢ Є8Âþ¤ÆGÄ xl GýÐs²F»C0ÀµÃnC ®S\Pp}:@OE å CüVÿÀøs8èïrfÿ*GüN†XY99:Cžp„ ÀîhÕùÐhý„8 œ®ó!n¸Äòð»u,«€\3ü“Ê wF£øPp‡_¿Ê\Y•wrt„!Ð(¢_ý)À‘0«ë¹{ÿ¼\{„“;ó—e G@­Ñ€º:õpW˜ŠÂŸ˜kÑ?> Ì/{( €¹`V¶À_<öt†ý‚~¹¯9xcœœÖ×4`ÞpkØõqƒÐHW˜7æ?ÿ¶ˆ@ n…XÂlࢪ_»aÖØ×÷„{Lù¯åðÿúüýôäZaP'„ƒç?ðßW ”3–×5ãþ“òßA99'†WÀ+ $ DDÞÿ®ó÷þbÿÛ« ÿÙÝTTAX;Äþ q=½¿ˆ¸ý© Ž?׆ðï4®õ pü#3~a~«ë/Ðÿy ~§üÿ´ÿ«Êÿ*ÿÿîìêàð;Îñàÿ‰Cáž"\`ÈÿÂþXc9'èÇTÐë½EØ8ü=<8 ÷€Aµáh+Û?¤ó×쯗ÎŽ€i;¡à¿^3^?ÿÅ®7ÍÊþúU‚º¾¡ß!Øõ"ýûHE„•ô×Æ ‹ H$Ä“ˆÿZV èz5¡0ߊùNè뀳+Ú`í„$úu‚ :ÄÑ ùåþÃÃ:ÛÂÿ±®mÔ?¶€ˆ´uúmÿ«+W$òz9KäºÏ¿ìßoÌfE49îd%áoWåß|R!KçλÜ7ÿõ}H”qW°0šíKÆV?Y鳋œ´”¦/Q{¢0pxÜ‹ŽkûdÀì-õÅ\‚r:K±kIn—7óãé~› ŒêtÏfP41déÎN´™…)®){¬áøy€h»àjÙJÎÅ{Ôª‰ØÀwt›).›ö~'Ô˜‰å+º×Øìv9Ò4Éõ±ŠþàN(<âX®9ŒöÔsÉ4âuïìì‡Ôî®DÁ½,µŽr‘Ò@ ZÊgªLi}1=´TÑ©ªoQXlrÂGòE†ŸitF¹Â™<$òØ\o°D´¡µÎµæøÜmI!•IÄg-XA`Cx B½œã̺ªüU|¯§qD•û×Ú´”´*3Z«ºÅ3ïaôJǃ%"hªÀ%Oc€µ×®òGÙVi¡mr…žÙða_»*£i pZJ‰úž!D;12ö!Eœð>È“mþn ‰è¿™\,]ƒbÉ×:¶»€— 1 åI‡Ú«5Õö¥nüäL¤¤.Ù\²RŽª!½«Z¨ì`A6›&ü0CLàîlHíAjT²õoÃ]|á9x’4ŠðŽqêgÞØ/½ƒ3J+BÝM“áxñc¯ço„\I)üSÐéø_í)~ hM¾FB‘€;‰¤(ÃÏLVˇ¿µŽãÒÀ ž ®<½`÷á88k·Œò¤ôMÂFڳȌ8.ªšŠ¥šÁû?ËR4¤S?úºÊ-õÿ0q9ÈIÔ€/ÎfüÈp[¶+GˆüQQP˜áÓ·©7±*+‰H0c7|¥1ËŸ´Ç|Zr@øYY¡¬¡@²éˆ÷Ÿ‚+¤KÊúžÄ½I"‰47µ{j÷“JÏ^¸Ùíé€n SV‰^1ax:›(…—nÎd§-Âù}ke´°¾rcÍNÏ#§·XŸä\„Ë‹-ñÝcX®Ò„Ýœ ?v¤—M¡ùß­kL¹ˆŽRoê2Kèiºÿðã}Q§n,ðîûÙ\‡f6¿`ê=”R9Ý3ï7¥ä?{¨yÜ“öÖƒ§%/3½HNŒ¼€l÷,«€Ë±¦Øƒ€ÒϦYùíþüwrÂ%áS³b8A«1ªÏ’O$RX>-UÈ«°6™¶¸o°æ׬ÀÈöFÛ¢biÓ<?Úok<)ÂiàÙ¼ý®Š/›|þÊUp色û˜2ØöOolÏÎMþÐ`º×íjðù*·¹ }Ò¬mÜÒ$Wth-+-—tónÆ#P  ƒ3yÚÄË/•<8b}våfÀÔX"â¬Ã²Az+¬µCÑ«:|GzSeÊeq©ïõû#-uy“‘ö¡ÙÀÎÓxEÓ™|ü7ȘÍŸŠdÎ ¸ÔMù’Rô÷ú‰r&´xc¶sÄ Çu1‰T³¶[¢à·NÏg'È®^ÞB ZórÙÌG·î\näŸoO 5¸Œ“¦ã{}SÇ%ehäÿÜËKØ9¼ï;I³×ÌMEÆ€›4Sê[Ž·9ñÒ‘™‘Úë£ó:>MÖëK¿e‡°,ÈDÂÐJéÌá—s”¾…õ®üüØZžœ”»Ñ[|œŸ-§«M |nµæõXøvâ†sënzëˆñ.涿ÀÚ^ ½Ëõ5é¾Î¢zA—L±™kõXÁßÛÕ²Z-c½ó–û<ø^DZ½UªºªíÇ~µJäؽnŠ}(ÐG2=ñ’®£éžššä|Ö(®OÊ(²Èä7IzçÍnÜÐú{ÍT1¶ÑË{íAV÷µÞ0/|WÁš£€Å&Ç õ4¬´|5µp–ÈvGT…¹¶Û<9sNJ&´B%²¸œ/éÝ­ž[‘ÀÙ‘|´MƒÞ·+ø:TJ¦™3(!º•"äžúÀ³"ª²MÂ]ù@—Æ¥ërE]¹~y¯-DIãÂÚÉ"ëë‰íq?&Òf.É’áSMåY<^ÕéYí—øŒ<—™˜EDð4™en=Mq™ˆŸt•{ØN0á †z +N|/ ÔŬ­z?;¬e#y“ÖûÀÛĽ¼Zô};U$0]lyy•}´ F ˆ˜vã kQ£”ÐРó²¼ØôD&©?†mʩů>ó¼EG3ôrÊ ¡”§ÑEaP¥¢ &¶‰Ä[»BXS‚mÞq¹¦MBÆÞבMèû]° X¡ä@š€ üoìÜÊUØq¹ýtŠD¾ŽÕpr Õ®5ët§ À‰±Ãëiz1FÌž3,Õ[2x’µÔzF˜¢8@qU^È=럽Ÿh„ëO…7ƒãñE8ºR&Ï8Ï?DãÇÁ“ҨœXr°ß‚‹…y¯Â‰|’–ÉüÑD’Œ©¹:šôÝœ×VÎfÑ]x½ (‰øý€…ÊÝEË;yŠ¾¶9¢”¶™ù^1- äÇs ﺢåŸ**“§4—’#/šš³ñÖi3 KR1W§7ãú6z|ÿ¨LAy3@*Ž +šº…ñ#¿CÔ•\KgCDt,Júìt›ÐÊçÄ)ýh|%G!Æ#å‚ϱ œÄɤ ^Hgºc‹;`ÉÅ]¦€ãÃI@Æò˜ÀNœtÄ©ÖYÄ(Óæ‡)ÒM¿fjlóÈlçîVºkû( öt';ò(Å%Ô!Þº"Ä(è¹<8 U­¼AËÎ#»k¼iSœßúZé>l[jZ›ÿÁE¹IHXLoŸ­>k€Ÿ¯2{:ŒP!Ÿvv0Ò»õÜî¾Åe'«{9.Cÿ:ßÿÓ;…Ûà ë>þRæÀL$£ûÔ³eÿÝò/[rs×O¦Ý%âKñÌÖúØÓJ ƒ€¦T²¨šÊ€vÕ²™1bÉËT†ûBú›,~K"u²}.׿;l™efË+piºc®¦ÚU)O±ß‰»ää›Fñ”å˜#ñ…±Ûu\?(p¾Šœ%jzº×¼°½ Ü#5¨!Z" ×lyصYóÁX¨æ¼±¹ÎâPzu6w·ô oÍ‹Züü'Ý}Q±?» —ÑxÇÏÕ=3Z«b´Þ H)ß‘ÁªJÓu¬®ãîx”¡Ú½ëáÁ"‡[©bÐÊ"ûÌ°œ>³PEµ(¸”3bÛi’¤ö$@­Ã]¢9Û® ¦*Ïk³ÜΫÏ[7Ç©€Ù ÉÂ3-ãÊA}e½ö4û62àÂ1ãz9&Ú]OæÃÖ‡µm%Ö ¾H—œÖ£ÁrBkÀòC¡Û%}dúÞÛ?÷ƒÌTêgwßí¡':ú;§MÚAÄÜ¢§ø^Þq•[§Ïú^¤[¬+¬M¯0f|#u¾¬óc¬Ïw÷®ø¢È:ùÆâÖÝêÀy!ó† CÚ„ñ Áy‚¦Ò±^mË\ùêé¯ NÐ.ìÄ<@ÿþUÿm­srßÀ&ˆdá¾uá)¡Þ ¾rŠã0°ºKÈR Íq0Œ³°úEÕ”¥ÝjzŠ€x=/+~%wkÝó\!ڟІ­ój€ÊÒq'dÑÕüãƬ©s5×cüÁ-Ò¸eꓶxÚGéW”u´c1køG túu¡/$ÈF‚ Š ¥* ´ fQCŽ¼À)¾ØñllãiVhÿ«¼×ÖÊ“j®EûsquH†ýÖGó•X¥)Ÿ[xÊ^7Ý%ÎO…¢ˆFçíf©ª¾Þer·Ê—á,µçÁsåóÛ f¦\kžîËÁaáàXÔN%)£©wnÕý9þî8E„ÁgŽ9{ä§Ø†€6‚ýuG&?¤n©iD7–ñ&˜ŸòÑžŠœ }ÓA]ˆŽ­s#ºRùm|_ë­> »Á“ã¼àuøê!…ÕqÓú,ÝO·*#©õ#¦7''óÔ›o«ïô{ñö†û<¡_[Eh½J}DÏ!è×"€¼õm[x/ùΨTYîÂòö ¦ä #ˆn³Éÿ‡Þ>÷|1dڱɸ‰@ùÎÖP¾.,v³wTÕëÊà’¼\ð¥îþG¼ÍÙGe6ZZbf“÷6PÈöš­i¼’GWËwÕþà)ºíÙ÷‡´™¶:zcG§MòFÏìQ)S8ž£¾eùOïÚØ„Œ‡¶Ø¿ú<堦·EH ÞJQ r¼|^ýÉ}Ùšh9ª$…x~ e¨8Ìs€'¿°Göý ­n:ûÕÀ6…P}ÉeL}ÿ‰Ç·­þ1±Äêk”ƒ§[ÉD0î“•Ö•Áéøó5±}Ì=î40M¨U@J0jˆCÅ\ùö ¥1´äéF\`ÀéãÌ€Ÿ„]IÚâˆ0Ó5?ù„.òkV”(§yíø¬Xmk€+ñ|²ôÁú+Ò“ú!—ÓWÆTéÔ'”2Ž/·õ]ÂØqÚ#4ÓY(×âÃjÙNÀç}êMߤf%Ï]áñ°0ä>Œ{îýIóýq¶LÙW:ÞKÔH Ä}íC!³×…0‡<¤ð‚†õ¹dÈŠM5žÝ„fŒÕEâ¾f?Bû3øružèàæ¸{Ò¹ÅÒJ>Ë ·g6©ã´[‡—y5Š´žfëJ¿Üˆ²óÛík²Ê_šÙ:Å¥Û0WÂó9L”êç‹b´Zð²I°@âeu!_#yß#ïÓ#uÏÁœÞðŒŒf—bdžPÐg½í3ôûs,Š…­xþ UÃBž®|ZZ\ÑËÑsDÎ-gÁµWÒ JxÁÍH«XC IiËž1WK¬hÜIM~‹ìÊTÞ(¨°}RØ‘$}Zõ¡Âsw8÷ Zúy%Ä…ŒsLë%σ@Á!þŠÓ‚NÇ·Îa-Œ q_ÞÞf»ó5‰Rrï91+? “úˆqMGóm»î5¼ûBf@éÄ^ë€ÇøQú÷ïƧƒp?+T‘-5ñ"3{|ÞÓ‹ÒǸ*Ûô¬Š[ŸçªÉ¦Hm´ª‘— cáÜb?;ã|e]oc e Æðé»6œ)‰7Þå;+á¹r¹F ¡¹†³3K‚«%söÕS“‰¾H÷|Ÿ¡n3yce¼H ÝÝ®mgwöO&'v)£y¸·æ8¿²ï£Î¾"r¿Â>¼È&¢ï‰ëÂKFM)XÐlÔ¼Þ?=wZ?W¦8 ¦WÄš›H–^¯yú 6»’Kù xìÐw!C=ùñ”&·ËÝ}r¢(çÌE´I‰ñÙsó_²Üz]|ÙÖ£›Ŷ}ê¹ÂYlÒ(üBõuÂÈÊQ;ë§æWБÊ̤Çh²_–°¸?Œ¦Á’ñå÷6Þþí¸­½ûùü}ª¶*øß‚Ì?u+<äí¸ÇÒšO¾Ÿ–!Ú;&¦G÷-±,¡+;GB¢-ÓAþ†La‰TŸ ‹â„x0*ìwÎJ’l§Óc½‹ìC“ˆ¾ýêb‹V»SÛ•|ákÉEÖÆ©"ôDÅÛB’!€*”U$Æ Ï®X„ :âMRêõŽå½…y:ÝW7ÎÃ{pf”µÞSTjà茺r¿r m¤(º®ØñùC£ô“ýï^Í&Ý,Hå[G¤Óú{iÛbmÑ™³›ò}S|xñ»ÊÁ®YØâë&é©pgŒÀ)æS~ñ…-s­M­[þ,ú½Egvª §™ÕlñŽÑ•|gŠ)=ò2[æù=|ÿx= †¶ê[Ïp< v©º{>¶Rʘó0±Ì•VvÕ^Þ4ü2®Ôïö’ã‰pÇn£Ô7–íÛ´¡êLé•â˜2«SSNçàN#ÕnºÖ§·ëãì"?kMvÝÈä¾,jšö£¸ô{Ε ·M%Š¥L@±¶9ʼþáðà " 7eU0Ú2oyÑmUÙ½#\ÒæÛ™õó.3ÕÄóÌ牆’¦É|±(0aÐP´ûöÀ$žàUâ)SQƒžjz,Kp˜,ë8)%rô@ÙmdxÍ@DüÙq4üµµDžÝŠŒä;Æïá³ÝgÊ\LõeÁaI2HÙÌ׶Wǽ˖ï-½Ù“ì{l'OœzZÇ6άÏù61ô©Ù´Qáñq©B‡6¥bµœU{w½Ð’5Ü©IëÄÄ þ)^Ÿ—¸µ '7ÔÖTk¦x°WŠæG+ñè2p¨……”WÍg¿2¸×¢ºÿB;biÌ»K:µìxÜ´ÔÞ#É3p¸€Ø]™‹¿x´ý#UQš*u}Ø,ȳ,s3¨¤€ùV§M³‚œ[¦CSIýú/‹‰>_×­L~±ó}ÚÛZ9ÇA¤Ùw¥–SÒÃö§ßâ/ƦzµmækI#øZ¦ìC(}]_}ŽÅmz¾a[nôBrZ–‚aC#­œ®´ørYõ,ĸýµƒJy»%÷œTMÿn?ÓÝ)K7ÔN1ÿ,P“}àÊK8MMÒø ŠË[5 °$'彪­¿¦¼‚Žô æʆê‡P¯Mƒ›gÖêO·kêÔ @Ö3ƒtœí¡‰…âaùbþÑu˜ {Á‹,b52%ð&ÏEƒªE-µÌk#ÁЙYk™~øð1|ÂT ëñ&ãÚc&¨¹C„löE)McÁn×Út!š5]•œWG¤‘Õ.غc±ÚýÊÝ×÷4+zËÍð:Õ-0QžâxgÓ\ž<³_zOY}" IIÛêôÐR}/­¾ÅÎr¥)µpüÔ¸ƒH‹KëxßÚÈ-E@ÜmVš*²{þfÒ ËÝÓÛ’ä.g€ÆFVJüxz¾B/-ЙÆ–û1žSÝg3Éó¤ë\¤ W5_tiƒˆRº{‘uö{!Î8é5Bo×ÂM‰O,òrûG)¶0\4ï¨k3ùªÃ´Ü½a7é÷u $ZN—YYñY ŸP9á‘kÂÛgÅ<ᬎªÙžBSNvžÄ4;ÆO¾(Ü÷6Ž?Ž/8éž%èõ^Ò÷õý!rû‹’àÅnbŽ«ï»(ñìÃë„ö_Ý\n™I1WÊ;˜´ÅľQ“q'»¿ÒiF}ù.b«5I Úîyh!1 £ÖïÔo/W©*m›¡>ßÁ^+1ãûln‰7Ö\Ï¡¼_k*†?‘Ïr*ŽºÞ=´kLJ>Zˆè¥èç;ב[g¢ÿ”ó˜û\ÊùÌõI›[5:ÁW2¢ÚGp >g"“Ëy¤ÿ=mœ)ˆw(Ûä ãÓ[÷þm¶”Ëþ›-K‰ZO!@/²ã°Ào9”1x>ÁFñ{u-¶z ®®×)Yš¸cE¦ùßtÚ<ê¾7ÍuЕtŠE&gJ*éu(¥æeF«»péZV— åÌqèoËÔò]œIá.â²Ä_7ãzÖð)¾ÕŒƒZó‹À–³ÃÉlNE‰Ú –¦À8tÅ iŽ;ô„pyÉè%‹eñ {ì™ËÕkm+½VÝÝ'ïV•óæ)”ÀÚË?e pHù‡ªÅ _=)ÉtS¨ýñül †å0 ý놩tyjLDùE¶~­Ë¸¾Rýƒ*6©ÚÞ U!u¹¤Žh{®Ž½æ(¶ŽQ«c—¸Œq­qÒ©#Ö—;Þ]ãÀ'Ÿðw™¯BH™&&ú„¾cc­K¤ƒ–Ò *ÖßJ7²u³cÌ|³ çÖ»Ò«KÅÆd×y÷Øð%¡y›÷è`PH´Ñ¦]‚D20Ãu:ÚA}¡¶µ\CJÁ%fZ—‚o´&=Akk›Ú<ÆU¥„Fù`OÇ*W­`yØtþ¨½{ï¹Ü\(S¡s(>gjÍYäºÊq“Þ"YÚþÅ¡ÐÎn!3¦#S†¨¢ÑÓ° xvV»óÆQè‚EÓýs5c*U]?eX10-bÿXÒ¥r!´`fÞhlD-˜moÝŠñIXâÁFËßöª6¯¶Tÿ¤6§qï°(¥BÅ‘?º,Þlh(»Ì>–±Ly·€ â¯@\¬ýð,0ôñ Z_ó­P¸‚îf‚ʈçÕŒŒB*‰Æ\gëT¥v’£×"Ã"tö\ç*&Ë‹m¢&¹&QpxÒ%›`ÚÛÐ¥þØF(! endstream endobj 1161 0 obj << /Type /FontDescriptor /FontName /BYCRVB+EURB10 /Flags 4 /FontBBox [-30 -247 1178 723] /Ascent 702 /CapHeight 702 /Descent -189 /ItalicAngle 0 /StemV 114 /XHeight 486 /CharSet (/Lambda/phi/psi/rho) /FontFile 1160 0 R >> endobj 1162 0 obj << /Length1 1424 /Length2 6811 /Length3 0 /Length 7776 /Filter /FlateDecode >> stream xÚuT”k×6Ý) R3t*1tw ÃCÌÀ0äRÒ¥„„"%H(ÝÒRÒ%-"‚„€êyÏùÎûÿkýÿzÖšçÙyïkïkßÃɦkÀ/o‡´…)#h~ €l¤¯ ‘prÂÑ.°¿Ô$œÆ0”‰ú_Š(}­S‚ ¯ý´€º§ ( ŠIÅ¥„„ !!Éÿ8"QR%ˆÜ %PG"`$œŠH7_ÜÁ-øûÀ å%%Åù~‡ä]a(8‚hAÐŽ0×ë¡€ ‡¡}ÿ•‚[Æv“ôöö€¸z Qr<|o8Ú 󀡼`v€_€ÚWØd$œCG¸Ç½Òí AÁ× 8†ð¸ŽðDØÁP€ëÃjš7â³æ>À_½€§û+úW"8âw0 EººA¾p„Àîè(k  }Ð|Âî—#ÄÅyñ‚À] ¶×¿+‡”åõk€Áó€¢ành¸Ë/ˆ‚¿Ò\wŒ°SDººÂh’_õ)ÁQ0èuÛ}ÿLÖôF`þìá;û_ ì<ÝpwO˜šÒ_.×*’t04@THøz"@Ìó: þJoèëûmü­¾F€qCºì¯AÀàö°ë Æâ Qž°Ìÿ6ü["vp(` s€#HþÉ~­†Ùÿ‘¯‡‚û,„®¹ýzþþzxM/;$ÂÅ÷÷ßóTPT707¼÷ñß6¤ÃðƒDįé ˆƒ„ÿÎò7þÿ`ÿ­Õ…ÀÿªM蟄j{$@ò„ëÞý†×_¬àþkcxÿ>AyMe€ûæ[ ‰ A¯€ÿßüÿò£ý¯,ÿ/æÿwAÊž..¿ÍÜ¿íÿ‡â wñýËÁÓ†úoØŸýU@ºØý·M ¹Þy„ƒËß­ƒ{(Ã}`vºp4ÔñmþÓùëus#`ºHø¯ûÀú/ÛõŽA¯ïëùü6Á®WèßG‚P¤Ý¯]‰Š (Ä—DèšR QQx½”v0Ÿßl è뀛':`D‘üš¢0@Pâjkù¥þ£º9Âÿ‘AײÇ?2H ˆrDþ–ÿU Ô…º^Ëß¹®ó?òï;óAI>N!¡ÒaNUaM§òLÞüŸ†–VZ¢ͺ#EÑ\Ó9GM •qw»2†÷iº3ÅǦü™x÷O‡},Û³‚SU³ÑXàî5…þÜþ³oí°ÛvÈ[#âiQkÔ_’ôX(m,ð,î>5™ºïÞ|³‘Ùâ±i.¹ p|§bKeBÇÉ¿hÙʼn1cãXA˜YRÅß@Y<ó4T3¹Ä‰ÇŸ(4Å2žù®Yħ,,ôeõt§ æit–‹•=¦a¤õSg{ñþI/#]R–z­—‚èwÅ—q½ð Þ86é.O\Žø¶´Îň΢€·#D«òÙV¬ex,ƒ°R‚ëüŽºâÏ”_³ø*ï•ê™/ª,¡5«?ÆÐìk$vY +¾†p{ÿÕ~ù69‘ýJ½ qcœªLçl”_d–:¡oA‰!âØiO “}™RĬ@_®%úVrñE!K…xX–hpëûeò‹a"4ΫNÃŽû ZA‚¼$>æÜßü¾Iø2Jê1\VÚ²k0ÚL¬³D ?ó|ºedZ¿ì%{ã… #¡ðM ±–x…ëAn†¾÷(_.óã]¶·|§Z´Ð®ÐÁí-5©D>y°Á<«1^tñÚ´ £”5dԋ檹„Ó¤ísçÈVÿ\å½®á*”Üfø©°uCÙ*žÍŠ‘èy¶ÁN 9FP¸‰*mÕ´Õ¹~$ã7 zJêâvä“ÚÐ×9 [ ü˜o«ñHF’#…$„PÙ)1Äöh6[‰»³H†¤`ÙÎXË©¬PP};‘i B`õ‹;ëN¥ãÚ»ØßR©Íßnùµ×Ìs6qz{÷Æù<‹‚Ëc¡…+/£žAGkÚ•ÌÝ7_–Ãľ¥½óêö,¾Jœh FÖEõÉË° NýÁ¥D°D¿ñY9Þã ›öl‘Œzš² C©¿^­í¼aùÃúpá긳S÷IÒåü«§DoÔÈv’\d𥾆žø`¿ƒ´ôÅö2zÄš¦D‘ÎÓ S8vÜï¤{¿Ãµ=á˶“5ØÊ4¤ qH±ZnTnÇöBÊ!¨“pòc.G‹ç Í)ºµ4zà$õnv¾‘웃®ÉS܃ã2ãn)‹Ç|äAO»¼vwø©oG“™÷R÷ýü¡•_ZÜu#Åè½6vyãA™Üb]‚m1š@øª½:Jhþ]b®Ù²Z{Õb½÷„àÈ<^õÇ~ÿ 3[bpìí¯˜ >9Áú^.çÏ!²ÊO_Ä,ŽiëáÈí‰[fmkàœøÈmâÔA{]¯K( õ‡–Z÷X=·ƒ›Ëo•T+dã…éI+ËÅW®¶ÐÌúgëoz–Œwúg¶Tí…ÞŒéÖ`²¦«oS¹i¢»‡ÓÔW(-[üÕODùnÜOB½Û±67fˆW$¿R¿V-wh5êU¥ÝgùL ';áÁŽ•= çlŸX:ØmYp¼Ÿà^Þ/Hq$ëÜ—þZ¶8¬¶¦ø`¼ÿUdS׺çÙÚÁ H|gå=ÿX‰+–­#t6¯° (ܵf"AOc³aš1þ¼@ì^jˆuÔÏéy‡eøáÆ'²"Ýt¦þ¸¶ jAªoAbDl ßP¬wSuQ?ÆB~^à£'k:Æ1F’í”ÑOg,š__õ-Lö®GçúÓÛ Ý>h*?ø~ ЖÓlr~!îNµ’MÙƒ6v«¢›8RpÎîcv'iÌòS>-ÏÞÛ¸r äXµÙß’uöÛ4Å#ýÚ„Á&ýhèá^Ì9Ä,§þ¨è­æö”W¹r¥¨•5—½™õ2ë>íâUIñí’ÀV9 ^–Qf2Ñn þX¾O¯Î oÝx¿dóŠç?ÿ']ÞŠr½Â”‰²<ä?Ý]ÀÎè¯Þñ}ä…ÏtüQžMA¤b’óÓ#Ü—ƒÍsþöÆ¥>DM!ÒÞU9n7Há~¥£Ónô¢ÑÓ}øåÂQ£g%©<öb·µZYÜf؇jZÕJº–>Í-,¾·Ðð×}ˆ…»ÿ”¨ëžÿº&¾™ÖÚgýðìáYÝ4Ë$ã Ö‰ ÓÝn5RœêÒ¸7VÒŠFI3 ßaÆ5»(÷M#P¾%Û/ž:ïð=v’Ã-.°Zaù¾„°•z&éyÖ5±iAÁ¼K܇Û»r±jgŸÉÕŸ§Ã&1WvKçrHœ–mi_i6!,_pJÞ—à4_z¿¨ËˆÞ˜û1k Ÿ9{5™ˆÓ‹ëdn•ªØ.%ߎ|'ˆ+ˆ ¢¸ò $šu6ŒÂë¸u=Q/Ž’3ÊI«¾õ¥>ø¦g´ê¹Ýùžf³ï ó¥Øù=<ˆDY£¨0³ßùAºœ¨X² ': esÚF—¤´èˆM{?j@Õ´ÝG+ø„ƒ|ûåé„(‡78ÜYÿ•Ý5^ðCØ5µ0CE+ˆÏ Kë Îvó«®„E× ˜Åwбäá3_p<>2Aç¾T` ‘˜½Ä¨Z¶¢&'¢éé=4 ÞÍ·k7v•‡Þˆ‡­Eœ ÷-Z(Un½4¿+üEø»qAŒäÕK™È7Ü®ÊE+Ã+QD—ÕñôÎc êŽÒQÑ{n²óÍâ3Š €oÔv³m†9õãY¿^ëŒ:Kû‹ùDS=©a‹¦9½ä-Únß‘ÜÜZ#ØVVHFõñêòP>ëžìÞãæËo—oéâ‚,¾aæ݈œ-•ŒOɃøÆX½“Œ ζ ɉ&d›ËgSêDùãg¯RÏ0H°ò2¶€¢ÀÏ!-QCËGåF#³K¦»`R„Þíõ0Ôð{Ù±æ1b]þÀiÿϾ|¢Aíª‹±¨¹|}}'º]3w²÷g³ë–¦`Å?ÀŠ×Ñ>¼BZIååjµ¾ÕŸ{e^>VšN1ú‘Hš•â[íG—¹{r=34’WRÛ~¯Âš7ÌRKòéÖ•ŠÁ¾£æ×h‚¥Iü éf6’nX»v Ô…›^JCšïeÃÆXœ¦"“mﾑ@ˆ†‚¹8(»íäÔªZ~/Ì@€O„› [9iÐÞ’´ÇGÓͰL{5}n‰:Ÿiÿ½‡ü¹I,º,ë„:Þ<ÎCM×é5‘³Õ•².ßG÷éS3'æO ‹DGE ¬;;ü„JgÝEWµì^·ïöqß¼RÞçÔ{²ÕTà@ ‚^¼Ð¶)<-Úè 2‚€‚¥²Q«Ú÷+°pãöí$x(‚nf@fuåBœ©úò <±jÏHEï”™ðÖ0ánØrjÓ8ê…ðÄ=–Á(1(»ÄDfåV¦Ü8ß„õ¶ We1§ï ¾$åõ¿¥lNp†&Æ]þá0þ ci© uX)!Ĭ†`ð0Èhõ»0Õüt® S:ín²ä%^æ µÜh)ööbÖ£úmÌm²@®l-f2ÁlEõiã•Ê/DÔ‚««FZaCÐ×ü1¤W*#œ·úÀãÁÖc±û ½9Ç°‡ÝC™"ÔSU³dφ‡yFT‰ø€aMkâj\¡Ñ*%Cë!êûV&6@gónË™ì—!OºŠqT,3s©ì|ÈÓÌ¡¸-­ ±"C«þa#d®¼O’ÂïßbÊ)kºá/Äo,™†l/-Õ³¼‚Ìà€oªM ˆ¼*“¢±9{)g[ûd'›'xïë+´g±V¢å»SªâÇ",…šÍI$¼ ;F ‘¶çNt™ ,þÖ(‹Yöj²;ebÙ¤ Šn&Œíy&§š6‹Óñv§›ó2•§ƒZµV Gqq÷Ãüø½J-õm}z™D“8Nc‚0ºùÕÔôIçþÒ>àÞyB"œ¯ûìëñý,?êqïÚmJÍSLDEþÜÚñ@³‘X¢Éº='ØWT°Ÿ÷ <ùÜ 3ÝœS.ykJwª¼Ö¼¤“cÞÊ’­î®É6}Ô Â¢Œ˜óÚ/{¬.›õ¥íJõ©ØIËîùó ªîc€áwÓÇÄF˜D¯”¨óþmGÃØÍ{ä8Zjyâ&]Ǿ y{×›íc—<¦C‰Ï?Ï0·Byay½´Pîûºµ"wP'fO×î¨ßÍg |×ß¹3)híµ£üô©ù*„"¨káµñÚYó±˜ÂÞ¸ˆÜŒ„ßKúº§°Ö­Ó¸s1c¬¡Æ: )–w*÷£Uù‡n_¨Auf•‚îφ%zMEãûóƒÍÓX-åžÞŽÓe‡fvg?È®ØbÐ)æñÊ%0µyuÁŸ%=¨ž]ÈÎϤ¥‹S½—½Ö~G&b`÷À¸[ÀǧE™WfÁrê´m¡ÄžèxµiëµêšL!òãìºó|EàSÉÐ}±Eªô{ŠŒD=¶à ]{-ÚC†0)”u»Ú¤åzÑBK1e÷÷£ ˆÃ½:ºµ€ƒwÂOZä1ƒük¾Vk†ç1(yõK»5{xN9_SëFi¥xÛäzhŽkÞ<½Ç°ßÈ·o×|Y°: {Îg’#ÖT.XVÓ60Öªù/º3‹Ë26μ¸OœééÔPíµÜÉšñ•¹F¨k–¤þ’@1Æåöh=˜:ú0—°²¥FíøN]Г“‡úŒJ/¬\ ¾‹ÍlîÊ›ãÞÌA"ž‚¦(1 ;X»ðýÛ”û?ñ5 rŸÁ<ÔèÑŸCB‘˜ú¢éñã-Â¥W¬áçÎÆ™h–ðª–Z‹ o[c‹éž”¾Ãw‡Æ/îq¸ŸˆJF8ÓSëЇV½9)ÈMˆ dY$µ¾Èœw•·Ï.³}E@™ŒèçZ\a8¬„ètª²ás1\3µßiÅnqFï”4–OKô[H¼*lp¥ ¥ºõ|5®§ñR\•éãR….'ÿ=qÉ‚œU¨«•ƒc ¨F'ÍÓe„xñd*âl4Ÿ.b²³Û­¥K súÎéë“ŠTŽ1™òÅc³¹")¿Ožo2ëL7ñ/Û0T´²ë«¤Z+}˜z½.Íe;á«}`–d—ÙØ"zeîíŠñçƼ¬‰hë#{ Ëî™åñŸÄýªÝ‘æ5')F]Sâë‡7‡ÎÝG³m? _!Ú½Õ alÏȪ̵Ϲ¼˜”lu rô Ûý÷£Ä”7âñ÷s£Y¯+Q’@Ä­O¹5žû鹎HëµRÒ·~ÇÓÑ Q‚ë§X/xº¨Ý°è”Ÿ¦ãç5ža²]Hdóo ? ¢eĈ–ÃvRÃd>lrS7⥣søµ5xG%S>Qïsm-&t*’“VTg߶·©8õ‹Î½¹râ'êôÂñƒ:ï—Ææu˜]ý’eèž‚ù­$^.÷­28Œ½­‰û_éûœíè“ED›{B}x ¸×Gf‡—ù‰Ç;lî<ágã}o6ÂÛö)ø"è`³ÂéÅð¸•ýo“[I#‡@ε²cÜjr™1¾ä”›yÏ{”´Úæç>³6~ƧԮh3Î#oÁ=©™øó^Ó%ý$z’×Ø]KÂë8ÇGØèMl7\ñˆ&³ãB‰øõ’xTÞzw*ÀÆ2¨®No/ Kv=öQÁŠ˜eVtY´™¸zwëÓGw²An‡H(\>¿˜Ïßæh Ð¾Ëq†u”¸ÂÂO|Ê)µ#_V#ÇæâÃÐ_@kùHBË~¼è)Edu ñéHŒþ¦¦Q&ýë‚“.Ë«cÉšBþåõ=m¯D帯¯ƒÌŸ‡&¸d{;¼»ØWcĉM¿ÏÕøNb²“aßß^Rˆy;©ß«Z5< ÿ¢‰ÊíÍ—^¥òè-d§ ­ÇÿT1h¬ˆ î›Ú; ç:X±b¨÷°Ü»³³›“éÃçÁ¾ð[Òeb`ç†Ø(®z¼^}ö·ëâ’í ÿ™¿GœoF¿LÃ" ¨…=~»çë@˜®[ÈÃÇ—'¥·« ¤ì£¿Ïœ%g?MÑHÿ´x²Æ“+òÝö²%ŽÿUùËhñš† XÌÌH ])DØzÓ’è´»zÛ —z¶û}0ó«#Ë¥Ý Ü+s×[Û4)ädûŽÅC'ɶNYÅF*Ùíºùe–‘pP£Ö§ö—LqÏÀ‚P^Dz‚–ØŸâåSÆ6WÏ>åíV욈|¢ÂïTŸBX¸R ½ØÀk8/2Ò8©64BZç¥5L¯7ãsˆ—_è#9h¢’Â^'$sÅTòµ&Å´´¹˜¶¬"Ús–MHcœŠGRÞé½M˜>V÷²ÊÍ#Ù¥3\¼·£Eí2HGÝsÃASâ$U|hlvꀣÚç"=ã$áÉ_–Ûù‚=Ϩ˜ùôã2ñu úêÒ½â¨äT?E)+Gp† !t{úYQ´„ê:µ’¼Ãu*8³D2'¶iÿì¤kó¼8ÀL"/¤ýÃñÌ€ÁYɾôÚ!j}ªNSg ïâ)A¯ðq|èðxÛqå–¿‹‘„ÇË)¡dEÆ ‘Œ›§2÷Rë”fUcÜ8›5É¿v’ £ íHåT®*“Rë¤&vi >·ä D¾¯8ÚL½ p e¿Çdžᷢ`=õ9Hf×à ÖJê1ï²´¨¾´È„¿±5ŠÞº+…Ñzõêçb œ$;i·äáE¦tæ5¼d6äsàSږ´@ükÞÙœ¡ŽºÏ¥wìh¡L›gxÉÔ] §eCa¦)ØΖ-Ï8 †¦¨ƒ;ˆ±é d#ðW8´üîMáõˆË©ê»NNÑ,/S}ÍÐI’Ǻc®[ÖëõˆÂ™Ì,ä’ÖÎüGÔîšMÊWÕÛ3]—¼;>÷2iû±‡Óµq:äCÉæ:ˆl‹T¬¿ñÉ\¾¡°˜q±Ø¤ó‡#žéÏ…=ÿâÇ*4­Ž)²S4¿e:¿yÎËPŽ¾ c[¾“p6¼¤«J§'à_ïÖ¡ºu,5.ß½úÆ$$#AõØƸB¥ðÞÌœr¤n^·‹a¿gIwñEß µ8l{ïNûæNˆ‰e 㣞Uø„¦÷„#øE€JýÑ­qíb{**„•Y}c’êЂ Û~ìŠVn++Bbټ΄ÎCÉ'wþZÿúå8,ËÉEG<Úi€vøm`¹DûôÉéäi@ì§ k ïþ#²ø„vIâ¯YÜq.•“ï“É#˜UzÙÉx2z®n]>l0ê]ÝQ £¹+TJ…·q1@›É¼žkÁÏŸ'÷ò”¢¹¯¥ãÑÑŒPrÖã­I³Ð›d=7AŽl3$ÒòÅ!'|—Èýñ½n65uPyh+î"[Ý4WXx&758óÇ-Òë@á¥Á<±õz=4?Õ¹pab6·nB‚]ÝøpÓü}ñ×þê9-Â~eƒ€”9MÐç)>öt%õdA¿ÏéG/×\È)ÆÕØÓ„Ì2Š_|Ý<Íõ_÷8L]§/ÿUÎÿVC½ ÝpÛ\qˆQŒ«¥Ó12Žm’Š7I!©®jÖì(Q‡<¤K¥ÏþÒ"Ð # rFñ˜F££sñ ž±t[ɺJ¾}Ü1:õ†bn„2•žs ú™W<Òâªê¶Ç£påL­½ r I.§ýa¥E¥*àj ­I 8N–YkýÖ}tÉNÇž¾ÊOÊÔý>Ö\ŠÇÞÙ¾çOË’C ³ŠTzcÍúà> endobj 1164 0 obj << /Length1 2174 /Length2 16322 /Length3 0 /Length 17636 /Filter /FlateDecode >> stream xÚŒ÷P\ÛÖ Cp÷à¡q÷àîîîÚ@#;$@p îîîÜî®Á5Èãœ+9÷ûÿª÷ª«º{L[kŒ5çÚµ)I•ÕEÌíM’ö`FV&^€„†ª+ €……‰…… ‘’Räb ü‘Rèä ²óþ#BÌ hâòf7qy T°d]m¬ìVN^V.^ Ïíxâ&n s€@Ö tF¤³wðtYZ¹ðþû@cF `åáábø; bt™™€ &.V@»·ÍLljöf  ‹çÿ” á·rqqàefvwwg2±sf²w²¤e¸ƒ\¬ª@g “Ððe€¢‰ðßÔ˜)êV ç9Ôì-\ÜMœ€€7ƒ-È v~Kq›o«ÔdäJ@ð¿‚åÿÀø·8V&Öÿ–ûwö_…@à¿“MÌÌìíLÀž °%Àd (IÊ3¹x¸0LÀæšØ:ۿ囸™€lMLßþÞº @RD`òÆðßüœÍœ@.ÎLÎ Û¿82ÿUæMf °¹˜½ìâŒø×þÄAN@³7Ý=™ÿ}¸6`{w°÷lnñ sWf 0ÈÑ(#þï˜7â›%ÐÀÁÂÎÊÃýt=̬˜ÿZ@ÝÓø·“õ/ó_o{€Å  /Èøöƒèílâ¸8¹}½ÿéø_„ÈÊ 0™¹L– 0âŸêof Å¿ðÛù;<z,oíÇ `ùëóßofn¶õüþ÷3KË)©¨KÐÿ›ò¢¢öoFv6#ÛÇ·†eádp±±|ÿ·Îøû¿­Ê& ïîeÀöž‘xSï?DÜþÝ4ÿZÀÿ® hÿÖÏ@ÍŸö×gá`1{ûbýÿ<§üÿëý¿ªü¿¶ÿÿÝ‘¤«­íß~šüÿøMì@¶žÿŽpµ:ýß-à¿ÆXhrµû¿^“·É[ÚþW>³$Èh® r1³úWóüGý·±³ÊöΠ¿.#+ Ëÿñ½Íš™ÍÛeâüvF»€o£ô¿KJ€ÍìÍÿš96N€‰““‰'"Ë[c±qp¼Y߆ÓèñwO˜™Àö.o)W_€…½â_'ÉÉ`ûËô/Ä `ÿ/b0K™ØÙ™ü×ÀÅ`–ùƒ8Ìr'€Yþ¿è­Ž²è|sjþAokjýAo±ÚÿEC*Øsùs½Òa²7WÑÒóþ÷ÇJ)ÕϾ·c±QnˆÚH7Г–€Ž…é( kÙõZÇÎPîÕ$gô½Pj¾ÔÃÇÜbùª]XÕ‰|5ÈøØú ô”)WõG|R÷S¼£3µs±“‰°nKÒR•ÝµÛW(Bè‡´Ó ÌŽi}—©XÓ-OE7 M¿ôªÏdŒ#~dþ4~¦Ò¾©ó–<” ©BÈ{ýE¬"Í…Ù¯2n÷íªê'ñx–µjT"Õ&¦ 0®¥–e; ó—‹?Öä§ø%¾2UÏðî\iiÚ%÷X*Ëyµ§’ïÈ §,éÖ¼L¨±Â½IïŸ^( .")K‘iÚv˜• j[UNÕêj9É —kMÊw´XR¤zŠ“›çgµ¸eáMùYÄþ/.An}ŸÜ̘Ԉ®íòaÍùw ~´ûnY6Ñ h/lÆJ“Äð™xŠ@'u:\3hH”¾*¥ß鼡æ5ã“œ ¯­±Ÿ˜ ‹iéÀ wï?oÌ£c&E°“'®<Ü]( ¯ÀóV§þ~划þE²ñà )›Cc-çAKÁk ª0ªÀìµ ¶vM¢ù䘠&H›ðiÝu[9¯o,NZ’ð‘–ôc‡ÄûTyNFKœy à;gìðÌ¡MÕ^e0†j±ÛÍjàì˜?©8·SŒ›?!NiÕ†‹|¿oý¥úÇ$ò~ý–ω’“üTÂñÙÛŸŽPŸ“t¦"=qê <"‚rBr§Ž·. DNù3±ˆ{¢µGCgk|:»?A>¾‹óÕvòã^"øe9ó‰5 5£îk˜­Ö³å‘ØK¿Ø*3„¯[ž“ø›=™¼ÒŒ¸¼ÚU4î¶&’zâ9øýÜ ÄÑû¯¡=‚íXuܸØýv$ªM¾)Î !#Ë2Yp¯÷eÎþˆý' sËÝ\Ÿ½…‘B÷w®³vöɨXduî‘~Õãä0ÆcÄ%n„O×µþЗ™=ûnÙ¤}¤ÔKSexl½â¡?¥/; 2–õ7X{Åól^u5b£“K2_4  jZ9½Æ«Ü˜˜AæV†ÞùÞÔëI¥X/½—git_Éë^–꿳…_(D†¶‚-[$:Ü”„‹nµ~ñ Boj*íB¢¿ÜÁ>¦áYËD.Ÿ†oøâÑ$47š¨h+1Ój[ÙtŸ<Ä6ÐÙAgƒÄÆÐ2ØÊœÁFIRW‰p1c홽ùBš«ZKa5ç+‘Dò®'}7ýþ¶Ÿ?Hj<%èŽu‚ ÷[à¬FTãfóÄG«Eø놵‡ñfƒ7ŸšþgWr‘¤g*6¸ ÛÜ‚K–’s}Ú¼®)Ø3—¶Óâ6ý$Y1h¼ñªLPV.¤wÄNâ=½Ö¨=ÏÅs.y†Bó.#Ïí²£‹ø·ö֪˲R8‹ó¬ª±ˆyµ¬EÆ¡ƒº>ŠÐáìt^ÜJ®¡|0zhï‰Ê½ b™•/üïsª‹¯Ó?ÐyD{ë~Àþ® l’twÆöøùé‚»w­“ÿ¡|÷É0jy3™êÉ&†ªû"S×®F_;BϱÞ4Ú”1d9Œß@Ϫ«= âªFXE6ʼn]¾ëIs‘š¿3BÀ÷K|ñ&ðXtzæœ '>Y‘¢™ÑRV„hbÑm˜Æ½KPn¿ö©X¯ë’ŒÙÀ"»Š^¨F\n¬êS65›ð‡N®A5: àðf×I-cJ!4nâkw)vKb¿¾v$óBØ–ÛQ{„ ‚V¶gäxÍ«×nέ±¹J‚õjÕëÈ!ëÙös¬Ä e`_ù—¯ôçñ¼J/ÌãÆwÂ_ög²g¢wq pF>¶D7”Ë"¢Š"GŒS”ë˜ü¬‘ÆA‰ýlÉò^F%ÿŠ6ù·›ÿ)'ܬÔ⎰K”£vp¤àæÈWÜJÊšÚ×n‰£{\¾¬ <¶å"ÕrKU~$œ…ðIIñ=èôiµ¶±×Fœ8WÔIøÇßÙWŒmèÔ‘›hªJ ²úÐ89Tõzh$*ODdIÙya€t"yG+ùãÈ0º/eÃÎ ŠÓ‘v fßË.˜˜–‘CzCg.3UºB¨Z “†<ªvÇ[½À)¨]¡Á6b%5ŽZŒ‹‰Ï_8ÃìáèRè™n-­musœ>X¹~bçe+tY¼Xô¤"”L%uç¿u$9C=­ù&ɤ0ƒêWºÉ—&Â~4«rëÇÒQœ¤—MÜ®Æà¤Ê¡8IÿÃä‰yDî€v7U&ørm]侓ˆq‚w@UŽ½;›íU;)eÊÓ=>Kfˆ[ù৊^€åâ£Kj8€Öé.a2oQûã×H3Èî5#êãGBfÓ*Å×j‚Ô[(¹å^öÝŒ‹ËÃZÿr%«)Y‘¡Ô´úvAÃbL·ì™O:B8j;—–™*^sEò³ä€üùÂ=Uœ>Ò 'sÒU$È„ôÏ¡it9°þ];<{9¸jUâæX6 }q¦5šCÁe†.I}­8€±kÒŠ¡NýÒ•2… JÍpœÔ…ø›S¶yõ$Á鈯ÁŠJGxKä&£JêY𚻌ßhšIe›u;=µ×:JH ª½æW„癚üJp9Nà ¦¿Æ‹Ë7p9-JïV:ZW3pÚ/@?°œ¨V¦ŒðT˜u%u{äÌŠ<›¤Œ¶Måæ=阯ÒÒñfÝÛG™‡æå‰(<*›F3‚ bC«] ã È Ãm-©Ff€%É•hI¦æXÓtJ1(µêVø«¶n]<ßh!fñÇ)ÄÒÓ‰ÑéÇiZ¢~¶/)ÿÖŒˆSrÈr13nû™³Å’Ý_çWÅÍQ¡#B3@}Ê*›†D‹LP•0#0 ’pˆŽÞ4qÄÈ ×¯~8¢¦3ú¸HÜé+§“[Mßh>'hh¶¦Á9ˆÉ"¡;èÂ0üŒ…O[º…~(‘ÔWÎds€1ŠAÜ¢/t{ƒÉàa’‰ßïÄ•_Ûr€Ø27Ë'†-àÅn$2AœûÜ<ò*w¡‡¾Í†×‰ø¾çM—ÆÌ÷ÄËݤF¬€ö™~s§:ïÂ'sä¸m†ø—ô M±ábËh^­7` oèCÕª.¹‹H"ªŸþK‡„$5úƒ•1+;ßóXåEkʧ0Éewê Ï»¯^ú.ZˆpqYp&Ø“ÚA²˜±r\ºòâd&ù‰×{2¬%DB³Ö.}Ô#ܼÌ4‘2YR¹ÙKp°£œáCÐø!)(ü_èsëJn÷Y Öþc¡ Ÿ«¿ñtízõÉt’B 69a¥»´DÂ’>Âj8˜òYÆB~ˆb”çݼcÆ©XªwÀ2YH‚GãÝþÎêñ0”p9¼c «ÙxA%!ñ.Û_¨É uîYl9y(•‹«ŽÑxh‘ººRÁmîˆøè½O}•sä$r§6i¿%9Ø|*D/Þc'ÐLh¾l?kOrWlt’`φºÍ¿ú­ á]h-ÑJÚ»! 12]†ˆוzæâêGºsèÈ?ú¬û›7âA>JLû<î¦jïÚœDá¼b6çÒ†HÓØ,ÏßÌ=as– o,O˜Â™9¥´^.Çj°¢~![kç_°6B~ûŽõ—#îQc|¯(Rº)ÀƒkÉ!pªeè11NãáßÐÿº”%ñùCMÏ’ 7.º¯böõœL´øh`B\JXWyZ!I© ÞÈ»5?2_È&Iì®îØE u,¶a;U?BQ¯»É{–¡ãÒ[qC0ßÂÕ2¨´p¼5 ç ª`—^QÊP7ú‹™“SµÌýQvYã8²=c¹ë ™‘Ûæ`5۶ߞ ‹Å­}—þ™èt†’¾$äBÈÊ„{Ñ-~Œ°OgŸ’ìú>YŸ'dÄÁKºiHÂäΘÐáô©uòn£¯u³±¢ˆÃBç|ö0y¨à2:P×,œ¸Ð6›üTÚ?›òa*˜»§Ìzáï:ˆðôâü”T]ŸO‚»RýSH…^ÕÖÊcÕ ÖZûþ ïU—ª\¹QN¾Ý[ÛÙ"‡§±g“®œãó8ÅDÑ]û¹Ië8(êÍ K44RŸ”Fù>ù.— µ`Ü› †À¢Ž'= X¶uÂÁ@Új<Ù»>‡gÓcÔ“©Pz ôS”h@@ºò7ãÈrªJ%ߎ6¬Œ‡¸Ù­íÔãRÞ<ô}Ë®ÿmP! ,æø´bĶ24kó|ƒEu<øéX+Å5S(“tÙÁ¢bÑÃÜ{‹þ›ý,= Hª¶åÃ/ù/j.ŒãÇÎ{#erGûÉ&1m|QZ$¢xYUPŒ¨m]ÁÐ÷Ñ Ø\^Ôµ…Æ=ex€ ðSJ–íl¶S-ŠÅû@Ìcé''¬(œyŒrgQÝ¿³úð…m­¦ 4ç”Ë5 LI”‡å>=ñ±+â”{¬ù4Я6d8ÃæÓOÞ—ŽLqw²r>ë‹èÇJ„Ic5Uý&hÔZ¦îŠ¸Ïbµ±6»žXåŸÖ\¥Éi¸Q§¹ênhëÌ2¸,ùâY,¼â~K²#ÄG}Qs—;øŠ×²œÂ=_{Þì˜X|ñ+Éa’¸fREÏE…\‡ÙõO†1›ÒðQº,¡oÕ ¦hpšX ‹©"Ç<7ìs'æ3ll2Ïž:ñqðŒ.œ£ ªd(ÎKé‚à¾T¬ßeÄZ&;xÏ¡ §²ÌJOÒŒ5¸juÌzù¤«7lkÊ(lW[ЭçröâÀ›íW-ò'g¨rL¸1ºWÑí* -| }¸Jb ÐÌ+B…íäÈŠZåàg«½ƒyïH“7|´B¢ ‚Ì!´øÔŽÈÝÉ8=Ô1-|Oã{’Ž÷¶÷^=VI!9#P¹²6ÀøÌa64×Ø+Òû…:=#Ô9ì ‡>-q —ÿ¿Ë|bÚbÏøº­Œ4qΪsfŸÀ™3­f|k÷ös«ìdY)]Ô|MW©¼^js”3™åðÂ'; ·1îÕÑ!¬À`u]ᯆ_Æ¡Šs@\ÍålM‰*òåõ\uÝ(ÍïÍ…÷iU z^­„–ôì‚„ÈO<ò“Y Ý÷„ì½ê/Ó›Wàe¶\E¥Ñ¨O ëÏ£q4µÊp"Iµ9giÚß®˜?û:ß{Nd¦Ò,ûò°†v°tï8/¦„öÃ޹„í]ãî×8åK¡x†êàŠ]ÊBäÐúS­wÛ)„QBoçŒwÁ¥éC‹tõ|¿är«ÈÉ—/Æ,”me¹¹Å¸IWýôo÷ÊÉ÷%ºÁMy‹Kªúi{W­Î€ò«’‹ÿ篭»Ûx¶í"6p[¹˜‚#sÙA›$¾M¡FSÍuÛ¯0¥WZØTû%´0ä“ó~5i‘s¿ésq)FbØö@F©¥¢™^nÐgÆDb‡Cfô~þÓ‘m 0t_F«vYZÛƒïrçõ'†_`íI–¯¢¯ìÚ4K ÕõÚ¯`óçLϨœ¢—:›'ès˜]ŒÞ^^¨ýŠz•i—®¾g9ÑÍ`l~Ô r˜§ÈA"Òð&fŒl*™ #¼dG8.å¤æ!š+û8n‡™§Àä®mN²@ÍôµQl1_ùüPP}lf4Ž$ܨF.åãÜSxôÈî³ütéw§p-Ó«µ½¤±o_|…õ·$Hµ•@;qoné™ ¤÷‹üìpm¾T¯m‹ýU\Û•;/K ©õƒ)›Â;`%Ênp~þ~¤º¬íÁÝ*r,VÃLÔWÅÉ…ÍñL–˜ÙDóOï…JíýBh=õ¤ÎÊw¦_ìTÝ´xm îÓ?íʒ굡¥®¶/Pu&„\%øòÃ&ý!ý¥? 8ÅpÕÿóžD$N2sÁ¦ÙTÿfû ÎMfõnoä]3T® ææU&¤xíìÑÕo›{ç¤<+Á¦4ã Øl“¨RåšOÖ”ÊUÁ-Ã#‚hˆRZ¶‰òþÓBïR3CvÏŠn¡WcK©z­\o ¨:>ˆ@S GDÍR>Fó˜é+#“IÑ­!l•=<ß4ožÀŒt÷`'þ@‡ð ‚&ŠÌN¦'‹ßñ#lÁ^jùR¦¹%žw³„ßÅ*µ”}'í¾¥ƒÖ~¨^7ñ£|Ò9†±ÏA-¶¼G¬¾.Ç‘1Š‡Çhk :ãí%V Ò&ä©Ô¦6Ç´G¿¡NnüdL¶ðG´‰Í°aIÜ¥"úbX·z€¯ê5-`v˜D{oi-²Oíb*ªlÎ_½]hø#ša†­/×íç®Ý’,il-U-³ZÝ»áÞw ã<¾ÚÄz˪ì›"lôŠ_ÌÉ¶Ü $9$nâôΘÐr'gtÄ ‰·{÷I¬ðp"ó}H0Ïåö9~D0qÐÈm>âÌÒáÿøÍó§p ÎLfôwÑŸ¿ôtHZÚn*ˆŸ0"Ãb/&bI ‹}JoT5œ8Ì)0¬£îÊÝ«€Ùm•Q§4”LÆL·—(.ÝñˆÆxßC³n¾‚Yw­h`¯ée jæ†9øÁn{M©æÝËb0ž{¯»’+âÐ=ª „‡Yxfê«fþlO…t¯#w ÛHÁûEÞäååüçŸé×y_˜üÕÒ`é‡q…©Ð:Âcõû§ º3MµÖsIó× qLú~±ÄY|m·ýº þ6ƒ(!…Üàw æÄ{QsEŒ±;{¢‚-¿{€½5ž©ÕHõžÛûC¬˜ÆB3_³ŠË áv¦^U¸îz&†õ`Zƒ”ÊG’á=1~]àÔÂ/\ûÝGñ —Õ î™ãÍâ×zÕíMb©/„Fxaø,ãŽÛ¿L‹np‘Ü£§÷ñ¤#psñy5¦Ý-öüœ Å·¸•èö0Q;æãÜ‚GîS|Ö©œÖMo¡š]ÕÍšÏƬ ó“_Š ¶[g"Ó¬¶89sZ6`YnÜã–.gRÍâírjãM¶J¸úD`žæN¡íÔD†•Ç¯£3y+’SÊà‘øsjîNÞÍ? «œú‚ä*,¦.|ù"Žç„-ÃGñË_'ôtòýÆ…§¤Úz7?Ä[›±V)ØuA—»Êº½ý Ó9gÐv’_¦^¹¯E×áNvRòBª­ˆ_ð úBZ¥ÑŽùÁ c‡œ¤]hz©Jæ‹Œ%&V«X.4“V²=a33ʺxÞd¶5w eky:¸“X¨vûæ6°Y($ T_ˆªÉ„ùöþ¼Æ^:‹™§¤ÆZuïÂøÔ'£[c³À®…š’Üüž4˜´4/÷²t”úљҤe_/92(úY¾¹ÅÞ’í½¼wí·™$Ÿ§ÇôxþA ÆuÏ\ƒ(³ïøúO®Ó§ì0žb#¼i°12Æ‹Ê©Qu0NÕàXE÷{Õ¼|sËqYáMK7ôŠ´ÍŸòÔ5TªFŽµ¾˜1æϲ kùƒ+8›Îߪ¸òâÙ÷siçØ)ê‡$£×ŠÎæ;Ø÷)‘©ÖŽglÁÆP~7‡dzàb¬$)òN^ùg)œ™»Î<Á<Ò‰…œŠ%ª~õòÛšVÃÇnšØc'9ÓDTÿaéâd_äˆPÈfŸÝÛgÛYí÷SBÖD-#irTÇî}(e|Œd‚Aæé’·…ëºã”C_¾"Œ„5¼$6`)ÁŸÇóö¿zrL•$ûG®üj92ñüZãù)2j-tÙÝBîBŸ(LfŽšmÞ˜bG×áxŽC^Ù5¨W¶ƒƒGrÈ{‘(ÈwóžGõÉäÉѹf3=®}ä-]ºÊS ­¡)‡œp‘Vo*¬þ$ Ú 5Eîé¨WÌÐó›õ:|&­³oŒ'ãɺےÿnÿª,Ùh:"‰Ø›‹R]s‚Æ wf¨5tضÕ9»ÎÇHêÔÕÂ÷`›Yº|Î lŒ#§¾{}]L”߆Zm/ô\n„_X–Wwf}SÓ„…¼…Nv—¤nâ eÞÌîQh˜VÍ‚%¦Pž”êË>ï¦;e.ÀJj)¼,ÞÄüøaë&¿A…i4÷xšÆÝoVÖ1ÞA3¢FÜF)TÐÀIîª{)Y4ò<Ä©`Àª¡»•7>>熯oy—¢äçŒT”¶Ì´ýhÄÉZ¢¸V~Ùiçg›ª¥ƒ®9ž¿ à•Î"2¤u -ݬô5Œî`€[äÙ˜f&Ä‚ôª –™Ò?(S•#©¥i$|ø‚Zo4…yN +×g— ûáöƒE\Áh601¨ôüؘªlbÁ-JtšŸ•` ÿ>ÛþÅ×ó°w]¦Ú,nµìC){WBCCÚ¯Áöø50ezù󣞧ƒ0É›6¸òÇ^æ„)c²€"žÞt«P2#Oè³âÕu^4ÊÀþ®XÑ9ƒy„YÙl—Ýœû=ÃÀ~%W+¡¤Æ\có¯nâpŒ¤×Ó´5 a‚ã²at±Vµ¢~±$¿)n `ÛœS]È`Oæ.:(ù8˹½XE,É?‘(šËJ£¹—È2ä3¶J74)mßJó]x³£Í§:UjÄ-.Òçj:ÈÜ{¯ €—0w†|ÚþI™5Æ·zî6Ÿ;¸Ayí£²‡À‡ÿÊwÌŸIç©YW?-‚îN¯˜ A\\ôÔ‡æ¸R~úÐàZ±Wis"xã[çG Ý~jøpþ1sßö • Ô;éGr9Ó’6OzŸÑ ga‡§ô›”Ü?J;ú`ùÈL^TƒƒŸ9 ›«TF)v­µFÞq¯þºÀꕇ;á‰üpâo²?¨šì .¶{~¨ܳ½õž»Q2zü ¨7ljz°ð[ ¾YoÉ¥dˆÏê^šþÙòöÊi6žOV~VÈËê¥Ñ@Ë w^½±¹N‘84Ú^Ô‘u@öß^•žIF=4Õ!Ï_‹Ñ£'PŒµrÊx®æ*¨ã> ‘!2ozRR(ì%å)ZdÀéý«ÑõS›s—$.Õ¸Å>"ï»âßpyMIÝR&_®‡ú‚ ‚GO¶š½k&O²Ôek³±x^ë²Y¶ïbÝï Xåj£[tÅNæQÔ¾;°Ð–u"йÉÀ9ªSª­_Ê5ÐÚRy½§UYDëhæ7"¸¡;`Á) ZÏ&Œ•P{jENéiÄ•óöh:µ Öm4 xû;É5>ò¬Ê%x[¬r á‹b/£™¹ÛkL»©wùù!rà¥Ïú½eÖÃmlzR?ë—UêeB-«0eÌAñÏ¢>®äáÏY¿Õ×Ö¢´Ht(òçÕRð.ñ ½Äkz¡˜ÂÆ<õ'(äa;K9Cû¦Î¶Œ_æñüƒØSH’$‡ŸÏZ÷²0c·4˜~Ï™V…ø¶|Ì›C·#ÝêøΫ9O -3[{GÑ¿¾Öµ•XÓ 2ÍI(&û“5xgpió6ïK‘¼ÙŽn¤ºõ>XDÞ©>H¶¸øv{3†–{ˆÝÊ•Z~ 3T®ô2kù›œáEy7ÜçågËnÇWXâO5I¤·úYãa| dð8óJ16½µÂàRQÜ®µè\TiÃüÕ}FøE–ˆ ÷¶ZÊ|J¢:”Sãoж– ›9©§¹[¡£:…Ð …üñB¬íòÍ^/Ç{HàõÛ¦Ò‡º=•ºŒÁàoØŠa4Ûƒ6ÏÂÍ#KÝbðîÓ<€JÜXLÊJj$sk>^/M8ÉôjÝÔ„~?vZï„Áç$£0®ä"ÚE‹ïöö‚0¼ Îà‡b5\©ŒeU…wkutwÜ'"Eèáj×jÂUù@PÀù2Âå‡ÂHVïw¸¢tØi)¨8hqªÚ탿òïuÎvºëíŸ@§~šÛÿ±»ãóÞÓàKŽ„ÿS.VáX‰ïV$†üóѨГóéÏÊ‹“˜Ü°#¤»ßÃE·?ƒ,趜7Ÿ…Æz¨FT2“ÓoNÝË¥¤®[Ì°“àÆL‹ µÖ¾]aVþ잺O/z"?¤¤ÀøJ’PSÒXÀ{¥µë}u¦f;ÃMJÎ{QgýXÝÝ)”üraZ>%Wèƒs¸ÎkßïŒ ¢™ð¨¡0&ìCw«ãSÓ94œÆLcd Ö€£Jñ;?j*ï­'Ãs­ ›ÌŸO``Õ0šŒºfùÆ (Ë–F`ËÒ_“¤9Gv<Œ˜½3 @ë0îO*ëtÎè1èóðÔ=€¸¬ªçue½Y\à Ýá³K©`L똢öÒÁCôÏwÛò° TÙßáb‚¯È‹.üixs_º£6?yê–åmQkœ_>Cm/Eðÿ¶êî­Š„}º"@âè¦;`­Qycáe¯ G ûê«]CjàBºÝM³05š5\ª®YéL<:¬PEèU"ïßÑ|é±þ´hl²îJ4{ožfü}…D×'H´T—=3Ö¢}¥ «³m–^ÔÁqb¡^ÌÑýöJñú£¥Ó}º³gàþ¨è.¡…iõ´÷Þj2žèísÒSY.¿ÿ4|Îý„; ÑqÕR}âN›ývH%Ó@uÔdéÛ8·Õåü˜n$a482³$(K#hcFFad¦è†GQ—¥/õl¼ô«è ÐãÌT%NÇOF suæ*Ìì„ š˜N’nãÔ'hŠ>EùŒˆM•.ËãHº šÎ›6Ú^™YàV8)†ßßÈ5E<¶iV~§dX±öÌ,Þ’WÆÚm¡¯dr”p«1t[ôŠ!¯›¾£Ó À®×Ú‚,]ÀXŒ2¼Ó‡£á4™žø-/wµEL8àóÎ@ú—*AR뮨o'ª¡jE#€¬÷Ê´Œ~–§Á ‚ÙGc·ÿk´½ûrÄÇHIìOe¦ ÁˆŸQ¼‡¦ÑÕKy;¬~HscÞUÍm­ç1‘~þÔ•¶]¥„iTË …®gpü H‰>¾^@õÂê¾LîŠô9÷Ëë ¬txd6b 4«<[Ö Ráy`¾2 Áêi»zoC̲R[¨‘wÜŸÊæÂQ¿¹D9ÿ½ž°çþb”\/+">ë«-LuˆÎ¦6üX ‚XÃD(¨C »‘õóZ— ÿzá÷`c^”ëÔõß¿€g6}‘ýukƒ-òxøÒ÷HÀôê¿„¢ì!Œ“:¤ÈhŸ ÓØû$Vš’’L¶Û®¡D ÷äÏ;š\Ô…Ü£¬Û2Ià%`í¨7c;7qŽãdµ¯£¬lúÈ‚¾÷vîòCºZ$zÏ*lQ=4ßbaø®ñÙ”äMéc Ùúíû„öõà!*ºˆ‚»#J«‡%•Í!Ka'ÃöäðÉjIgq0y„ËØX÷Q0›72Ýd­Ç‡˜tQËÞß‹Eê‚ ÌÖG'õ—Î[€¨óý‚FCçÖG[s«¢l™„“ë º½ÿï&v¤¿ÂÔ…,‰8ë÷ gQŒ7¤Ò½Ô‹ ·²wpÜ»"¢£ÈúäÑ3ýMÍd¿È“ÛØË Q”"2eÖÄ\M¶¿úŒ;–ɵ(Tú‘;îaX¡zßוø‘užZémÆŒ rˆ;OóÌ(Òè?ÏÎÈjP½ë¶nŒ1ýЩ6eÛ¾P>Y.C»µáxŸ0–ï<ëópn—pëNÀltÂ$æpœðÄW‡¹uœpb騷 ü-«[}¿·‚ T²5¯´m¶ßX-oƒ´§R¿c&´^Níþ˜O9M*D}&kîÖKÛDZÅ{XÒ×o—|6ââtÉ0æ"RÞå¾±)_âN[„IÕÉ_¦”îätäJNÐÃ9¢ñNi³9Õ–£22ÉCëµ3¨¬ín™½Khô3+,„,¥{?êCÙåòÙð¹óI$_ÝßS:Û¬t,Ì.ž”¿Ñ¨ÖtÍ èà8î‡ýpÅ9‘ÔR ø(ŠÃsÆðîþ^Ùðƒ™Aƒut–öÐ1m¨.Æ4DiÏD«.\b®û,?†]!TîrÙ¡ €¿W†ƒ¯,AkÕÁ¶¢”$.^ÆW}T‘KïžÝö±–¬OÛ)ËÎh~±ƒQ’ÜJaþsŒøv)‡÷_ìF­0Ëu?V¢ì‰–¤+‹2W,Dø´óì÷ÅP–õ\Ý—,V±]¥³Å_j#–Æùü"cö;¿ aÃr‰àHJ!fªþJB5kÊG¨·€.ÙÕ’ï&‹a²ëÀ°Iì6Ä.ÚF!=²ICdBbOúä‡B¥ÛÙÁ NdFk‡¹êuI¨|±¡vÍðl•êtà¯íŠ³Kc7•":öŽ/WŽì¨¿ ¸ãB×9©ž6tøtí;‰/…å“éÜ!ä%U• ~5Ò@DGá…—o¹Š«lB™ëƒ"HÙîÈöâ-ÚÙaV,U1;»Ap¬ç—ÏPE Ç‘[ZŽ÷ê¡Bð¥Â4¡¯.¥©% ‘éf¾G™£AÄ¥Ñ>Ñ À¿½^%"÷g/¾“Ûþ8´ú†@Sª óÉ+²#þ|aÜLYNuÐuÞ0ŽL泎f ‚«­ ö—W&²Bò¼:¤ FS9'š+0Ï-Øþ²[ éN¼^³þÔ…Pê¦îF-Î^›©²ôYy5T„²äÛ8¦†>efø‚k+bŸW…PPý;Þ˜mÆoæ”!¦À€r»;Œþ¢Ð´ÂeÁ¯ ¶c`©ÈÿÉ$ìƪ¶ÀÄÆ÷q?ƒalöò»Ò§›Éœ6ˆ‚Òîa¯FÎN9b‹ÕÜ/gÚŒK˜´šš‚Ö»n_gIòZ¨‹Nƒá7d©÷3wÒ_nåûº Zµ‰'.bp¿wX—¿HAÅÿ¬­ÓS;Â$mò‚U“믗\£%·4|~é>„š…§ÑgÁq éçïPÛºñvo‘•(d®(~;B¸ ô¥.€úI¤¨™_¹zãÜù(Ê&Ì»šæ4µ7-ñIš-'ª áã’ö»ùÉÍ-¸UO·s˜†èP8‘ƒgôNRpÄ*HB¼áδ Ɉ2S@išÎÐÊÜ@õõ5½Ä‡1ó¼1ûÖ¨~)=¨öx†c¹H!¶ÃGô'h¤v¡Ì§M²N Žÿm>ÙQÆ?C„dû´<µWIk'%Ô‹Ù,‚8_ÆŸyçÙš š«s»ªZˆƒb%õrõ=c-ß{™ê†vÓKèSZ*Ž·B]íûoàÜNE|¤×ÐôôEÀÝ~´Dqªp#I2VÓ…¼ºmäsãδå®…*ôpl_e°®B2"÷.¾Š~~ÁsýBÖ¼ç3¼—MV#R_©—ˆp2€¤`õ:k¿wÕy>‡B†‚’#cÂœíz.|@ð•×°¯ïëÎ{Ö¾•L‘°c¼¦cŸL„96•ªÍ†|‘Ïäë®?ÙO/VåÎËÉÏ·—f£—ÄÊ‹”ôØØœ¨tmaÜ.Гë/§keMH1h"9GÊ‹ýj—X5ÆôXÞuØJù¸½ÆȲ…»¶2JÛŸZzé£BÐC2b[šò–þ ,ËEPÜñ)“°­ó‡·Rñ!¾ë®›C‰C‚ú bàÑqz<.îmIÃ2³Ö-l{Iï^ʼn…&Ágv“MÜR8¨zýz+}«Öõ•ò9÷(Í”u.W<èZóÕïIò¿¨¶Z?ãP©šZÄQ þlÙ9çÇÚlWýƶHDºF” ÷µc[öóÉ` —Í%Ì1PPl¬×Fçë, _‚NgíqÁïªPVýχ7’UnŽ×L±ø¡ÎÚwÃí•4ù%Ó‚ ':[M>¬,‹ 0( k+%›Gæîm]X6 ‚„ü-M§oÐñþ”’¯()è!RèÆ$À÷æ™r$çPC åš‘RG¬B*Öçúî Yh¸1”Ñ‹< ¸6Œ*û(ÕmFH¸fZ¥ÑìH+ŸÂj/Ñ;-)›R¼'ÉŽD^$¨«JË’O…¥€Uaèï¯)6Ý-ÐÐä*ŸgïßåXS•ðLÀg;-K¼`‘$H%N(Ÿ2ú¹±" M|ÒV³°í QÞtð÷=úÌ (ÇvãtãñE†!¸ÔÅ”ê´èsÁ§ L˜nuǶL—ç8Cd´]áÇ*,áZ²Ò¥ò—¹~ïC ûD†ï#N¦É¸8- Ú/á¹.Û˜{>Žãî—›tUŠÙËOfçï‰ÔæBîà”DýºÅäµf5î‰Õ¹5pÞã…&z!N$˜hë?Ât«Ä{?…ìåקN 9~þq«âëiafF:-XÊtlND0ó•nØH/Áb‚…&?a0>óÒJæ'á{%âÍ)ɬc¹!“lŸ›ˆÒ¯Ä›úc£—_Ô~Æ"˜ùNg(Ø¥´ŸîCO÷ q°ªÒ]»ÂõËP¢#3«y8iÝ7”KÁ*Ü—Mc ‚ÛÅwú=+š2Þ,« Íº¾ÑòÁ¥’á½ÐÆÉÑáQHâ4ô-…„ΗȈþÀ±ˆ*èx[ïwŠØ§‰Å²Y§'Ù5»Ã!úÒ~¼–Ó0¯Y–ݤJ@R‹œ¡³Š]wLøôkžiǵ)*S|Ê©|mœ³n]ìP-Ÿß1<óá0]l”¹c.—C}£±²¥ q¸†7¢úFåaÉ|^ ]˜ÑÓ¤„â¤_~#±°VÔ¿ª˜1HxÑÖjûËX–¤V<äÛ3ôPêرܔA ‹Ú›ŠJ=‚aPéçˆn^"í‡ ¯c:+rh¿"ã×J¶÷w(€ýÒéŒw'þ>i±¼èt;t/O–å#UeŒkéKsùÒjDùždèÙæË5vWó3mB¨„ ÅÌ¡dFàŒ´y=…Í£<ë/ÑÊÈ”˜z¡g8fÍAË­C¿¼¶îË¡È×ìD©­¸BÄ Jø(ŸTðP³î‹Ãå$W®AoLШëã¾ëù2E*è ì¢.N ëj·xª×æX¯8Ê}ü‰ÄmÒtÏðäº>91å ƒæ4÷ f¢JV{¡á”S¨þ°qúKX"CPGb¦¬ÊÇÒ0ü ¨£Î#¸l®ñX‹õ•ˆ‰ËsäsÝÚ× öHBk‘ÁðBŒQ¿ØüGH1‘kwÛ|¾JvMŸ«“Ñ0¼ºŽ?Ä]CpEà²V y/ľ«%“‡……þdªfÝç#}8#ó…%ÙƒØ-VŽdZ ç&ŽnbŸ•|G§”ž&ÜHô‹þ¾§ÇÂ!ž§g°f“›‘{ ê䮧}f©YÐÝðõºpÓHW.§qæÌôŽíg4!îY:¨ß¡Ê­êš\VAôËýz9ì»­FvØ_ÊÍÞóMßJe²šž“&B{B…Kûi‘ÚÀXA _'%Ž\ƒÐîµ ÙÊ?«•T6;lòr–gbÊ*9Õt1´—Ýfý=Õ°g×–ÿñë¼áÐj\hÙ™‰ d_À·¼ÅjˆÍÒ¡.†×™—‹· ßBÞ‡|Ÿ?@g‰!¾'¡:âÓ9yy…×Ì?RkžÌqZ!bt>¿F(4ž@kÜ¥gêuलÖhÃÉmÎþ®Þ—¾ÉG´áé°£?ÚÑ5®`—5äcf ¤ÅoÿrY®üC°%KKo?©¥&O¼Z_8÷ˆ¨¨×&Ð>qäÐS;CUpåž6²‰ŸäÁÁ¼¸äï[tûA"Œ`;C Ù&7qæ‹#Öúœýʱj$ßòzâPEW[ j…Ü Ä>A¿Y©Ò;£ Â]Î…Ã2h9Ô³“‹TØ(Êõ,®B~'o­Ì“`.V\)j?l¸~ ÅŽèδt:>û¬Íü±ì÷¢‘þÅ.¶v«7 íù‘²&Þïfñ*Q©¿B"pNOpœ×GpzŒU¶r3ù9X“‰ÞÖ téÒó+¹ó ±e•Ãûôõ7bñîÌ—xìй£]•¾‡?7-&`9Kv÷mäÝAfÜ]‘k†(1î©_Ës˜P’AéÝ ð¼ºýH×ÞUâJð´. Ö´œ°&dzçök膠xç¿àÇãX¬.œ…>çúáÓ¯,g+žD}ÞIjIŸ¤è¹Ü?þiÏí!º—›V˜ÈT­÷ ë&+ª“òÄØÙhAqk–c0ìçG‡¤ÏªM„·ĆթxŒ\Ü>AÙcᶅ­ÚZÈ&^U|kéÅõž˜Àh“(9Ÿæ;+”]A¬ ¸6ü:úº 4tˆ‹Œî](RŠy«Jµ #Ÿa±¸¡øo’Þ0õÏóí凘?yx{ößa+_“ÑaukŸ”ðsI‰~„+Þ/ekYæùL [E$ªfõÅzä´40^–„ÞatáaÕ­è{ñÏûä5Ö´ÉïPF¶døýôÁÌ÷á ~t³‚T³•#)¢îð‡F Y¾¡øø×V:æâT]ß~Ö1$Ï{åeÁ = ò ’Ürá¥üýÖ0 žržçm¼ñå1ÚðGÎ+ŸO\ƒŽc^´ûa³®Îäª05tº[Æ¿P¸Ê4¹GüÐ_G†TüE•í…·#¡ dc>„Õï7Šý€ƒýŠ¬øñ±¯Lÿ$j©èx RpÒ²ºüÃ&,S߸MwqÝ9D|”¹sŠ¥Œ#z”F=3‚(39#o+m=Ý B=qB¯ÊQ¿> CÁ]Í«léF»žöNl@­¶!($¹ˆÅûhÚ×XùÎ_fQ¯’)d¿f¼WôõÞÛr秃KrTõ“VhÀ;‘¬ÌÞÙò¬¥Þ2ªü3]LŸíŒF®vÐ|¶ïé;%›òßÒ¯½ Ìsa©—íTU|$„+¤ÉñÃÛ4'Ér”3¦®õ-º.±ZÉþH»ûÞ‘ÝjÑê"kɲæÊ_;ºK #‹¡À#íQ¼27ÁÚuAªŸ­ãñ¦:ºwŽŒ^è°¨jx>vÛ»²yInî;Äh¡ Bw9Ž9Np,ÖbëU3³x[1£Ñ…&»ŒŠ,ÎõÌì N8ðgoH(«jü¼7\,Lrž”ç(ŠHÜÀ¾$Y¡ªÓQëþ-LèUÐéáLïEvâѹs{ÒÖÌ—j ñƒ‡èc; $ù"þJ1ÚNâEØìy}U;á8*ËËsY—¤†×s®H´ºppñó²z¨Ãf‚t:Éws•x[:ÄO(´ÙæSý¸„w°¿•¿©ñÔ´tþÒB³­üÕA;u÷«l?Ïk‰­Cö©h+¾ AàÐ?ôS”äjÙìË5 Žÿ\šy)vß»â"TÙ¸Žô»žá9i‡˜‚ºT6÷TÙÈ#Ã@tdê™ã¡ÑB9ÇÂ%½ ³v“9µ†Cn†oY}"ÞTcßé‘y¨Ü'#E=֠±/•Q¨Ï[rÕŽKK HCÍ·éìm;ÏGQÓhG©$).'sìî¼àÒ]I‡í¢]7eg¥ƒ·6O¼O9èÚ3ã>‡ÊBƒY qň¡²EŽÃc¼–óåôï Ëôö¡G´nºÕþÖ^ªßX¥›´Upg ŽÁxpV³¾MKòméè½ÝxúÆœO2›np îiãþ ±Kõ äã’‚#ý5O:ŽRÕ£ô¢yi—£²VT£šê©d¾dè÷TŸQ2V,ì-ãÑñoš¸À *Ò>ñù©~ðIU –99¹£R¾æ"3ŸzåÑD2«&l"Á>÷IÂèoÝMéþ‘ŽÐ"7£Ú‡¹«0ÔÌOj¸˜ß‡,$EÇ£*4Ò5÷KP ÈàLZ‡ Ç–—4dw©Eòô‰Ìc ÕåÃÕ×|aÙ’ýÔ<Ég8¸¦) ;©ŠæòuDIߤŠÛ1+§º.3°žòÎüÛÛú|`ŨÎ1ü2'ܧoPÎp Š+ÛˆGþŸW¨ü…3@ì–rèÆ…[ª¬” Œ¼AAºU›„~!‘€CËú_Ühš%a=w0sÕ5wWAúÖ¢´¸-÷Ó‡¾,ÝÀÇ’ÔfúÌûv^G\ÐàÂ>ó p6öX»a ,Ãh· =û ±x×`bêÀñs$|š€¢ÌÖ«cYFmŒš±Ñc*ïCÜÄPîáþ¾—ÄÍ4&˜œéåƒöye¯ò™SÀçÿÀ–ŽD<—ïa#øþsßz½¬MT£žÀ¯x²Ù¬Â>åæh•ÆW¾i«Z¿qõ«Å92«îÏBù§¼SŽÃb8î ehìN#O ÝæWÈý¶k›4ìIÞ(-.–wÞt%|âåA{PþÊ—Ú6\‚%‚ÒYŠgÒ½âP•7ä÷–Âճᇺ³ë oO¯U÷6úB¸«õAú“EOÉQä1šu¼Ýiáh´Æ{jåÛgÃò^}ˆdÙ!“Ö}Y à)ohÔï0ä!¹Ü÷´C+JbrÔ–w„D¬<‹w ¨ã›O¶ˆB,7g†IqÅÀø±e­#ïÙ¯¥™‰]finKf\s_½%º…8 ï¼:L ”ч†K]êê%§*û7ÌRiß±Køš €æLòùUx&¡¢žx*Qç` zœrÁž2 wlg\Û¶ë™[.ˆÅåhI<Îñ£ …eG¸ëÉ/.-ÏtÍÊ••òë±3,FîN¦¢M3оóäµNá&Ð`W ðÊkT`‹eÅœB·övëØ^_½Rð»Ço¦[€-c¤Ô¾UJ‘þZÌs’yÔšnF,Ûó5v¢ U.ÆÊ^ˆ6I |¥´W{œL5&¥=„A(à‡ C#C¸*0ª_7hÖ'Æ¡é»T:šê»qßÊi-›¶x× D>…´·Ž,˜ö*ÇÀ“ ‘`dÍŽ –Sã)ÉšoÇvÝS4ˆHA¯€!èë˜ÀÄí@É<2aÛ‡±@î°wnoÓâ=¿ËV†öšHŽÔE¿hò3·\höÓ„ò½¦š´Ï"¯3!â…FÍ0]fQ·-·—'zNê[åX?Æ¥›ù——Ú<æü©ð#›È~Dv]ö:˜«eâ½ C‡?:ŸzDà endstream endobj 1165 0 obj << /Type /FontDescriptor /FontName /HKOQTE+EURM10 /Flags 4 /FontBBox [-32 -243 1060 720] /Ascent 691 /CapHeight 691 /Descent -189 /ItalicAngle 0 /StemV 78 /XHeight 459 /CharSet (/C/D/Gamma/I/K/L/Phi/V/W/X/Z/a/alpha/b/beta/c/d/delta/e/f/gamma/greater/h/i/j/k/kappa/lambda/lscript/m/mu/n/nu/omega/p/partialdiff/phi/pi/psi/q/r/rho/s/sigma/slash/t/tau/v/x/y/z) /FontFile 1164 0 R >> endobj 1166 0 obj << /Length1 1939 /Length2 13780 /Length3 0 /Length 14985 /Filter /FlateDecode >> stream xÚ·PØÖ-LpwNC€àîîîîÞ84îî,¸»wÜ`‚»ûcfî73÷ûÿª÷Š*èµeí³öÙûPMN¢¨B'd 2Šƒìœé˜è¹bjÊrFFzFFfxrrUKgàÌðäê@G'K÷¿DFÎï6Q#ç÷89@ÚÅÀÄ`bçfâàfd032rýO È‘ jäji £Hƒì€Nðä" {GKs gnÀß”&T&..Ú?ÓB¶@GK#;€œ‘³Ðö½¢‰‘ @db töø/ J^ gg{n777z#['z£9?-ÀÍÒÙ  t:ºMÈÙÿRFOPµ°túË®2sv3rÞ 6–&@;§÷ ;S #à½8@EJ `´û+Xö¯ZÀz`¢gú›î?ÙYÚý™ldb²µ7²ó°´3˜YÚ â²ôÎîδ#;Ó?lœ@ïùF®F–6FÆïžÜ .¤0zøyN&Ž–öÎNôN–6Hdøƒæ½Ëbv¦" [[ ³üçµtš¼·Ýƒá¯›µ¶¹Ùyý˜YÚ™šý!ÂÔÅžAÍÎÒÁ(%úŸwü?6s 3€‘åýF@ÐÝÄ‚ázU{àŸN¦?Ìï |¼ìAö³w@K3àûx/'#W ÀÙÑèãõoÇ#x&&€©¥‰3Àhniÿû»hö~¿|GKw€ãûì1ÿøùû“Þûx™‚ìl<þ ÿó~T¤D…D•iþRü·OXäð¢cÐ1³¾Ï*33€ƒ™àóß,ëÿíZ,ÿs6Æ¥ìÌ@®¿$¼÷îd¸þg*(ÿ³1T€ÿ® ze €òŸÉ×edc4yÿÅôÿ<ÿ¦üÿý,ÿ·Éÿßw±±ùÓMù§ÿÿã6²µ´ñøO€‹ ÐñhÿÚ_9 ©¥‹íÿöJ9½ï„¹Íßͳt·tš*Z:›Xü58ÿÓû÷…³±´*‚œ,ÿxatLŒŒÿË÷¾e&Ö﯈Óû ýé¾/Ñ—³3™þ±mÌlì#GG#xÆ÷¡bfcx1½¯¥)ÐýÏy0ÐÛœßSö.Î>3#ü÷ÈÎ`úÃôb0ÿƒ8 "ÿ Nƒè߈À adkkô·ƒÀ õz'’ý½g*ZXþßê£÷±cÐþ½×ü‡ô}iŒlì-þ±p½sÿãïƒé¿ 3€Áì_ð]œå¿à{]«Á÷RÖÿ¿;mþzuþÂ`°ý2¿³ÛºüëpïÅíþñ¾7Àî/Ë;ÙÍÿ-æ=ßþŸ~²¾##GgK#SK³NÍòÎjÿ¯n1¿‹°ÿ²üAâô/÷» 'KóÝû{ÓœlŒœ,þUù]Ü?²˜ß{èlô/!`  ó¿únrý2¿Szü ¾ŸÁóOø_³hââèøþ2ÿùF¼êÿà?ÿ î@øå O°U]pÇ}¾Ýî8ôÆVWDœÖ@8›3Åb®—…,LºÄŒƒ°¡i%ÎhŠâRYèô‚7>õéý„»nOfÀz²dŽ3˜ØÀ¶ð9]ÞðÃU9È„g>É‘±v¯Dˆb¨©ó9Acá9„£—e¯êwáK—Óž6×ÀP|‹J’Œï-”Ýu¥EB¶å<¢¥‹ƒî¨“ê¢*¥6ùes'Üñ÷Ác['&mdmm(sp …å2_¦¯š½2ÓSš${ôë\¬øLéf'0 a¶[‘²–³¹>ØúohúPcª›šwßë3%b€=—ó5là&zóäÍ=îŸX_É%”ý|nGã£]á5Ùn 'ÌÑ=lK’ÑhVöÒµ]áœkÉN¨¿Ã?ó¤<æ•ÈVíB+¨|Hˆ¯¯@-J½‘pQ}Ä•#q;Á9çìÏ£85M¼¡•zE~* yá¡‹&])™Ý—•bjfõ»*uÚŠ#K®ênEAÅá*®gÚ~±ÏŠ]¯ë&y–|z”#ZämÉpçNèk¼w¼’§:ÞH³ÕC)oÈ£S#DfÀcÀ*ˆ[Ž1²J„µéMB%ˆvqW®àwzêç`ë=ä7–»\Ar4ö3á©ùÖEïš=£Z‘žž-–ÔW†Üy°¿«˜U«ÚŠd#» åYE +>^ørRé´@ÉO?ˆ¹H‰Gðƪé•ûhÔL£¼ñ'rLüŒŽR×|ê¿y]¾ôˆ"“¡ÛÛÊw©½Õï~nÃ_Pûh(¥—0¡:,òѦæW_³®þO,nÌ#"BF(“äÝø9¬/ÑËO¤åÊu»•.Ñ °óãÌvL "]•5Í<7A¸Gʽû…OkDÅ‹»²øúÆ_ÏAÕÁ“ú“TN—h .¹º±~Ê|¨ÍoC±"à–y_U|*­n1+ì§ã¡‚4)ÚüÀ•qÖÑ18›wÞHlLY[+ÐBˆ“5åUÉ®{,}±z¦]È7d3+YvbÏ« Xlög w¯-Ñ¢ý3me¼’£&«JZ3Z®78õÔPí¯o°œXvVó±Õúr%úE:›y]ƒÞ'aèY…ášÆnq,^¦4ÿwBûã‡Ù«3PŠqÝt;HN©Äói÷|;š¾@õ’Õ»¤9¥ Ù{ªT ÄyÝ ~TÍô>^#óË.€ª*“ºfowífVcýÀåôÃ}$|©lP»+¿ie©ˆêñt³¤£™üÔ›ÂôŠÐ½´0Ø•8Roq 3 ûóKsçk„ƒç>¬a¦ÎV›ãâ†AxnÌDóÐú 4Ú€/¬þ™ad?wD#,¸·ƒÛˆÍe²`÷îÕCÉŽÂx®Û©Y꣇X/éUùV‘Ó©R´ÅˆµJÅê|00o)ÿaÉg¢Ž×à†Q®5NèÌHWˆÌؘ¹†Ã»K:ç,.o±\ªóì]{k^KÃø!"·Ä:Òu¸ Š\ BYŸc_‰3ZòµçvÒÍžà#Ú0…´ÖÏ*¾*äÍÊ3…¢8ŸÇHÄß;p‘£Ö4¾ögߨ,ÐbWß¿ ÑÕíQ)_oWdZX”2¬ú8ð»…ú6‘ï‰6LªõW5>­¸,œy<™*³a„+fõ5ÐØ‚G{˜lÇ3°¯É‰¨‘YÍŒåÝÎÚJ™Éõåw 3ôò “j`”ûƒ-ŽýT N†gìú >L"FªáRÙ‚3(Ü@§§GüNîü5W`?ª~éU®ST”’EXdÝEÇ“cŠfP–‰zB@9ÎPÞL ™/a£Ý(«D~:×M8V…HïxJ‰¡Yx„¯¾ŸÜÁ¦í“ë´Ý/¸rµ|Óq±n9½‰¡jð“­3™!t•h ÂUš›ä0'÷ÔÔL¸»¦Æ3m™+QçîK`j‚§Iëë©ñtê4Ëér*mªêYiûâ÷¯L¨›Ã\wü®õ­‘?Êôj ?ÉÖ –ëTÔÀKmõ Ð÷ŽéK=#’E¡ ¬½Þ?žiN0h»UÉŠ¨ÏOÞ¸rÂœ£³¤bA*ìœ)>Êe)0&c5ë롉k·ø7üü¿É‹Gœ…ÒgÖ_$#Þ¡ê ï/lç.)%T5[–„îÀ²ñ_‹&ÚØC¬K¾%©Âóšà8WŸž54#HÓVðkŸS.—+ï÷õzq §Æ×9¼2ŽõŠ-Ž@ Ç|*ŸøÉ/¥ÊjBÌa„ík7‡PÒOáv?&âVr5xh%¾*0c}SüQ)˜Jš:RTõj“ää…£Ìþgô›Ks%ô”ϯ^n, ¬Ý ¢Ó¬‰QáÕ¸n/à§<‘;`m“^('Q'dý¸ÇÚs\¶(&QYÅBûP²x‹íÐ6—TÔ2˜pô_55à —m~wd  .ήöbá·²7iSª±…àÉN&ê/› ­ÈZåàµIoùgÜà2C¬¡Ü±Èèܵ¹•""“Gy}KŒK5Q†RÖýð†zbnôð•5ÖJñmSJÌ ”ÚÛuJ×®Ì/caÄ"I¨ëµÈïêë,gÔ‹±å¡t°`Iæ‡ðAã¢0Š££MQoæì`&p›:a©±a¶ jšÛHã“Pyµ [²ç©„˜ !¶€ý¼âÛÊÏP÷ÖÚ«RJÓo”CrOá÷t#Ü »§Ž‹Ü.WÓÀî5Îêž+,@²¯g°UˆlºóØ5rL¢¦WÐFÂ6rÈŵ!X›2NêKBÚ˜d3ìLJ¡gc6{œ^·„ú‡2 µ>½Â×Xto÷ž“ÚO—‰:SêèþþA Û5ý /ãßè4ÃG¡¾qÔñz1ªUšË¦äŒÎaiñ J.´³î «‡ ¥L.Ð7G2Ìî^ËoÌA\òøXˆD¹ƒV—««´cxúÝwy}<›Á! ÆR”Öå|7Æ°@r\jO¥Oãø¿–±WWnã¢;iˆ4^O©Ø“ÇÜ£ëFŽ}å*   Zž¶ÖÔ$å¯ØZ%üJ6ïž}áËPÛ†vt9ý§òA”¨;øŸ”.bžêá3[¡Ìwt±—Uu&`†¾µ0Sg/Ãa{w+ÿj'Jž-Íïw"; ìK“cv%ZŸà×hž ‹Ïž‹:’æn|@b=¹ å°k®És¡+ ä-º²f¦ÚbZ™´o$¸qà&ÁïÆÓ$àϳŠ­¼î5«Éã]ë–ê‡Y •Åé=뤗úÈý:­”ú›4¨OŸ€¬VwËÁ öá¾@T¶¢:Spocš¨ÌÙáé—R𡈢Ž]ãô2¨9‡‚E‹¢DµgÎTNÉÀ0&3¼ehCÔå– WXØ þ{ ý º ÞÞÊÜ7`Û/¼kz*^=ºjz‹3y³½üÈ37^œæB²^žÙNxc_=“–%„zx@¹)æ:äÕÙ‰5,%æ*E `j©®•°©k(ÎÐzŽÇQã¢)ùÂ+i«gŠÝé”Åqús æØ1íãÄ$méqß³O|=Ç"´î>GR•¦[:›G±EõwNG²vXâpÕPOÜ b™}Hº£·tÖC‰îbà|àH‚tä:“t#¦Í¨’›µ‡¨½|AM¢*[g æX6«Qx²#ü o‹@xZCdJdŸ]W´D¢ãÓê=å3çÖ}܃…z|¸G\¶$—ýõ8d žJZ,šß{Ö•áZ¥¥¦kS@›çqæÿùcL(å«:!Íë®ËæGí­×é^˜…¯\nS¦åYˆã!|«ðf«ú€ ÷!Ê´Zv98Ô(…ÆÂ8înÌwKuIa—öxç\møÌ¿–ÙÀVéÞ0Ô5^UÖ3÷UOõÃï…€¯ =·y/Ï“éoj<œzÝM Cƒã~g Lzr=¸1„/1eÏÍ~Ã6k‡2β”Ø2YÓ }š³³S:gÔJØ„ß¾ Vï†öóY·]*püP¾§Ì•ºÚD }R£9ñ\ÉC`aÔáHê§J«Ö±l±¨.YvUÞ{ ÉPlGo(rËÖG¸“1ÚãW°Á+änž½8Ên"Óô峃'Sm°·„?úæ=:pŽ B¯ûVèÇž…óTµü*èkšÄÂho8CT Aµ0ÉÌl%$&Ž»dS᪠|±îpx…‡4ãò¬6Ñú,¸D¦ï!8й®+'—€ë[mg^Rƒòiüšª²Þ N†–P+¬eöÑb…²®Âf†âhôR:ª¥é‡8Í!ol›j+Ý"Ò–ÚGýB½(ÿ±ähï)IÇê¾,uV€Òeé…òÇÖË}=ÓƼi<ôÂ7­ü¨Ëý‚fÆ´bSý¾4…ìâf:krÄuIÛErUH$>‘Kæ†[ï<é¥ÊßúÈ^¼û+®Ú»Ïî9_W&tW‹±Üm¢G–%8­©>× \´L´%D®îÇ—ô¦£?öô Ò§ª‘§OÚ]àÕã¬dë«À¿}ÛøÒ>‡i ž¸´>zV”È•‡ÛÝpö–k] "BƒóÜŠÕ{mc¾U÷¾½Ê¢Â~äI£š0Ahga4cŠÂ©sOTT°ó¬ÉA¨ÊÜÏ)*›ˆl†€AD–Õ;JÓ«µ¼t‹ ˆö\²—ÿ4MµóB¡ XÌžcº½3+ÆS…ÍUĉßDã(ëiG&\J³|°û¸‚w1ÐmðØli :w¨¾YëBeL¤#CýAÐî‡#ÁƶÐ3mö¸“Úš÷ÚVW*ȕɨfló@Æ'¦ñÉt¾cMš"-‚7ËGâ xµ}aƒ¹ÎáÕ]Å •Åzs!|ø¬;cF 'S–^ =0lW|p‡¥h@l×çBŽG²Ö1b^{Yàa_»´ü´ˆnðþ––$ô×ɬaQDÒ•êÉÒ“®Y2§¹„Ži;â©+}eÎ{2¶õšÐÓä ëveÅ“íì~1G@Óü6)T‘®|þò-ìŠ1i gU—67ت5ã{`ŠüŽë«ÿOj©ÕŒ¹ »6©ñ %“W¾ÙúíŒ3€À­G2LM¯&"|¦š:x^¬…q¦“DÊÄÁmŸ\FUÞ³Ñ2#á[vÄèW¹Óh·}“PÏ©‡ðùÝb’rçvLÒE¯%öi>h^ã­7Å1òa]¿ƒ@i¬f¶i£(Ê°ÂÝ‚˜€ü;2‚TY„øÝ;”ÜõõÛM±±ÇùØ¢*¾»Ú³îô=8ó°ⵌLµ'`¥é ØGe¯ßbòð”Ñæ€lÓ$åªì.xSöçQa) Ü¡˜_pYkÒqæ̵êìÕáöøý*i‹?R¡ò2ž}lÞû@E]¥ Ö]£.j:YÈ>ûº-æxèøž¾?YöNBÎn‰6//&q¢7èÍ@]+Ñ’×¥\¦½n»tõVtJŠAôKÏsŽ 18F#f¼«·ÉË‹cÉ Úf¤‡ß²S MÃN]ÁÅŽ9 ]—X>w|©Uvåb(›(‹R²gÏQ¡ˆ$a“DN𥜥 ¹Y!‚ÂxÚÄMø›ÿÕío ÈñƒöŽ¹“®sÙoFPáöT¥/cžEøÌÄHèp(†ôh5ˆ¼µ*§»â[Fp8ŽèÁ‰*å&~N4F(Ÿç–RlZ{'J¥¤#N — Ä_)¢¨RlgÅkßFjßìÂÂz½ÜEèP¾´ý:òÂy!ÐíÚ[¥W a|ÃI(d ª³t ?PÌ[Ò3õAÓiOÛ!  mïC-èkΓÕñA@ k^ƒOîÜQX«Q6ojTú…ªÝÜ¿HhICîM®j«ÓÆ»N¼q…øcp§+åŒQyÿ¸àÒï ‡¹VXVÊØÁxåÝGeL3W©B†¡wÚÌÈË~h*ðæ(s"%$¬]SxóÄw€ÿi½¶A»¤J(höÖâ6ÜÔ㵋mw³7Ð,9ô*FfsÉ3æža¯£@AC¸7_·ºè´¹íLSìÙ?aëÝZköÑ3coírȳį³Ï½žNuÚ<åÖ"î´‘D;ùG[r![üB1ÿ£Ö€Tâ\Qû®R#ånRPl¼|?o•¿BùÌË.»g!N¿C”nM'õ1n¸y‘ùQ$šF…jr©ÄBŠÛ¦¯«y¡ÑPVÔ>ê­hxk‘‰?¢™ðêÒÃzÚhJ6ûãÒ÷,½…ä©©½»È½tlü,>­¾¢ÄUXêÄIðœ÷´d·_…þä;ømI¸9²3CfÛp3¨„— û’.rÒýÝÌãçuÌ茞Žé²N´RrœàX:¡øè7ŒŠY(êq¸ —õž06¡|çR™ ÈS싵_¿U¥•\©„-a&ØÔ±¤ìRÔ.•nVOŒá`«OÉ{Ôa¯%l“¿*JóØ-ruiùhÚ–ÔÓƒn\`]¸Cý#ݬûÅÇ7”Òˆ ¢§f–’ë(¢:=6ûÔ²"Mèã‘ úõ"ÊL‰-•[éY»ï Ñ..ý‡gÛm®Ìá›ìÇRšâI lÖ”zËP;h¦a]Ó8j¶k¶ }ya­i}Vò²ãv±¦œ/|IšÑÒóù%P"R* ýJùB‚5GhA8ue£8’õų!Ùñ³P `^”Ô¿…‡cA˜Ê^«2˜¹› 4&¸nù¿Yƒ;P$±I§6 ?i€bcè´Þ ˜¬F«×k{§!?ïQÆ&FE»´„D#ŽJ¯1qÎÄÿ2iT6Åø ¥…à¥víÕ;Ú 02ëí‘sA»c¹®³¿TC>ɃdF}D³s…"ØÇRQ7‹qо!ãéQaKN12WÑ.…¬Ç?."ÎKÉ8)—«Åè§|ŽŒ/n0êk×%÷þ—ÝœX¡*Œ°Yæ*Í2*uØâY`iųþ˜‘‰Ð•×AÓ£ZÓÓêsØb©ãÔå ¿~˜d€½o«m h{/ÉdF5"Á«É\“:hg0ˆÈµ@NMŠ:@&‡ÚÁŽŽ!`&úD £ïÀ’ð¤õh]Ÿ>ù»Û¶2C«ò6zÝŠ+ů®iHz„Üö³ß‰ §çYœÈ>cBò†0Pjñ©ÊÄ¥.»À#‚ ÄØÓ«í Q_â÷¢ø IKûæA?Ý Oê.¼>²×üßW¡B[ëƒ ˆF#’ðƒÃlÆ^Æk¯!à;­ßë…žìcëy¹8&rVåž”ñTsÃVš|wIW²>ú¦\­gŽ8Ÿæ¹¢Ë³”‡Ë’‚zl²‚%«V 8F?£·{…ëÃ="Üö07\6Á¬øJ«Ý°¢ù:̘>ÛÑà{=©;+ɘç?}‰Î¯ïd ý”¹×v¤£i%û lnU]W[ÍT s_V Ð"áÕƒžü‚6 ínFUuöªZÉx4J|J/äX‹ŠaE¦‹ÁÐ µs‘Í%ÆÍcOÙÛé<:G±eÈËtðPÜÝ%#ñ+ÄÉ$í$ ådj‚Ù´·ài`ËxhP™b‚_ºp] a&"Ç63»Õ´&K(œt<첄`BÀ.ô8oB8]©V8¡[ue·I]cêƃA× »°µZ_4paŒA?U"HUHËEt: ¦ÇRÿ¶Z~L•ÏsùÛ@ŸãøE¢‘ˆÑñÄ; ^ÄI’±F:… _žì¤ò¥<‹†Ð>FFÓe40óÿ¡[üÈqÞŽPäyæò4&_QðCîV%#¸œâ}ºsú¿!Ö»D|õZŒ×>ª‚rÑòT—„Cl|7®òï[>-ŠYâ´9Ï™•!’1Ð;ö£>u‡)sæ.Pº(ðù÷)¿j'99'qùìŽ î$áßÇü"’¶ÌÅ-1ù:¯%—¢~ÄŸ=üXÛº¨Aè¹ zr¨´v_u ˆõØ 4¬ŒÉ®òZ+1Ê9ÂïÛ‹¯š¦ŽÎ«ˆQ>TêîCÏ/“‚Ä/¿9~}ä΋Ì;v_X½šŠ¢«Òç;„;pþ´Ì„›oî,M™Àæ×,_/O8w/Þöõ$é„-tgZÔÿ4¤’¸5OrpÆÃYHÇÂü>²5Ñí‡,/—¸ÐýŠµZžàêwUn>mëÆE’þ ýÐ똽ÔkA•­âD:oÈìN]>R­f S4ƒ1ZȘä4žŠ¸5iIJTÞhj ¹›1 äÉG sO1íò~­p°èd2œçÅGr¤†rRÈxÏôUo¨Wƒ „«DÖX`sÚϧI§Ï»šÁNP÷±¶a7Îßqlá:¹®E`&L²°Ez˜¶Db”5éû±²CÆ!Ä\, Lî ?ïJP+MvÃõ%Æ'˜À7Ì]‹©.zjÑÍh±´£z!·¹x%ŽÁ$í”+è¦y‰«_Π\õ•û#÷Ù;fapQµ{ùÉŠjŠ¦N$P•®pi¾b˜@ß"®nÞ(¯ùoÝ»ü_qt+|á²F+nº[õM4N¦5]ðÔ:Ç_]_‰Šv8›£ÄSÁÊ$EØKfîÀçšK3ʬ2M[-—Ô‘?‹Fùª †É~ï.AÖv*¼KZxÈeëfåÝÛjÝÕkp¢JÀŽ=a{iÆÄ1[”ïwÈ;t6óÖŒ=3´¡¼!Ç<½ß(”zÐ~yâ]± l×<®wø1œ­Ÿþ%ïª!õ±>Ç]-ž8î¦ó^š2ìV錩$q}`Ààmžæʼ8ˆ¯uãáz+0>vØU‹ eTVz>Ô¼]܃Z¥D‹P/ÌSF–Bê5(VBd«.®éFôã^“…í+Í4dW©ÌU1©Ôvz‰Ãj.Ço÷“ –¼Ë³~Hcsb“gWE3€¹uEí¾òƒÐuL-®‚"¡äò½ÔÇžÝäºØ ddº<´ÞP÷Æ'Ï,ÇÚ¡µüë]Rœx©¦ ­FbGQ×›ˆHäm`XÞ÷„6êÓ½ñ—EoÞÒg d<Ï4ŽZ|-\´ž îë—p¤‡ª:¶Ú,O÷‹[$1 š¦<ïšæ ïÁ>ÁJUDQˆdBYš˜Ži˜Þ¥icŒeÄê ¯ }œRøäUGå|죷Š•3d=4ßòžƒDý6±x¬ ãœ]¯RX¹pÅ‹m2ÊQV¾D^€-1‹¼°š¨ò ±%ÑåÁ*$È‚Ÿ=yÑϱ«°ú0¤þœŠºáéÊ4jÌ3Ú¹;¸žÂæ™’á·»Lê1ºþ¡ª_ÈóqW†Î \ØO¹ƒ]ŽœŸz€Ÿþ ÿÂS`ׇՌ{ôrNi8yf%€w"Ð:“;È:6Å‹usá«ž‰nüôÃÝÇ”û‡Ú¡o+èügºM}ÎY#6m‚ÅÀ}¾:£OÂ2×¼Æ)4åðéLÒ‰C‚ŸJ±Îá^×pÑ/zÛ@’gq]Ç”I‰ 5lý²Cšc§ºp–6ˆÌ¨`­œE s%VâYì;K¡ô#>çãØo¬òýmG¦oŸé`¤·sRˆ†ŸŠD àp)’‘Æ€ïû81<äM7ó”ZH²6<-˜9+Åe “—–ÚFß>Î,•Qk¬“F«çÅü¼øÍ€ÜQc{¾ÂNŠraI¼=Á·LeËÙ R“+Ô¥ç¡7{ª|ÉÜ&Ôª%Ά’Y?ÌÕ¸P.nT³Æ‡ë«p@«÷úœ–Q²Ú»×ù1¼O '¿GáÀ¥GåÖاEôœöJU¤a¿rÉØ"h|3í§oëûmÎ(«?óÔ[écU$*”÷,N4#¸ \ü:Ú¾ÆD07ˆiTâV°}A¿[ˆföëeÀ¿Ruí“JågÁG”íhÑ-x#a»#µL¼Æí>x0Ëb'šÐI¿ DòôšöÝØ6>ëvZ}R($¾XUTx:Eöey±µ œ}1äfâG 7RoQSÂm[áÁß×vˆ&ÖrJ;Ö ô×D<ŽC]ä{Ø… ­aį/Æõí‹S/"¿{_&QFL´Tû©6Ò+HÂzÊVpú%^!ojέø„˜°¯&·Sµ1üܪè’È0Ó3R¶ ÉyX>fÀÁN á½­u/­ñ ™k{]yÆtôþ:fõú˘;–‚ÞŒDs›t¿9¤G­}‹}“Û!ü¯KÚÖŸXƒéšM˜€áìÏëJÚûdŽI¿i ‘ ÓL‘´Dâg/“„²#Ÿ±èž*„ °ì{k•Ô„˾ É"¦—ÄwubÄ×in5 PQj<(¢^ÀnGãUŠìŃ»ADL²Eõiq‰bŠm››‚ûh5ûzÝéutk»út‡>0šívß©ŽIG@&Š†ÊõJ0¯?WF¡Ë t›Ä¦V;žðb…çD–—]ÝtmJšMU\0ïq}Xž…ˆêøÔÖv–?9¨Çágfœ7UìÒíPebLX©@!¢‘ÿDIrîÄà^ Ç›]Œ>×\Jm±Y5³,•5š,W™ÿ‘öFÔ©~2Ìa œÿÅŠ­ŠEUåK¡I–J€PW¼d+ÛµõZgö~rjQ¨å¦\q½žJ¿‹û6Ù|e‘d›U~ŒæK¶èà©5Ú Ëf<ÁêK@…Io¹~²ü¯+Ò¾Ù!6øÁÎÉñ1®é±Ò+:ô‹uë®ä§ -'¸M IF©Ø8÷ÜÍòS©ÄÈ“\W;…ªI'¸ÞFEæ%£"÷6³bÆü©]VÞÞœ>DÙ¢ïN›Ý”Liæê:÷¡†8CÅþÚ2e˜Ã õž²ÇIl]yÝâöD‡LË÷8ü8›<Ï>TónŠÉ.Oè ÷Èù¶­G"Ô‘) I8+¥â%ôÎ`°)w›)¸«àˆæ5Úä¼WQUµÏ”‰:¸SJèT©©¾ùZê9ƒ:bDçP¥ÚÓ/cüLæ{/ö%+ßû‚ŽOè3*@ µªñ¡®¼¯÷´SÖÈ–¤?Sçº:¢–;ñé£ÀXE÷CûÚ^*[ýÕH¢Kè;»¶|Æ­¦Y«›]ËÓv?+{è@+æOœ£ÝÞžºÐá|µ1´Å\SÑ\u·£‚s~x¤Â.Þ…›FÝosѳ %ÇEà²ßÊì™2 j åp*V·’²ypLk†j£ÌõÎîª{Å5[‡|ŸÄF¥p‰¦»dß² \¢†O$}ÿÊ~p’¿J°ŒoFÎ¥èŸóx© ×ÇÂÐø¡M™ _"ÚHc æ7ÿlHš g¼9á^ùú¢y£noÔgÕN£x·ô"Nzüöå(ׯwàFH0¦!™$6nüqŸéÐ äÎRÕC±Ì‡ bÇä¿tâ«­Òp"•#“ ãœÅq–³Ú°´Èë»<&)oâÃ}lQÛp\DxIgFÆ‘MBswNhëå…мA×Q}ÈlÂ)"'¿˜êTÆmþ¸Áâ´äEZ…—ª]Y›ò>®\B©]ˆÆFb’q±DFéRŒhe2L í9&Ös䪳ÍÅ[pÅLïIÕ9ŽÕ@ >V„;aßsØÛœ|« d‡inÿ,SbÒCÂÊ‹¹DCñL®Ô)@sa²`õ걄‚ÉâÆá鹸+%+ÿª!áÃIjBägâ¨Ö,Ä vÒô¡e³*¢Åvª¦<½º‰ß"¢3ÌO‚Ifö3„“ ³åkÄ}¿Ó¹+Tê:EÑg,L_çßÎBqYmCÐÍï|0ïl¶M¨l£¡YâÀ6€EÖY£—>ÿnAªCÂ…LªööxŽoo¨Gÿ> endobj 1168 0 obj << /Length1 1487 /Length2 2960 /Length3 0 /Length 3912 /Filter /FlateDecode >> stream xÚT 8”{ûÆP‡¤I›˜x“=f±¢±…P¢3/3Œf†1†,'ä¨dí4*{Š²E¢"”=Ù YS„ƒP¾!Õéüÿ×õ}×\×;ïs?Ëïw?Ïý¼²{Ú¨`TgДJa¨ àHÀä„ ‘jp$RUHVö8‰A¿ãB²¶ N¢RtþaDq .fŒcp­¨À›  Ô”¦JK‰T‘Hô÷@*M0Æù€° R@º¬Õ“E#¹:ÀW@¯ Ðh-åµtãÒHx°Â1ˆ ÷D<Ž ØPñ$Áú¥„‚‘ÁðÔA ˜L&çA‡Si®úŠÊ“Ä XÒ|@°J°Æy€ëÔàB²Àq"‰þÍaCua0q4àd¤Ð¹)ÞH¸§6æ–ÀOò-Øò[€2°ÞGý(·ž½ZˆDYKÆáñTO…E¢¸.$21µ„3|ÊŽBX Ä‘éTn>ÎG"㜹kWǦ˜cŽËpO#y2èp:‰¼Ê±Z†Ûf ÁˆêáRt¡Õû“h žÛwb}¸î*“Âþn¹(—UoOÄ ÉË47^áBB?1Wh ÕPhmmô@_<±zÀq–'¸æD­Â\lOª'àÂ¥\@ŽóÍ `ÿÓñ«%„Bž8ƒ®$ŠÐÏê\tùfsçO#ùöH®üPrõ÷ãÍ«0•Bfý _1ÂÄÂÒÄ»ò§¡!Õ`«h*(´*€ÖT´Ô´€_«üàÿûzGZ¿Û?ê™S\¨únï¾ÓðY×…ÂúÒ(¿ž`Måª~Šÿ R‰ç>Pÿó ¬¥üÊ_­ò_Åÿï™z“Ék~…oÿÇó ‘YëÞdöï;ðÛ[’·Ç¿½æ w/0Wòö‘è¦$_p”ÄÀ¿Iç{÷¹KG&QÀ£T:iõ3Ãù/wÓðîÜO ;£5È]¤_4¡à©„ÕSÕÐp4Ž%„äÊJUC`£¸«I}× à*ƒ›xz3*Mhu’šÚÂxZ³´4„åO‹ë³þaq‚8ùÃB!¹¡Î8Ú?U.@ÃáA2èÂø¬¶À*þ ¼7Æ]é5iqÙ}·×¾ è â…ºÚ©xÝónçËò0L•áF=É?Nªª4¦;þƨ6i=Ûcsãz÷á,Ó®*”©£[…µ¡×ÂÍÞŽöÐ}©B–úœŠ”騫”ó•ö•9Þ¶XöxñLÄUkŽôa½ÇŸ…b˜¹Ðw5/jßn­Úó$2ØÂ'dxJ&Œ±ÕD·f¬¾1×ömªÈrR¦ÏG”í|h›ÑN«´«C8oA¯ñÖæS Ñ‚Kµ×'*²¡E=z¦BÃÕèÒäáþÕ‚9~zV*ì…ãoD‘a+çqö M’CÕcR&w 춞ˆn™îÈÑi9’1½ãÉÍ‚—ñšAÛqöa^½œ: êR£<ܖʳvßÎñ<M,)ä<ýô§dwU“*dSYm´ÿD_l﹈ț_V1$v΃âz‹ÉÊ ç÷tªÑoü•hß…äÐt½Ó\  þöxn^åØuºz%<ýÏiþçoÉÍ÷™ÝÜ&Ò£gÞwêJ+G7ä~F\RªÅ‹åŸÔy½¸ª”žTd@ð°¨1Çå •/á6Å{*s "ê õ‘˜ÀøK¡‰Vú•RÈ ›l·÷ôo7 _éî d+>øôµyë›ò^tc ô˜üê!$¦«}»Hùq±d“9±â“`_cS¬|½¹€È©ÃðXÌgç›èìÂäe`p‡ü,÷ù /x_cù°¿›ÈÊxÞø¿}u„y©ß.IþM£—DÔÏ»ƒf{Ç%ù-ùŒÕEgO7CÞ/4’}Œú _²ïF—-l.k¼yÔC~v¤L90[ ±Ð½Ÿ‡ùÛ艊{FEÓq Ó³­YþÙ½s®þ@f˜lÂØ»…f+‘&伂$Ïrª÷DæÕ8ËÛU'ÙÓsaŒØ´SKÕçø+ƶó¼­þºA0ý’RE°—]xW%ëÝݶ§múcü—Æ+o=5þ|Æsr±wòÂæ†ëŸZößoNT+è+ÖõàϾÄal2»£{+òlA‚ÛîØ#Ï„ÐKƒ2ï“:,#¢ßß/]µgÍ4ž|H¹n~å`ðÔ”ï0ßGü]Þ¿‡Üg^žÄZÏDä±.Á¨^HX™þêàÐòì^NWv¬û²¡RüÑÔžó‘’´Òßó#›´Â­¬¿XO½°çµzâØ¿,q7éÆp9[ÓRxæÃæaðt·^Wù' ¿Šc¤cÁRËÎÊ:êÓ#’öÈMËÉËeìòåýûü}Ùö]ñKqù+e¹Fõ¶›‚üøeoÂìϲs~’(7YjÄËãÎò,Q4cÁ¬ÕÓGE1Ñøs8Q$é^GÕk¨nC÷—-&KʷÙ™ÒâL¿DË0ém§°zÄjþòÐ;‰Y^av­„ÅÒ¼Ž¿hæM )ëeÁ¦dîÀ¿^¥Ã*ßåÄÈÝcò·]ÀÜV¥üÄô†ò¤µ™œn?,½ÕoH,8Ié(õ&Wu÷VÞ+:¡™®½°…£,2Y‹¾*’¤ZÙØßqÐ…^?Kg>Î2—·Ÿzö2òq‰ÄæÆ–óÛ*òã°[5)"a9/ΨUŠ½b‹Íäše×_‚LJèêi´è'îŒÉœ#=¨Èíèœ]jü•ð>þôÔ|~è/ ­‘Ðý¢);üGtd6>ØÜ.>ü¨à)xg§w”؈+¯Ò´xÈyÀg-rãõ‡”&<“’“‡Ô«OéÏÇÜrÝ»°/²·,ý¡.EŸoWt¹1ªh÷K}5s„Ó’Ë®M=“:{têêx5ý4š³à‰÷³OŽ¾†SJàÌnlÓn“\:¥ëæ!XeæóɼäÃõ+èÌ÷šEì½Ïý¯xÁ¶ý­Á¤ÛX°­3Ð<ÙøþOI¬üþ•Çñ5w*þLɲ¸Ú9JÌIî¸5Ó¦,KŒÀIíôµ¨ôVw!çöÇ=Àz›$Û±zŸÊ8|LÊ+î" žœ"‚}aŠúì„®w#\¦KÑ Â>fò•¾6sU&÷}¾î°+ÝÚã 4òZIúö©¢‰t`vÖ† iN»ÖÓkNsNÙ7ôðlÕÔ 8ß‹k=ú5,Ÿ÷ÛWbŒç l¤¼MœXg•C‘‚vå:<þ>bEs£‹‚" ‡b)å!Éjq„À~ –žÁw&(Æ?ÊD5=U€–›nWÚ±¸·ŸÊÜ¿¬qÑõø onÝÓÜtGh^¿"£T‡˜é´~ó±cärã¦áCj{ÒX_ãg™7Ìt‘÷¢ˆÒwf8´%!n<3ë},¡ù*Õý²!¬zA4]³c†ØQ‘ ÿSA=–ÞÂg÷ˆ,¬ Ö5EKcÒ!æ"yOgN’̾Y×kSY.œ0ÐÃxq [‚v`óºø6ø|÷¹76t~¿¼ÞÅȃd‡s»E5#íÊõ„o”‡ŒÚ)Í¿á÷®¯»zô‡9ÕK]ô˜„Þ÷KŽ¶#,¹º0F|çòvE¯±™Ü ßÂ[M—lÇæàn¶OêýåòˆáO¦pÑõþŽÁ¾|²ó®Ü®Áìgû~8ç¬Û×f.×Ä‹@Ô?äÏþèÔ¸¿2aß•(?êo¬¡#´¶¬~Ìë„rèV+áòîiý7íÚ´ÞšV} ¬£Ì²³!þP4sb.Ã=¹|™¡{G£X}׎bU&,ýÙ+\Ç--tˆÖÄ1Îi>¥ÿôäX×½þ¸c¢—G‡¶Åø¢ÑÅÆäMp׎q³Ú¼ø¯ÊéØŒ…ºðÓoÚÊy‘ )çRÿÞ7`‹TG~€©˜äw¶¿¶M œ’)8:›-FÒKoöŒÖ¸CÒ¤‡JN\ –¹ÒÌ`Ð?ÙÍÜz\ËŸÚ¹,d×9†ën°ÅH”4~:íÃÈ[¶ògg°yžPÏ­ú¦FŒ1 ÜðA¸b§ù'uÅ‘Cgª6÷¹AÛàcòóõÛvLF…5Ö¨(v.ØØ´õ”hï »Œ¨ggo="xb{·T¹óCs%ü@Œøáî+Ï%] »v%Éô× õOЈ`¿Û‰¿©ä ”‰HDZïNõÁœcb«ë­ØÝËÁA’äAKå‚Q‹¥w½Ý”²Óx¯˜@χ»Øl”h­½V*h³8ÿ{ ½ 6“Å % Ÿ’¨µí ã,ÉŒ~ª=­‡•Œó ^rÌ"³­ü*x7CýJ_²ôFŽño-fçŒ×`8ŸÚ²%t#`$ë&øŠvüÄ<âÈ.}·¬íT0æ6ü"p^,Ï„£$í“M/£¹_0ºsz¡ëRãʹŽ‚MIÝbé£÷¿×å´{êðtUéfâ9 ^2¤Ëò؆œe9p·\·+ç¯,î½'h'çÍÐ\ôŠÖæ´"Óùò´?¿w:xÆñQ€çüÝÖm¡| ËTÓÑ[ÿPaÑQ§nñÖ…GºuO+‡ö‡ÿ%3Ø;çõžY¹õÉ]ùË’#WÄ¿Ô^bkp. ý©•aqϽ¦{§¬æj.øxÖ’3Ý\9l‘aêå.`ÝÏ4·…N•C»—ŠÛ~åC;ŒÇ\Ô©T|dÿø–^߉Ž!µÙÓêÓЕvTŠËóDˆË>V»÷WÅü”¾'o“ .úÿÖO© endstream endobj 1169 0 obj << /Type /FontDescriptor /FontName /EJLECR+EUSM10 /Flags 4 /FontBBox [-8 -192 963 738] /Ascent 0 /CapHeight 702 /Descent 0 /ItalicAngle 0 /StemV 70 /XHeight 459 /CharSet (/D/L/N/X/bar/braceleft/braceright) /FontFile 1168 0 R >> endobj 1170 0 obj << /Length1 1405 /Length2 1737 /Length3 0 /Length 2632 /Filter /FlateDecode >> stream xÚT TçFpÁ(¢=Äúƒ" $$È–€e#»lŠ€’IHfB2!ÄH‹ˆ´€ Ðò(;"¼J¢R¥`T—* DW,‹ÚÒI@¬¾wÎ{'çLæÞûÝûÿß½ßC}_Š#‹€Ý0§Ð©4&p ô÷²4ÚF*fN24 @p><ë&Á"1‚¡Ì¿œE0„>'p^ ¶Hø€¾Э˜tk&Ìi4Æ; &b(á/*Ø‚¡°˜dèŒ e"„‰3ÁÜ+0f›:ƒamªLŽX„°!xAx$, NdC|à±—}PÂØ.Ç…L33©TJ…b*&â}ab ¤ ¶ÂbX s€‚0ð†ð 3*ÉD"â¿?ÆÅ¥„ƒ°aTLdHP,ÄáÀŸå |„0:öœ˜‚ÙÞ:•>Wn6[QA•É› „*CPà"|ø¸yRñ8Ü@(G„øbŒÈ‡b!„EåÍ!àæè ‚à,=1[„q1UŒðÍeˆ.»¢gL €Q\LRÜÏÁl¢í2³™ÉF£˜•Ï\åp$8¡Y ŠÄH`–Ë,„p‘Þûx0,i‰‰X8ÀqìH3Eù™Vé 7Á ^.Ä„€K€ã.Lü‘äb(¸HÇËÿøÐ"Ñ逃°qó”ô¾:ᆹ361|vÐíÑMñ›{ #äÅÁP¾ì=\9_3/g–«{ІÆs1'',ÈÍÍ…Î0t:¬7Ú€ø«ÌñÇ]éõ…Ù»ÑÞd¡\ 0f(½{G#vVƳc><Á#¤ ã÷Ê¥YÒØăþë_™òßd¯¨ò¿”ÿñ…Ü$|¾2l¬ŒÿG |Ù,@‡E‚á™ýõ‚9ˆDðq”…CÄN8¢<þ\ó±s|œ9#œw½'Ž °/&F_bT4ÚG1bËØÑÄWDLLH‚‰%úðHW”qÛfni ‘’‘h„¨Ì--œN¬%ŽSê˜QQ 'R€P‚Ç.&")æheÌ\.¥eMXÞsfÌ" ‘ÒñÁ©l‰HDì R Ä•ÞÙÊ…‡á8˜MºÕ‹±m“¢j“š&kõ¤”¡.»Ïr&·™SºŽ„«ãí®WwögúççÝöøÁíVÝ-<ªÅÛ)f² ïƘ|°nu½Ìbœ²Úí!ouDzïôø¼kßÉ'ȽËNªg;}Î,µ«ø®JÓ<³ØžÃk÷7\¶Û¨ö—›ßI×éT¸YoÙ×ä}úLMEž§.™4:°?Ñ0dùj¥Qá€àÐÞïM9ìkàawzŠ”)­Ö¼ÿkgÇ­¶UgÓöl‰Mz¾6×rµí »X,ªºSòÉÛÂòØ‘\my„ûrg]¯ÒÃûsïÀ1£dÛ/e,ÎõÜx=°å˜æOwíÜHC팟‹†ûO}Ÿ»ËnÓtËŽ¥Yù‡øäi½$h`Ç¥îÏŸ¶¬Nq­¬ Ö Ì¸òòFóŠOÙËOÏÔ^ȲJpè…v$Ó—{^Ïé$¶¾ËH…„©’½£ur…a[ ™x„dþñ͵ž˜Œ®¾ÆMø‘󛯷¥•$j|£9;æX¯J¬]Zp¡?Ü*(Gç›/rèhF+Ûf¹ðKõO¾Q)^ð&¯Q}õ%ûž¾¨Ý,’懯((8ba§Iêj”k8,w©r“rŸî`œ¾Ìµ²Ò¬9г'2ÂáSìÝäJ8á\I£ÝÞ«Ë«¥eÅEVûç#ôäëwiò¿\³è³ý ]ž¾üÔn Úïé7~ýPE‹õ­û~®Ñ¶m̦.…ÈK:w² ´Þ¼Øí7Ý”÷¦ÄùöxÛÑHfÁ³]h_TedŸÅ‹'ê#ÕÑÒÑôGûÛ¬Zí=£Ö"åù󋸱Z¥5/Nv’ø÷ý£?q-п» ¿ùÐørxdkË ³ÖÊ•^’i mm™º`útQ8×¹mÜî¿6øgÖ‘º¢}lÚɸ÷Šêô6쫸b¥aЯ¿´Y^„äjôgØ>a»¤¬£G¯èÄÐû·ï5—fõ.é¦õÝ\Õþ¸¬C˜³0Š]Ì¢iìã±æëHw¯ïèŠLÕþªgecæ ×²ÉyŠœqNã®»Ú¥^õs9S£Ÿ]oZëMoM¸=5vÓºN¾ l’eS{~g]”Ýâ¼`fÝzkÚ*ê®ð¸çs¬VÛòeTwt-å+Ï`ƒ©æåcw6OIßÓŸ~̯ö\óZÍ ªÏË$¶uXWZqäñG)1šwRݨ¶#I5½&_—-Œ¯»p¥þ@ó§®þ½ PQø ´Æëé³–B±%ž.]áð`š>’õãdXzáy«¦ÍýÇŒ ¿AÚŠ€ÃÛ/y4¬tºÝòʇõ“IUPØÁ¦¥W3aõÍ™–mgírûî_p:`ø’ýµUä4Ê­‹ó÷ÜØ_Ÿ·ê¹q}QŠŽ~ÉAäõ`L¼ÁØ!NÓ.lSw§°p¯ÆwÅˇÕ3×OÚŒM8HN|¶~Eį2éùÊ×åO±Á³é #f=Ø°mi›c~X7#>ÛãUN„céÉîó:ÚÈ­»â^¢b{Þ§BþuÃïè–ßS„ çlí÷äX smºývÅåÔd›G¼ÂPnÏ>ÃÏÈí º¯µÇŒóK¼“ÕÊÖ…Œîž Ú}Ñ{èxX_‰Ñ2ÖÖ£?æ? 3ÚÝ‹#!©Ç¹¤¸Ðè5#/«sFKºŽ¹ª™Çž„ª^m –Úæ4Ñ Y| çƤ¿(güRï;ÞÏÒNËúì©{ãó‡Éé»·-ç—‚Ë2P︦ú刞±C–n³EëÁÛã>ÍßÇŠLüÏ_»ëh_=éÓ”><±®Uu—ñ«Ö4Ü™›üKŸgQ·Æoðx69¹yžÌ‚Ò~k•Ætjû™*­¦Ñ~æ°œQÈšóløÜnbzÜ\§ý´H^«3™wÝhw2ïÛŽ#¦ „yÍ~Ï£:Ï©M&•†ö°ÕöŽqî&]¸Ëáêʪ}m.{gÚ –¸¾®SëZµÙeóüÂ;Z’+ä½éÙóÁÍ·S'î9s²v'^ßé×·?Ž|âÄWÕÃ/—¼µ -ç]fÕ.p©vë¥ýfp$Ó¢Üj/7Þ5 Î[]ýydR'øæçjªÃhE]J7o¼'ÔDøÜ[¤W¹þM].·,5LumÊÉÜz²ÎÈÙFܘg{°¿¡â/L×,, endstream endobj 1171 0 obj << /Type /FontDescriptor /FontName /MCIEGV+EUSM7 /Flags 4 /FontBBox [22 -192 1119 738] /Ascent 0 /CapHeight 703 /Descent 0 /ItalicAngle 0 /StemV 79 /XHeight 460 /CharSet (/D/N/bar) /FontFile 1170 0 R >> endobj 1172 0 obj << /Length1 1363 /Length2 6105 /Length3 0 /Length 7034 /Filter /FlateDecode >> stream xÚtT“i×-Ez¥H J¯¡WA:Ò! C !HB‡”z)R¥Hï]A‘Þ»(Ò¥ˆ€ MùÑqæûæ¿w­{WÖzó>çìSöyöyy¸ŒA"ª”=T …ÄŠˆ‹ Uq ”%ÈyxÌàXôo;9Ï}(G!þ ¡Ž†‚±W6 0ö'…èz â’qqY  Êÿ D¡`O8` ÐE!¡ru”›sÂ*þyð;Äååe……T]¡h¸ 0c ®WÀå‡b}þ•‚_É ‹uSóòò»bDQh˜²€0À Žu˜B1P´'øI`v…þ¦&JÎ0s‚cþr€PŽX/0 ¸2 àP$æ*Ä ¢WÕ {ú#7(ò/°þ_aÀïáÄEÅÿI÷;úg"8òW0ØÁåêFúÀ‘0€#ié‹b½±Â0òF`PWñ`O0¶¿üj ÐR5€¯þæ‡q@ÃÝ°Q ñ“£ØÏ4WcÖDBÔQ®®P$Cþ³? 8êp5w±ß—ë‚Dy!ýþ>9‘ÇŸ4 nbæH¸»ôžÆoÌ•‰ü?6 JŠËˉ î¨·ƒ“ØÏf>nÐ_Î_æ+~n(7€ã hÜzõGî‡{BX´4Àï¿ÿ>‘‹‹ p,À ƒ#Éÿ“ýÊ uüë|uÿh¸7Àx%?qðçïŸ7Û+…APH„Ïà¿®XÌRËÀÄò¾ÐoÊÿ8ÕÔPÞ?9€ˆ”Ì•\%%ÅâWºü;Í?ø›ü/«1þ»¹ÿJxéˆÈþÅájxóðü- þß[#øwCÔ•œ¡þÿ¨ß( t¸zˆÿïÀ¯ÿ›ôfùªÿw¤å@üòóÿø?ü`W8Âç/„¦Šþß è_[l…À=\ÿ·÷|µªHâŸñÁ1Zpo(ÄŽupúK;¿ìЫ•øw¬&Òù¹;Ò20 ö!^ DBZà'~µd¨÷/mÄD‘(ìUÀÍpD¡É^‰œ@ÌÁ êàòÓüË"#ƒýü6AÑ(4ÔÝãª6|5ŸŸþ•vð@£¯Öê×í^õõ÷ù×C¡ÞPò¹i”ƒbˆsMHÛI•*«—ÈÚð ž5‹'"~sèv3j’ʬGÐǪ)}¯hß­hòÝçüî·Ý\GÞ’dÒzîa—`:¶ÖJ>;ÊøfäÙ¶jíkv26³»ëþßÝýï»6ãwêòäº{ÈQç3œxõh{×¾.] ›^3Y¯”Ñ£¸(‰1¶ .žäɳ:Å|‹+ÂN*HÿÅ›fòèx‚>gä’S7Aˆ<`'F²ÐÏjI"ötÊw±ÜLÓÅÂÍbÅÌNxD?8Æ맶™ªË4ãWRøîͬwƹr®P0—îùbóqz_š§}T$4ÅÊ?v*z™`£~ô1>•2zHS¶ûöÄg¢’µx–Õ•Fo;EJ!£[Ð]^9á*Ò›m }sÜžHUÞ€ʸ‡Oø\@gÕ7 fÕ‚ <ó"¾ñ  ֺ٬瀄>ËNoQCÙªÜ)¯ùG½i­Ç÷Ÿ¦[Èõé¸Q®+[°Á'´ñÑàqÎ7)hÌžQŽ½ ùŒJÑä'¨m2‘¨Qå¬$|a§^eÿ!/ã°g°†] Û[’4.ÛV`°øq(¨~Š¯À<4ÍnÐH¹ lµÔàK0áå#M”ø tAêÒÙ^a4xx³pÙïsµußöG9 爕vòŠã]rE1‚ë,‹Ì7¹»/v‹]ªœ´_ÛdûÁóœ×ï°R¥È¬½ÌŒA‹–}t£SŸs‰â~¥¼ ¬VŒcYÆÂøvÞ~ê56†Ë;[áçÑ4‹™ ˜ˆ…ÍÆ ûè¬EZÞ¦»6!˜õc„3¥¢1Pè  bC8ñÑRsÛNûÕ‡oðfGmù_Õîâ£(ÊÌLöDR‡GŒ’Ü—oùØÄ ¹«©±¨ZÄô©˜­?ý ?ˆÎàR„å„~?cýÕ*ËL»ñ@‚å7eM#Þwx,Ã(‰¦D8«PG²ü²Û·i¬ç{q)…§Õm¼—°ÃI›Ô¾`®Z²; F±gVΟå²Ñe¾ mË”•aØ’‚ÇZâ͆*O,v™é±æ›T-é.4™.‰|j¥ñà?yÍѹªp\GιŽï "u´4îmúÔ&1Í‘‚‚ˆÓsñ‡›=_ʱÌìYW&wÒÎÑØ´Y¾b[ÖÀáY¥ûùZAFMšñæ5¥èpÿ@Áà'RÓµð-¸¶—ÚÒiáÚJ©8¥êYWé:hT“Ù‰g~çÒýFd.Oh5,¨A¦ ¡r"õª* Ï0í2F×røXeÆÖ;f±h¹(ò7’ ]qºö* RÄ«½¾kI¡©~tšêŒ¯(‡›»ÛßÛÃÆÐU r›÷Li8ê©Íõ«ƒ$×;?<"ñòƒ1y´Þ®“_!ä׃R­F’+y»Ã&'–ó¦´ú¨5’$œzCE¨b ‘~•çQZêë3oZ€B¤‹|Å 3éNýž8t]vI»6*kjH ÐŸ0Qm5­Q›DÂÐ+Κ<ˆ[j"K~’cÃ|AÆÙÍqºT‘[:€9»÷F(w14,ãáw†Ðä„KLáó‹Ýpð BX¾À«•LN¡2óz]Î¥Ÿ1Í}Vµ¦g%õÓI7„ÑIg¼ìøç¢ìíR”»Z)~5 3ÍúNUzÃDàÜt@RÃv“®J®s8ÒRa—ZÎC>¯>!?™ßŸ-ëÛ°2yõtY§`ÀåÈ8\ú°6WQÐm[{6óäckÜA2=½2ORXX›waïëø -¼’8¶žùÅ¢ÝN2;#K)“Š–ÓV¬Ï;Þ•|eÌuð¡õ¨ê’/·[¦ù^\Ñz¹¥A³ˆ!óÅr’BªCUE”moê|b§ÓÔÊ$Ë‘ù'¿F“Å̲áà…FYÜ]®ëjLw@NÄù·Þ®taiâ‚ U›ù½†>>ŸºGÉ€´{)gœ¦£!)‰Jæ6ðÊKN!©}ð†BÇÞåWpQ¬Ú²¨d.0äìÓ¶ëƒÿûQµ¹ê”GÇsÈžQë]Š'âÈE<嶊C´z¨aGÓšÖ[ðBñjÍÝ´×úl x\IK|LÔ(f¶&á#ѶÏÝ>çúÀ•æÅû/˜G‚¸ƒe“*·~<ˆ‚îÝ÷#LBKã‰Üíã¹ývï"sÅ ^)Órª¯;pã2 ¬B9ÞÉ3NÞü€LíBÃœvåÔY!¥FÞŠ©*cž¹)ŠW~áâ8êײu©RBŸ}¶œ>¾U×è©þÕ§¢GcM6ÜØZã&=þzTùxw{« ,‹c#ŠJh¥.r8þ•[‘Ôæ°GæÇÒs‰/jaŽÙ¶“|E=XI{]8ǡˆøÎ ñŠr˜)PmÜÁÿuwÎö3ÁãgégI 54}·H_M)FhU}$ÏØŠl §Ì‹À‘¯¿6{ç&HìØIP³áîohúsÔ{R]y&J™í œéë-fB³ÁÑëû/¸…rq¾páÜe Œ k¨Ýˆ×O¾ãsÂ$,â,A0éÔNDâ6U?†“1W(Õ ÛúʧšÙN9jœYN9ƒƒ™÷Zs­fÒñš ?¤‡íߧ}¼ní•ý¾ô~án«ökd­M8Íg[ŒÜˆ­î­?=Ú³óÙù)d¾ÊiJ¤ ÚRWL†M¸„öp8æç2e]à¦Ûk×>¸Ä–(«ÀÉ©äx˜ÏøL_šHøVH0èQQ=s6=]u-\ ª§*P[uÙNn£"þ‘åhÁøPGñû<Ù´ •þdÖæÁ=˜Ù“…Ax£'i?›XÁNÜسêžEmBh€ö9øä‘8l/Ü@ÓâhÎ=™6AYm†\T¨ik/` Æxé*ìL^°Ñ7ŽÔJæ&lÂÓe¹p'Sˆ÷ÖˆThÒhìNÄú÷y˜?¢þ‘€A…Ö”ñÝ™*¶à9¹Šúˆçlusº&Í =bQiº,¤+š/Ô*UÁKeö“FÄÈP ß4S°©öòzݵDúgO‰ðï(8rùXá/¬ T»‰Ø¿ÔÛNÔ¹ƒð«j)@H_²LóÏõöÀÁ"Û€7/§B•®ñ'EwHo„+§+hér€©uÙ¼)*px.i³þé Ä´ƒ}LA%£¸™½ÊàG…+à Ú#Ñ«4Å·×¢Ä"’wdŠ_Ë Ñ¦Šæ×-+ÍÂrœ6WÜ©©â–ßB“ZbghŸ-µI8^çyeÒ0yžJÊ¡ùaÂ?_#ºS»Êåôçù s”ä‘T …͈äË­Eœ*õ öX1\É5¤œ–t¼è:“âÞzV?ýœ}8öwœ‚ÿ{ݺèjTW>j²þ0ímïk²Ä† ‡L,O{òש[4DV÷²6¤G>‚h{=Ù Ä6…ª«§”µˆlx’,ººáÛ§9=e´= Z‚<¨´æÙ”þ’K½%ʧÿ¤h°Ê×u“èÓñÙ0T,œËka¼öª´ÇúÕ'™0Œ4’éÈ};’KÔŠ·,›¨˜coâíö.Tõ“9†@€ÆÔ§È;³6Ù§)w}†%^ö-ùŠXB„ŸžˆvWˆÙF\hêVPŽu»‡’h§Îk–û­ß "£<£›Öd/ÓV’0`_ã^;w-·£ødñÒNǼ÷­(ËHfc£y^@L>)meœd|°30ɤ}ÏÛ˜³ˆše¶É*K´Ã­*Ñ·˜Ó½º”àL®y£Å²çXd"ù³-Ãì~EGé-ׄ| wÔpâ–~”f|ô7¶·ìÉig|ŒñG'¡"ݺŽ†¤—[ç QI„“¬ÎdkÖ/¢ì¬Ç£eçúý(6ý¡ÑGä-Ãq™¡¯×;oèÆ$„ ËSðð¤(+F4$$¾zÁaW!g²ã;íáZ‚TQ{ýFPÿ¹„~Ñsâ±ïVÝ’ð[ ±î$Öv`ž©Rñ¯õ¹OŸÚÎåm¥p åb… ²³mdŸ™0•|d'•'‹ïÆ_Y–öm®¥°8b0ôFŸ[Ü“—žªì¼ ž^z(á1¤°ŽwØó¸ÝùþgÜñŒDsÈû»N>…­ËøþÓrÎ$ÂôŽÈ{:²‚/Kº 4ÌèÅDt÷†3/Ÿ?-Í5IZ#uáÛ” vé‰y–Kvf=p4;Œvæ\åOê²ÄÛ¶[$T{w0•^`ÔñKˆ —€E±³ZÔC.‡‹ZÌ^Ú­oÌõÊÉ«ÂõgcýmN³»³$ÎÕyռΚ>Q®" fÖ£ÅyßÎõL>:%Y:üø!­½ÓhZx5BÊy|߬cJœæ1}4rßdˆ0ÛVEx7ÃîÙðdéõ±a’!çÞ>—궲"u½Îw‚xŒ%hDRÏíOÝ:µ.c7Îì®ýH§ÕÝ/·GÒ"Ï=iFï>b»P»–ÆD4@3¨ov‹•ï õBåüÂ䣯ÁÎ,‚'1®ŠN)À•ãDÅ“R )t R—U½½C†T!‹Ù(ìÆÍ ]ü#Ï¿k®”Åâþ|@å£i»"ëÕôÞt{ÕŠÀ¶Í|e.Á“ U]îͽ…HDš +J ;(Ëz É\ó^Óu"¤2?PXPÛ «¨Ѻ¡fæ‡ S÷Ò‡´Ø#] ⧺ª”V/[ήÏ>·ÑO>ooê„ævM>¨ ¼]¡ÂWøÌ]Žþãýýˆ†•ëŸ‡SIé5·°žÚ‰ÉªtÚ.ÉÉ·QŸ2ò9Eëœ=G¬VêHQ%×HÔŒVó†é˜‡¶¾7ºÌJŒúؼEÌ£<ä±Í‹ýˆ[u²ÑôŒx?Y"¡&ê–ùe؇5¥g¯;ò6¢—p|ßgbæ¡S\§§™MÉŽçP7Ñj°œWz‡Þ€!®l>GÉ® )G0û©œf\jZƒf ö"óWdNQg6í©³åÚÌ᱋‡D—MW‡£\0U#ÞÓ #nËV’±ÞF€wÊÛÂp7þÜô)B\Ë)ë Æ›äb‡í÷¸f;Mž“’²«ÔÅÝAÞ%ýF¾GÀüâ—8¦›#·íßØž/Hy]s{mæɤTÕg¤C .¡†ïâ[m‘¯óh`"Ý[åsº=ü‚„TgC´ÃÒC¡ZÄÍ*#9ѲG¬7öÊ2zÆ&´~´XS~³y³s˜®ú-p¬´Ž+#ó­oÓú¦/½¤Øm7æ±ALòL>•ú·}Z›Aéè‚ B°5µ{åÄ Ø…×UõÁnzÙñžÛ¸[ºÖúòz½/ZÚ›#f*?ןԉR³«&Ëú8¸ÿc*öÞ0×röúì1ÑèfÖMûÇ–¬Úþî^uìS¹…Ï‘÷)ãùßžð®vxª«—Jþ0ÿpÓ&Ñ×gÄÝÐWªWpÍþò¼±yš²'ldÞ“F¯çjF“J=ë"Ý—µ([|ê‘Ç~­wÇ‚Œ„œW²¥$Ï÷âÖË/Õšýô© ª€úë÷¾#[Ô¶I º¥×X¢ßô0fX½mŽ,,ítD“à2*%½ŒoÒ¶ËÍ´%iÒ›oœ>ŸÍC0R6âhQî6³Á†¡á“m‹-¯z?f?KcZ3G#†»@y5’{"V¥‚O“(²j4óƒ_k¬ûÑkÞÝ üAÈVÝbx˜•tÔE³K4x ñdRÜÄoèÅvÌ2–5ÁF•žõ!ÐFÎ1‘p®™¡dYš³z£÷ê«wâRÑ»+ª‰ :,–ür¯0®ö¶ ÿãb-h¶Ør>þ 9E¶NÈ+L5$£[Ó€ÔJe…sö@|×D+Äd9<¹ fXV“Žh›j[çüûõmÒòcGb!/ýáöêÕ/‰5®‡ÒíŒévߪ¯¡ð´‡ø—Ý"ÔP¢¦~v'¢|vw«$òÞl}æ7€çªï8¢Z)Ó½'šÏd7‰bU„íöC‹Å¢¸¦u“7ƒÞàDÛxwqÏoêÇ·È.c±-ךۛ‹–‹ÄqB6+¢*˜gBcMnˆé‚†K?#¡__ã/“†ú?—œÀøD`òÀMþ›èú&™½Þm¹WÅ+‡±Ä®Ckd×êËØéötcÎðXyLM¥= èÁoB•£sw¿ERÀ`†Ÿbš…Ó·ÑaŸ»ã™ÆŸ^«‚Câ>H$, ô?±Ã@T„câgÚüýø7RÌÜÝâ¼ ³Ç%ÓxìÈ/_sè"ùyʨ¸8|Ñ4IYÊ·_’˜óèæ_ï¹8 xµW)°¯»{÷¬oa´—À¦c ÍB²ðèéÌ æ2ÑeïC‚‰Èoi™15 Xù®Ú 箚ÒQc[1Øuá(ɸ»zÓ:‹$ó¸5Ó¥:Œö“z÷ ²ÚÐùL[±“€©8…aVWbb fò/æNèäÃê> endobj 1174 0 obj << /Length1 1370 /Length2 2139 /Length3 0 /Length 3020 /Filter /FlateDecode >> stream xÚS 8Ôk¯(´ »zë¦TÌbŒì×:ö%ƒ(ʘùÏø3 ³ÒT–*R$u»h'Z(K¸e)É’e*•}I¤MßXªÛý¾çù¾gžç?ï9çwÎ{~çýµ®ms2ÓÂ3m mœN4@£±H4Z¡¡áshÐw?BÃb±a&ÃðoKDäˆ|VDÎ Éö\À`FϳÕ:h´Áw “e¬ˆ¡08!=“±–ÌàpL à‚G IÚ0[µfÓ9bÁ$"89]t#‰H& †8á¿”Ð4àp‚ Q(‡$ÒÙH&‹jºI ð`NpƒØ+"ƒÊÀ™H‡æ©!À=fÏL ‡GdA@ä Á$ˆÁ¥pdˆD·‚#p †s`Ç9€˜À 1?ÊÍgÏ‚³ÉD‰I&2ÂaP`\ðŽHNG ä ‘ÆfŠò‰¡D˜Fôf['¼ù6@1œçÇ&±à`Ɇi3Q3eDc¶f-™t:Äà°3ýYÁ,ˆ$š{8jþqƒL#â»EdÊ 27åÁ€C¸Õ1 oçnáé²ežò …3 Dhãp@[WG¤S¬.ˆü_ëü˜Àwö³^W"<ßÝß*Ú1(L°uŽ„hz߉„Î+Cs~m6_opfŠô ÍŸò÷AãÐ$Ñó/ÁlÊÓþL•ÿ)ÿv„çÒh³qÍ9Àĉt˜>‡°æÒ Ö?Û¡¹5v‚È0—þϨ‡(Ú s•öc|0‡AdW˜C ˜Ϭíį¹Ö “<³<:8=@d±ˆá´H!:¢gŒÀˆ¶Œ …ÍŠ  &G”‚¹> 0Yˆ™'Ñ3(ë׬¥(ן–@¹ý´ôÊsÖú¥—Åm×싺ûnÏ®2…A$„ ™I2Š Ì‹)š¸e®ÊÓî®ÕÑ•©ø7*¹½ËZ9´w[”¥ñägìŸG[ÍK_ô|'û^çÊt~—:h×Û½)G wŠSü;µöÖÉš|}Ïväêô8<|¸ÃË S† ܪx£f›µBFº›ÒúmŒ7d¹0,±ñÞ RzQíñ£JúúXéE1ÅæiÅÙ¹æ/JìÖ³?™Õ~HTÇ÷O¦<~ªƒë¨xyMk!Üž¢:0 ì0š6`z¶¯\OÅÁ¯4jt•[+ OjFÅDjú³k&6C!·öý%à6}mŒ éôdL)EeŽyÛËÜt^íl¢`Š i Ú¨Û*ñÊk`âµ–ùš;µ>ºÒW_Ûø%‡¶Äï—³dÆ ƒZ,1ùþ)çnœ ¸rê6~zI›–ðñÉäQœ¤k.ol]£‚µƵ×UOPÔ“N|<…ïk{ͦ¼ŠTÓƒ÷rƽº²»N Qd+8p'<¶üsñÕö•‚GÕ‰«s7âkžfÄp6%ô¸ì³}ùv|»Ÿ‡©\:ás%ö—Yîïéd0š©W;x)¤ ³ïRët^É-ùzÛ(1÷Øý*•‹)÷FpÂý’w‘ûv¬;Ô=¶÷²ûËmWã:ó-"_y†H¡—ÑY7ïZ©2ázsõ̬âeëq $“ó&l¥…^ ;S `f⊒§|ÛòÒïr£6a—¹¦‰}½ïºb]Vá£@¿ŠEqõm~2üOn›ñ±Ûìö\á|VõK~’šCê•FºÞ3=! ‚ÝÊäÊ>1n8ø¡¤ó–§Ìëãb¨Œ8¥ãcƯ¼–ËOÚžþœÝÑ£ö Ä è¬wÏÞxB~«2Q,æ½[eV_û :zÑõö^xSG]¶Téo©±¶ GLu¬¯ûâ&'õæRɈ¢’ò|©kâÒ¡“*¤Í·Õc‹Fx õ¼³w»jïkÙû!\Ë¢¬-(±«`ç"·L¹$Ý V*Ú¸J¼¿Ø<2\òExÜ]]~p·*ÛÜIYíã­|ö:Ö׬®†^p|°Ë(±¯'êÈÂiÁÇT9¤ðÔºÔÇî5A(þõ«¯í«ì§¾5¿ô¨ÐÙ~Í›wòÞ y÷cÄ# ¦\tyáÎœðUAîÏÙªú—¢÷­9éÐ;Õ…¤¯Î';®¸,™ÚÔfünsõÓgçT¯:*¬98À77[{æä Ï°ýèßrfXåx¼kKêIÇŸ÷œØm.! Ùq»5B³,Z×å¡£u³§°]ȸóû‡ü*õÊ-uþñîûÛ †>ø$ñ!âDp{º¥²#?'`#AQ e¡ßî0ò!ª*V]zw\~ž`‰ØFßÖ‚JÊSe]ë£ºÒ m„V‡”#îS⺦ĺ&üù?V¾™‡Ãq+,m[Q-òŽ'mNiB/î%PóÓ?V¿bfW-~º4ãYvBýHìðDÿ®3Ý.n>¶y :K‚Ô¼Kà¿Ô*cõ´/³f6ãó;¹{Š¿µկܿètl¥»kÅ ÅÝì}…ùC벋vO[¬¸ÈH=8`ÅLªCèª÷Å<ò‘z{‘—ÓöÒ :{SBÿròîJÖãÌ åŽ‰Ì2þë–¢µ2OûïE{ЇþWˆ„¼…Ìùª8%µjbt8,ºÖg¨Ò#ø»BzÏûŒñ8©ˆpnyoZïžQi@1ø—ìÚq“ÖÞb}ÎA„ä çÄJB1~ãuÌ®6à]˜»¢¬îàå¦g¥÷¬áœÆjúd̆ÅyÓNbç“ÏíÊé‹”®îd;'ÕÆ»ü3M„J1Ù&K"_mkcíýd-abÔtÎ|´wòyEhî®*Í+iCÂݬe#Mòò‰êVŒ®SìM²µ÷uÉ+ÊA¶”-‰ó¾ñ„”µ?^ÌÇ4úÌ»Lz­Z˜ð™GN“-Þ~æ‰é\¾àÍ°U:É_ãÔ«žšVJ?L]?]PF@xß›–Ê1³ÅIÝ_ z¼m:uÿà^êko1Ë«ëF?'!ß{DµÚŽ >Í£[æþÛê+pfX*O}JÙa –!Ö5ýA­ï/=þÊPÁò¸˜Ø±6v“z¬7ƒûSÕ'ö™.e×Å’ Ї»ÁZrj{oÒ;§eçäU4ö5Z-¦¶F\­.ÒoFmϽ¯z™~*~ ®2n½ø7ñ[]dý®Zž»°€S_»;¾0÷Q&﹌¶ÜÅ_º½u·hÈMYŽK^ïéWE5*A¿¡=¥¡Y%úJ–¶Q'›W Û7ã'3„üî© ܧñoLb½Ì²LL›ýÑÒ+o5¶0êÔαÈ#ܯ=Û¦[—¢äž ú>/]J«v݉L6Ý: »'¡lW´÷_»jcã QÄ«£ÙìMõÿ# endstream endobj 1175 0 obj << /Type /FontDescriptor /FontName /FITBVO+MSBM10 /Flags 4 /FontBBox [-55 -420 2343 920] /Ascent 464 /CapHeight 689 /Descent 0 /ItalicAngle 0 /StemV 40 /XHeight 463 /CharSet (/E/P/R/V) /FontFile 1174 0 R >> endobj 1176 0 obj << /Length1 2417 /Length2 26418 /Length3 0 /Length 27442 /Filter /FlateDecode >> stream xÚ¬vP]¶%œàìâîîîî.wîîNpww·àÜ@pw >äïé7¯{úM=™ºU·¾³ÏþÖÙkË:)¡‚2­ ±¡‰˜­-#@hhgK+dgm ``ø°00À’’ ;˜,ìlE€N&\1 €,ÐÀÈ`äàbeäbä0rrp~8ªÉ:;Í>|˜98™9,¬,ì,°¤s€±…™…ÐÚÚ`bkdglb °™[Øš>À†Ö&G;S'W ƒ ÀÔÎ`ï`gìldakp27ùƒánob 42qX[™Ø:~8ÙÜ휎#;{w 3s§+…å_áþbPvwt2±q¤Z[”þø8~ )™8š8¸˜Ó}Äöñþ¿üñüqÜŸÃíMœÜv¦ÿüÞÁÞÎøç( ­ñ–…Óßc²sp¤ùcü æ°µsþa÷7&&Æ4gÇ?ÿÆŽöÖ@÷?6vƦO0f#k»?¼>Ø;9mMM>–®NævÎNfâfï`âèpu°pr2±í?Ð]€Öÿ(Ý>Ý˜Ö Æ@'àÿ—”3d,>X¼ÿ“úP‹¿òíèlèhalt°0ù‡Rüs!þŠå_áß ýèþ¯¤þåß’ÿ?NýØ¿fþ¼?Üþ¯?Ñ™Y|ôÔŸc>Ž46±:XýÕÿ‰‘ñ#h£?,Í,laéÿ̯¤­©ÝŸ™üËnìlÿ÷=ÇÙPüïy¦ü«ìl?ZÃØÄ–^ÎÎ飺 á¿ô?˜ hý]ÃÿOìþ9ÂÙÿLJ`ÿÔñ¿)$°ÿBHþ 2ûÿ–‘ÿ·ˆÀþçDä?ìcØÿ¢„üä¿•èØÿ¨9þeúaÿ³âñõìKµÿ!á°ÿÍþO1þçys¶¶–Ú|Ì,#Óß… 𷛉™éÿòÚXX»ÿÍÿ/çvP7ùÛlýøç=É°0´5û(7ÃßLŽbn&Æ NFæS µ£Éß쪶Æ&úAÁÎÑâÏ×€–‘áŸö>êddeû'¬Û2ùÈû?Åû¡eEK/¯%£¢¡Hýï>IþrýÓn)e§”ŒÿÍð׶ÐÂÖéOïý=à?€­ÿÏZèä`áÐføPDÆÇßߟtÿE»2äIprrlâÁtLÇ趥öÚ´ L€,˜‡LSÀ"y;pN5x‘ý¾Ù jÑgÄ,?›µ±kÖêXº·#+"·t^y>?|O# š ÖÕ˜â|hBÞ°Ùí\£|ö}cúͪŒžÉ,íÀÎj3]€Â º€ŽÞ<ì_dl4cæSªÏK8”˜ñåÜ™Ç/uÿTׂKªPRæy©‹Tl”›åz‡ ðà•çæøoWû\1Ù}èŽ ¯Á"ßÍgìÀ¾²úýŠž•dïG]!†½Œ’oþ¡P‹¸åÌèû¿Ý{ÌYLíÈ'¿Jîù«ð¢s8@€ÿ……ŸËÑnÑ9ÜnßKÙ£`[hɯh÷)Qéºg(%NQÂÄh^•Ýç ØŽn$ܱÒÄOV»ƒâœôºçª;}.áQ68±÷ÅbÒWeµúUã_RŸ ñJx³U~/* ͬ/‘–™¸Ü‹È‰'É®}ò3w1V«o¬j#¶QÃåº ­÷°ÜuV¸zw´ÚÎ f!¡Gù"ÏQIg~`zó”Em^~â5u5#ÐøýÃ툆©Úov{§Ÿ{»®L:„Ôi“bÕB„¥U-Uä[á‹ËLÙÖW¢l"'®áï÷ÑÓ1¡Æw Þ%ðøñÜ0`zs4ÖDˆcäq_\Ì"sœ– ¹û4÷/®¾g°½¿Ÿ„)—l L}iîw'ß4SH¾.¶ì|[I¨®ÜÒt[æ6o˜[ê»›Ã×Üm³nÛ³6ˆà7¿z_7{«Ñ¥þžq°Ø[Tr þ¤p½ ¾h‘¢ÜóŒ2+“¦Áˆ5!ãöŒb‚ ΩK¢³þ-U»åþº†Ã„ü./DÖÍëÉ°ýÈ‚#Šw²TB{–(§ôTÑoò cŸœ²X q‚'DKøÝ*G…72¨¡¦0õm~W¸‹E~šú¾µ?3̦¥öÂ-5¦ž…gJ> ªHy2Åv¼Ûº„ÛÔ¡âæ¥qû;dâ“ë@š92à@b<+‘×mÖ¶þÜ…È#IóE’žÁg?ƒ]•zǧ ³ ªÞegM\û×ÌG‹Åð;|Rº.ÐØdúœ}Ö¦´|j-Y½k›rš2ªóßñëB¶ÍµiÄÙ÷Ûb+F·Ð‚î[uó#å°¡Éœ†adR=”Ô¿¤©×©_ûÞ@cnõ‡• ÊÍëæB€däãÉš6LßÔ J2·êʺÔ*ÅPöF\ö]U$ñpðÃâ׶K{«_¦‚%_ÒÔØ Éô~—Dè¦ÚyX¾6ÏÜ#±¼æcÇ÷‰\kH¦Ý/Ã`¤”$í¿PSp‘ã2ëçó·7'?D,ÎR¿Z 2`éB|)ÚlÊ C4<€Š?LÍÞY¤m8¬x!Á éŸÇ`ÀÎ Ñr¢0ï©EL›]6íµÄ,"ó³ÏË#´œÇ¨GwÈs ¯1‰ ô^3\ÙìâÙè[Ë™ ؘ)"•ö‹²oaCökœ½g9ÓšCÂR³äqNïHì$Õ™¦››º—¶’ïQÚ*†˜'ΚARNw?'¡`˜"Wš5^Í ÚijٴSÕr7è ñ~úºž{³5ùpm:(,þ8†“y9‚—¦³µ¹™Þ3üj©ÕŠšŸ+ô‹–ÖC×õ,ªÖYé"KfÒ“áÖD¯7td°¥×7Ò5IÅ”ð£[Úç<µ9ÙZ:ߧÍàzaŽ†žYé PvXÊ|úS²^·Ô÷¾8ôAD¼ÄŽ(yŠ&½saóM-B÷]µ‘Ï?«©,Êvj¬l‰Ý–¤ýøœ¹™ŸýÕè”gë² ) f°0ǯ>¨Õ¥ˆ§ˆÈŠ{®¥N¥“ÈúPÝ mܾòÕŠ¹vMX‡.çp¯}Ýij$f4ÈÚ Ë˜Å MfKnÔè>€0x2V³>‰\caÛôÝÙfúŸ%­|×ÈÚG ƒ'hÍ7»²‚+lÿYózŠ-½¡Â™ qÀfUð#tE¶k0—(BÞiÚè*r;PÍÐ×¢¸Lºí.3…v àš{¾60(^ø)_ªæ™Ôæ‘ Øåø íq„ˆVGÀEùX¥9]BG³ŒM™Ûþµ…@r>R%a E&„CiF’7%¶*‚ú] g‰ ä Œ]OÌéqDšØš¨ Ï¢>DÚÙÇS‘ÌýgòÞõk¨ŸßH2( ß!×›ârûöeF‚}.sÓõ» °ÉK†-3€W‰ŽøWVJ…­S¸]>€Ü (<„fRo%0XrÍn[ö[¼°œiBCm£æ¬Áƒ-žéïÍe7VEˆfóÆq©gÇqLͱŸà_ÝöÍd5àxÇ8¢àÒcOx ë)î¹÷ewzöE ¹$9ˆ±Ó÷¿š#˜.lŽÚ~êäoXÃq«Á¨ÌÆ–H$o!ŽR^¢ü}<švF‡—üýó•Í\ogAª¢B¢àl0þN†RKHÝm ÚÐ÷“;á4Ä-†Wd&}V j¢>+çÍYN$p¾¾åu>gÙ§½±v³™öÈ%€ȽúàÛÞe‘e‰ÊÝÛèáQQ¥â0Å^¬çr¿CÚ¤’]Qä>¤%½%ÙŠßQ?îWRÆuH2TŽ)5+·‡7u§Poxì›ýJ_±£'»}dÇO–ôÃ+eëž‘žAäþÉ„(â1Rì %-+—:Ë ¡_–/xÕJT¸œx§ÔaßN|™˜±`{+Ê„¿bô@áöªXŒªWdÊ?)QÿÀ¼Vq°¤™ÔCŠh î75‹sÀº3õ?&ÙŠd:%Üøùê§Ì¢¾8ÞúbAíÀG)æ Âeú•Y—KËȱøNZŠ7¯£ 4¾«B¯àaC*°(¹zÊ \pGÂÈUÌêÛ¯;‚4šÉ¶äÔs«yOr€kæÐé2w †èÂCð5dà»´2Ä~ë’£ŽvõšêWFG߀õ+!Wë=~¨âÛ#?_Ï©ZžIé(-éÊ“:Á±Öö?»[mµ¬?þ‰kµdÿn'—ÛÃzèž>ó­xeôëà´wÁGyY}L™i<-¢˜ŒÊz=Õ¼gŽDÓ-û{gÇ@Qí½£È0I¨eúP²ñò¶[ðö’ì2ö¥xu„³Ü±Ñ l‘uZµñS`6eb?¯-Ÿ,w+Ö-œÁŸ[ wtÉèûù@»—˜Ë=Š©ãT­,XÇLÚü,fžüÕÂóûawžr|ò6‹( ŒÙûý17qT2ñZŒÄÙpMœç+½Ó’e‚µÌ«šÖ(±LÙmÆÓšrpY§#fáÞ1ú¼ý~Wº4HïW²D±Ïle·C؜եD†ºÒiÎ<·]‘tüm`—}à}Lä€Óq'à†À÷ƒ®óy OÝ{KΚ\ì.4:ˆ{©rcY¡=å‡x´„Ðܸ2kä)öZ1eh†h¸„Ed)yîJ-r¨v±ÃágËñ „‡µjõÝG·°QïÕ"ó‚Æ£áUÀìæ%kÅ$¢hÞݬ‹tiij¢ í5” ¶D:gJóYç@W8y^ ¶—…í ߯hGŽÄ7OºÖÉûüAüÃx<ñü9˜w•°ŸaZï£ú¹wF ‹+„ +Eo' ´ÜÝwí­ruΓ¾düJB“/3÷·˜„HÓèŸR;MÒyc@°ŠD¿x;î Úȼ&‡¸iu¸*_U~ßîh}hŒ”kã&JÎe¿ë¤+šãC¹ÊmÄhÛÈ}mÓ°J¶º¼y_¬óÀQ pÓŒÿÒ¶¹AÀ‡Â ªU]žÜÓ²UßJ>ôDâ\½9sÐî9©õgòàn’AÉ*úeV¨<¨§å1Ö´ìïKhI!¬³ño•kû\¹:ùÍ“`X¡öé¾Øî™ì5<™ùF»´uàâd÷ÈÍ»7¢å9£;¾Ç·á½ZQ5.®™â0¤ Ä¡°äIQ°É¦éúi³ .,˜} Ñ£dj{HÂ÷?ÒBô‘xlsÑ™ƒ»%ð–Ë ¼“Lðyõ£^Sq¤sNäÇ<òêDC +C³pÄd¹pÁàñí"I8˜Ò†SÐ`Ó’?Ô3NÛX†•)Ž—ðs]v˜…ŸöݦYt]~-QÌíMý:áƒßnõ]?ÏÄ·M"Õ)Ž¾^ë /3šÅD ïšÔš`çsÜ j¹æ4àgù\BÊ~ð2ÃÙc9°/5Šßëc)J1åؘÊÀu¯Ž¸`Ë™`‹„ÅK ˆ›=îAt¸éœÀHŸœ9ªu=CùÉÆ|ÕµŸ…&M LEHml4=ͽ¾Àˆ¶. h:³\œGÙ Öû0ùÍØw¢Nó¨áhŠºr.Å&%Ò Í•,ŸOGŠˆ™pÉbjZ¾Ó¯&ÚGY½9æËIr´Ö©›¤öø1Ìú‹möÓÃÏïìØ°„L)Ö*ïu„¢ä×5"ðÕaÔóIÜσäµò™Î‘ ¯úµ¡¶5[ŠN 3yÌkNÜ÷ä$Iæ‹é÷ncßs3ˆi`à[ì÷C¶,° ƒ<1Ó8âû“‰.>*μœðÓn<‡"Ù÷yâëŸÞS­Úè·ÆQÙ½5ª]Ô !#Á†ŠºÜ¤Zº-µÍØM»LËÕ“NÛn¿ˆqã„ÖŠ°EE¶I/ε×_†ì{ŵ YUHzà Ÿ=ÏŒãåy~á$|K Z>>ÛŠ=Þ,÷Iz{:t FĬ‰gÎ!0ŽöTgh(ÞÇ+í¶åžÚ·ÝõîåÒÂÀ(êTë’J¿eáýZ›`ƒœ}<Í»M=hQOôSâΞ)sõtž"ÁÔ7öÛQØnø G¨xâ|˜Pàû¤´Örœ†¹ Tܪnʆ—d håQc 6s¾±ç>,Mxg·|Ä*ÑS!øì noç ‹ûÆTAS…ªÉ·§D4%ÆÝÁLÆgûLJª ¦­ðÂGàj¾õnÜx°êK"æÃȽoÏ·ÕÑFÚäTà þå;’µM1>™½éOÁíßÆÑšŸ'RâÎÊ„…&y¹a±_™:·1îvÓ±»~î‰çzÌ|Þê˜7º>w3Ñ_P?cH+bmÉ'%ç; ~^_Â-øtfû©Én@£”SG\0ëà+‹ÅYNg•×y—¨]¬0²ðqI‚ â5˜Ò Hã2“&ÖÊÒæ§Ø€òR‹+Ùv»´MœÒdÒ(Ê­3(WÍù»¢Œ–‡`ØySM´+d8 r|Ãÿ+ ì8áS Ä1•YC2Ýè¹áHػϕÈYòƒy Ýèá)AèP·-ñaP…1`ÿ Ǿâ"œ}·”bC™uÐë#¢ª&ÙäP’ÀÔ Û›ÏÈ ¯™#Q´^ew#äö¢ž"+¢ýÛ>þôs´½i‚fVzàþótÛÚTÑTý*ÂÓUJÙÖ;ÁªuÝ6¸ôwÚœOì¢/~jÚ‚ÎtÒ/4{: =ÉC´—6ÃÍï½k>ýM@>+7ê`ÅÌã NÏRãhš/æMáŠÆeNi2=z–Õ3è+͸·Û&!PwÜ’Ù3Í4I3’;¥DŒ˜‹VQ©‹†&ÎëÊyØ;j´xßhÛYåëcË Yò¤{@µf[ IÐåÖV‹vÁý¼]“…Ø]øKŠÛA0ÃØlŒÕ{—F|ݸO;- ‡ï™C©Ê–ç›ØÔƒ²ßà ŸŒ‡»@H1*q _ä`7;ài†°Óú’'UU(ÙS­†øD؃ŽYXx|â­F9æC݆+K$´´ï¡¯!Û‚ÓŽEƒ1lN9gŠ890ßù¤9³¥ùÍûo#}ß5ɵ¯ŠJ„cï^ë "2MLa9wPÃ\Oö+Ð_²ê܂Λ£©‰.z'VÝ@«Hèi·râ™v Yr²Èm×½`x^â”A‹öJa÷ÈYš—çÜPéý¢« jXµgRµìXí¿Hm1hkK±¼ãj‘J¥j7ž¤°·7rݹ÷Bø"ó‚O4£´ªŸ%weÔ¡,üËÌ{–ê3xwE›ñ¹Í;ø0¹Õ÷Íï…•%$\2oöýN]ΈëÑ›‚·`¯”ž\G…º®ýéGw§I6±æM=»{ñIH•³K‚ÀÁžçÚî˜ Il8ô›‡Éª,¹tQ)µC!ŒáR¿6=™ê–†\¢ÒÅ‹Jƒa~äÅcÚdâ6aïœÇn%/YŽ Ú8r&­¬j¡4Rú]Á˜@Ÿ,VŽ¥rÁ­x«Ø"~ô ômÈ2Â_<ðvït®É%çÌ7?år·o¬ZdIü×`l2ØCëšìùà €Ä=ìÉò?÷tEý·5§ÓMµb3>«ª]î²É{6! ܽ›}ÜNµ†À½c†©æ"¸û '“÷=U²‘‰‘KÍ© 9êü”¢²ÜËà›6ÿ Í`záš3nݵo0åÐÚnËÌèöuÄÇØɾŒµÝ »¢`i¹$d¢“𠺛R2Â(ÑŒõ«c?ˆøMUš­"°N%"|çû¡ Èüßy\ÛhAPßräŸRõs!ƒÀNß" s¶ýC27÷-×`^¤²°‹<ص‘,K!ƒyCaö‚¿‘™”„tDûFOFÂ;®fãé¹+Ë/¥I·ôv8‘A8ŸøÆB ¾¯UÚ­×iÖ[È¥ŸBqÆv<ø5HF[ûZ»~•²Æã½2ãb%ãw‹ëNkÉÆ•=@ªV­Ûy›Ã‹®¤‡ âõil¨êï™ûH« ý—ˆ5eË5â×ÌÍÈÒÖo­P:u7ñÞn›Þž\·uŸæÜlâЃ&‚µ`k{ga$«OtW³_Š½²Ãö—‚ŽòEic†¿AËH-¾ÁÀ]ÓÀç„é¹ò—Á Ž¦O·rÛ¯*iìñ i±ø…sÇ^È5"dÏ@ícñ$L›ñýZ+ø.†Î°>€©Qüê0þaªÓ¤pšv '|Ÿ²>ÍœE¿&è-³ºÎú¸†Hëm“YÞ¯ËÐ-¤Š¸âçrÂ`–(9Q’Ö “!F—mÜù ¯ö­€Ý[…–®?–ÓWÒÕ¥¶ªžiä8Àîb€|4…íÖˆ)ˆ•?ØŠHÁãé¼+h49QéR¢ª¬¤àu3ÖTÔC©ÓòêaÝÄ~ìägT¹¤%\•––P¤ç¬ö¬ã¨Ar„Ö:«!’–{T}ÿ¼'¤‚=ZlZ ÛsÓµ4g=A¹ü̓êv0>3Ÿ¥ºU¥¹þ©) ¬‚™i‡ý39›_UºòWõµ‡B…‘ÑåFàÜÙR-—ŸŽîwl'K?­ŸLN+ƒJEáZ!{Ïjª@ÒQoçßF-Í(hVššT­p aìg±6Ü‚ã¹j¶«@ßbÌIiŒÅó/à»L;õµS\ `ÙŒí¨㪩tSk„^Þ-(õå¼ÞÔ=‘*LÚ"j0ŽØ.xQ!úÐøMrZœG]sÌE'„ã.ÜbBû’k;’Ïa–áòBvÀ{ð=w½|°úªˆÎü™Œ#²Ð °ü3koÕ3\ ù󯉰³„Oš=­¥6ŽÅÂçêaýòѾ ÊXËOÜÓSÂÔ¯f¹ j Æê¿6$äÅ'‡ÅI¬§+¸„Î0†Ç=ïÿüÁñÛ‚p¯(U·.µFŽ¢¹VÒY‘ö Ý †g<]£lÁlG—9d«.©ÆRXlDpÔ£Í#žà!o¸á}MºSp„b>à ÝL‹¼>â’Úâ6~üV&šè£™¶Y¹œ»³mÓ G> §2~º”îØÝñqôw®N~ñæT•qühO6-¡:Öc¡oÔ¯Ð}1Pé‰8Éš2„ÞåL˜ŸJ›’Ñܽé…T›ì‚“¼;q"Á])¢ÌrÌ ÌÖjy4Ü0.Eõ£CæóÓÎé#Ëc˜®1Ç¿ÒG—P¸OÇÛÛñTáV˜‹d²OŒ\c.Z­vë%fʦv=qí$™ðÉ¥Ÿo‚@Øvfp€Áté»@¿8ÕM°»…öæîy;>4ó*ô\¢•ð,}“ÛÒH‹š|ž¯ü¿r¾HI"ª·ßž…Ê* à ª_«Ç ƒoéÆúØ ú\mûÝZnÚ§ÔW¯~Vïj6@(D¡¹\òûøÜw)ú› À»Ò_8*Ÿ˜©ÄJîð¥é›èW’Ýñ3è‘žÊ%zš¨Q¬ëõí0Ä¥‹Oè¤m~ù¶N—î“£¹¦¾âlЇW"¡ÏVî_8îkÖ¡p4ÆÂJ´Ð Í°9]à<=Õ4[Ê T^¢ÏªEOfîvc颈Ï-§Çeˆy´Z÷q÷Âß ÚQŽ8æ‚gøWݜݠð@A¨¬$îOˆ*æÖÐ8îWHH‹sMœ³ýÜ^«2ÞK²"ë9LÌÃb æ -ÊÖøп ¤±Ñùq 3êÊøpÞÏ6t 2ï¬Pœ­Ë1|Ⳟ$íwTµ”Ôn¹kYälìÄÒiAîyû‰*€9]«ÈíB¾-¾T\€ãfÐò6k] ÷·ì+žüÈ›láfq&üÊ‚ÊY²2e²`„ÖnŸ÷€|‰,–k1t»ìzÓˆCÓâí«)׆Ãuþ2RpE+¿’¸7"`]GBO?‹œNtû;büºg–9®Ø‰/ ™¾ƒEZg¬÷(Ç ·7-!Eo ëÄ2fÆ„b6FaÇ3Û«¹Òîëvc1RF÷\ æÕ·E„ÎBu§Jx4¥w{vÅþpYk1ò¹¥d-Üõa‘Dö>j÷J”·éh µ¢%a¡ðˆº«õQ g½Ò“¨vP¯9µj­ö÷+´äBŠ#ÔKrØRsC—¶á‹y¶¥%^!¡Põ1œdŸ‰VûOØ Ä¯£ï¯BXÆé} ’DÇôcTW+Ë~ŒeŽ9ã•„²û±T¥"µäñZ`sËÚðÄ.?ÙQ»I#WÃ*a$ÞN£¸ Ú øL#A$gœË\eR1VœJ³‘§ÍéU Ö¢aIÞg°¤ôÅ0 _t72öÙ|Uð '0,æÑåÓè¶ôEDÒ%RM€ðN[ûÖž†ý=ÍA‡;#D'߆ŠµˆfÛSƒÉë1D_Øt©ç;ÆÅg¼{¨´t Ì,û\ûnµ#}C:2̾Ëp²öþˆ¨¨Ôç·õ‡fWµEËU6=NÒuqþ,Ú÷Å…õ&4Ø=%kÐOö˜Ë×ï9vaU: Ðm%ü´_ésÀ¤m6'¯¯èÜÊÛ³ÝVvÉõ!¶»lÓ™ÖwÉÌKmä^sVô`I©q„7ÍÃan½~I!'g˜çš¸ÞÚº˜ õ£n•wÎŒQ¿wœ—zÔžd4ý~‰uÕq?bjàILMGmTHøªl ;ŠóÁ<2Т05ô#Aa–»UOKá˜_{; n_±ÇX èú0\t® ²tÅûë­PQ‹5ìyҚ欦Ìââ—p¶1 qòŽöãö˜¾,‹Ø/‰¾  ëP ©hÃó‚×ÒpÔ™c9ÎÏý¨mùM;—Ùº™69lMô£íæ÷w¶,dÿA@8W\ôÀôêГe…ãIÚ¯pyŽGÎÖªÔ£ˆuÖ5[­Z}"Y¤9u‰PÁ]¶LÚ£B2&!x²²^aHØW¢Ù ÿo ~JƒÃLÍPÚT ê¡i§gð• $—ÚÓxlËt!£ÍQA0³õg0öœj¿êÜÚÕ {'ð˜»E øÈ7휞¿¶ÖˆÓ+€€¸Õß… S¬±ÆðùÆåjÝL ;qn£´[ýþèdjû岨9CðVeoYϯ‰Ê¡¯kaÈêÛÓÈÛäŸÃÒ†ò¸ÏŽe‡à–f-ƒ¼ÂÐa¡‡é5VuÇâ ¦7À~'¬ELF[H5p7&·ý=¨7'_ª±Ðçí”Ê.ÆÀj%™°ÅÊÌÎo˜Á6zä¥Üã0¢r“–%>6F¢ÇR|ŽUrÚ¡ÛR(¸%´“±V–‚:ùªŠ«¡¯fÛ×ú^_ª´]Lâé ~mólyÒ¿ÒZΛ‡Íò9Çç›jŽ«Èt#LëCc¬÷OéqÎ&º—ëý—7:  Ö1L1&DÖ®Åõ/µ¾E`sÕël:Õ°z­hÛû$ÁŸÅ{ÓªdóÛ¶¹£éh Ÿ}% ÂU`Œ=ºÖñE½pÌ°¿x+ëG͸‚&ÖØ¡·øa¡¼l'©t%òná•ÉîŠÉ aÉRÙuV÷üÒ&ÙîëŽ_•‘þ ¥2¸³. î÷ìºwnÎ€Ï œeåàc³4X›lK:è™ÚoFTÝGQìQEæô‹ƒµ¬Ty 7Ÿï£k ".ë"VÇ‚¯bÆçSH/ŽW·!£Î&x§GåéwTÜ{ÝBǘH×Ç©æâFØýº –̾íæÏ"Ìû¨Ó‡ôÃÀž•³ÔCÇ|CKàúºàeõ®™mÅÎ{ôâ .z}àvRjâ=è/COybD+ñ“į=Í-R'Ð[jþtôXà©Ò³‘ É~¨²©Zâ ËÝÆ|®ÇóaÄí÷¿Õ ‘NÜK‡´Óž+ož# êö ÊÕ)&ËÙ»ÙÃ)8M:‚°ˆÛq¯‚{¤ŠÈ"ëwñI%¶Ô–«Ã<•F™õ~‰™öwËSOSBH·*z±Æ§ßäÜz©¡†Æ2xè¹ÃªߤTÈ ~Ç)·‡S󦌠új }ê«ïñV)‘ÕT` QæÔ Ú"Üýlý9„Ö ´&Zι»d®ã3Kj—.\hJëùy{–z¯Sã£Æw‚ÔX»4™kšéjÌAAóFg~˜ö¾ÆÅFJÔ‹_‡¹oa)Š)­›Ü!©fo-u´ü)q‘ng㸆;,ÿ@ÿ!bqîÉwN¥U¡}4¶>ÞI<ü­hR¯ý>ÂIOž3Ò–òÝ(ŸˆÒO3òSî LAüQŸ!/•%5XòŠô«§>Kå ž€©ËÙ¬ùBµnw¯{žW_¥àvùË-qš}÷@ž±]1ë§;Zn{Û˜žÃ?í\û%—(‡JIcÂòðHe|Å ´!e¤“6¤Þݘ¾³dQÉØ’“P|ZÄ\ªXǾ‡å—ùMuK}ø9¾º* BŽŸùõ5Î]¥ŠWÌü³š$›9ã}k©s’3ÆÁšx'dÚº,Ép­½5Rt5zq€oìÔ‰•‹k-Þ]¯q´38ÄÑ»Ž4“µM½™ÖÅk`~Š]16UU±YÏOi$p§o÷H¼ŽtLº¦J(ˆÃH{Û×6švHY›ºËL½­>]ïj[›pYù”’ñá¦>âƒÃ5"¬ên[y)C‚c¤A€`óUÐØæW1Í)©¶˜ôs©É¬ûs¼,X×ÖÞ[$û*êtXw¼ÖŽ†ùXŽlwƒ¦0äI‡;ÜWuÏyè­&Ý®˜¦ºþœŽbŠªü¤9þÉÙòûn;yªb”È8³©$T-¸}´úGä—}}|½W!<ÑÙ»NÄéÍ×g¬q}¶ÛÛ­* 郞T‹ ­V`ø¡M†×'¥ÂÜu‚âÝ0—M5¸ûÙèjø@“Y?ªH„Ò)/¦FOè#' µäœýП.¦ä~Öoú²%>®•8òÀgU=Îl6cy„vÈ‘,Sä\•9â0wé‰ceäíR9©÷j,âtÜâÛ{‚f/X6–…i-j$ }ó­ùÑ¢»Kðâ#/†ËO1³ZôâÞ¡ÓµdG™,KÊ.â\Ñ?ÌfÜãê…òCSB\20úZĵ›K{…ò"³èrtç7Êܽë<9nx"Üf^±Bx]†·¶l•¾xRŠð‹w½Oº÷|³oºåëÞ²x¯\2!½° IÎÉÃ#Ô6e³‡\ER¨Æ 77KÏ`é¥r’DèÚ–Îzÿœa¥’ÂÅ‹ÕlÈ+ìgûÅGªY§›?_S£çÌBxf–m `g÷-9å\˜§d€ 7ô »0%U:"âå‰"kÏ*Leý,¿­b§|Ί1RdW釔Ù‘µIG–¬ëJæbÂ,.DTI?°C¿yÖkË8ôÝæZž©x]Ý•|3ÇR&gÿ ð,.ÌHßPäfô«¿éøL0¿P¼ÀÙ‹º¡a_ +wxpZ\0weªðEœ÷}UPq½µb¥5É iƒŽm’tmMâFLlïæÍ|áòQïBúÕ(¯QÝ%çÍ^a²%W¾®ÑÈlõ×n`C„^{<~—2Õ-‘î§Ê+Sl‘qûæ §c_ÎmæÉiD öTsn°øLœ|‡æ5Ññ…(à׺À§£.©Àíß‘àîˆ!.1ܧzØTãÝ ¿sû¦8ïÆzÝb?É´@Òuºƒ^Er<¿[̽4kµhë4hrzñ;£ccÁ¹õAEŠË|Mà‰›úÖn­•îµóì‡åD?â6@pC×Úi•ëÁ*ñÖ-Ô¾Ó®w=-Â)‘næ®äú¢Ò ¤’n. šßm»&=¿ ð—0VE“À£ ¦xqº—Íý9$óÍ &ü0‡NOÏÈ=²|#moŽ{䉳ÍH@ÆoB¾ œHEç)¸ælÙkêaE†Å–«ošZÿ^¥[±1ð`«+ôš¡sì>K+š6¡p|ÔÕ®ÁñDmžX„ïõg&ÉyÄ°žl!›¯˜f­QÁe<Ì2(°È«w,‡PMxÛûˆþ»q8S®Òeȵ'®ßÖÈÏè$ª #— ÞøU¢odJ*Èf4#>‡2Ï¥ˆá®þéfÌOV…Ù¹aè ßX#¯ÕáP¨ß8>PÕüZ ŸŒ/Û¬Àé\Ýz6È1Ć–z/Œûr;9nâ¢Nμ)u#Ø‚6ÿit ey­1S¹[íhHý¤PœzzQ›Þ2(XZôŠ³n÷so(}Òöyc‡du˜½k#±h›†òW#bm(j9-.´v?¬]Ù³¾¬;K¾ ¤–Ñ—>ÊÇ ÌXú•KmÎ-˜­Å›¦KU:,;qn†‚Ì;L¦êÞt ¸Í`[ToVÀ9iVÉVàÂM*Të*÷‹CQhfàŒÍËØdòÍž£µÜƒâºÍ`<CS"¦è|:¨nAÄð^ÕŠ¤ÿ1ì– d^µšG0‹?•‹¦ð Ó w‰á‰¥x'±’`±Z¤¯é ošÔY0k‡3"g©†bÄvY"´ÎÌh¯KË“ûœ‰,Ä54¨d€#š4ó“c'Üa (^M¥ ¼YLKZÙ…··aS+„  IAÒG´ÙÝs̃“¿˜ê$ pÔ¯ 3"gs ö«âh@·Ç=è\=¬Âèq-Oé?_‹àS2ÜЪv&–YdÓL„­ü>ÚU@g¿××C<-¬ÐŒÀd~99 æ½wêv°‚WA‡KÙŸÅ``œï¨µñ™%cÒ$ºÔ'©Ø>Pe|7[%ŸòÆæÔMÂï.e²ËåPUyð¦ìÛ"FLÝj×$p:bdqû¤ E‘¡©¾i}…õ¯•Û+a)RÅ׿`7W“ùV…Ñfuô+V]£lÇ'ð®·56ÿæ1!B‘ >r ¥!ÇÓƾY,ñØHpåàxîûYän<)ýæ½°s2+ß9!(M™ìXJõ<ð1zÕÁÜ «n@ßÅ'Åû,ñØùßTýÐ94Š5M_Ú÷!²×K.CaXûêyŒ§±.š ß5~õŠÕÔ{1uzpíSAyûÀOõj]éí–*h¬îÊ]kD¸R ŒhÕ èX;Â)´5;–“{åy6ë[Dwô·¤E;Ô•í¸îŒy]\ÊwÈ-(à‘sÐxPnéõŽ8À(½îã{ÛDT}ž{M§U©­ GGt4o,¶˜ó™µïÄÉp·ÉÂçRÛTÜ Œ”‚&Í÷—² ȲS(a€‘ê(à=ÿj€bYbá(1jü&Á£8ÿÙ¥ç­#“nÉÒHåÅhÔmSFÿdyUVf[ÿÀœï epÀ#“¡›®8 C`ƒ˜Jݪdª2—>ƒì‰éƒÂ1ñÝ49ÖCºÔýú=ŠÇðè×½R”ˆ˜7,1! ,ça¬ì<›ÑnAæm&ø†ll΀§ÒÙœ ̱¬åÊ@[g¥@E¥W¹{á®ìï)…vüthͲðú6d±0[Ù»Qº·â™ôŽzÇùXÖd‹®ÖErûƒ½§HžX9ÒwKʽÜæ@`œßµŒ>ñ À‚:.ª_}f;Àcq\pyx1*ÙÏ M“ ïá‹1z$I% }íYï(ÿ’Ì„­ÏO\Œò\1µ™*¹cùòÜÑ$Rü¶ô¸ñ6q |»ùu³Ï5äpUdr/{˜[(ÿH3­ŽP[.û·ØziQÄX¡î/á¹QáŽhÏLa¹{ÚWÓ\ï[œx <Çh(òë燠Ë+ùÞ¯7 _öÏ*¥O•½!x÷]³g !Ï8~|–¢ÍnFœ"¦36éâÐÕ¥î½<.íšÇŸK–f¡|ö¬\GØÖÚ~´fiù˜Ëí»þa“±[ VäË—.i8ÜWòëKvM¶oÖEL®¡*kªü1ÞœJ0S±Ànlz6PÏHƒ}ϯ$ŠENy=°™7}Áº©IË>ù|>)+GgðhClUØ‚8"‡› ¼ª V„ˆüÆ1/ ±jù»ÿeQyXY9’):Þ‹ÛÍI* ¨±Hì£ôcLäæé„ØRVÁš'žÚh¤{ºçò{Ày. »Ž"î‚. ·ô§ž¥AÍœ¶Ç¹vŠ|¥^„µÅüݱPÓü,•c­b±©~t#Ô‡âí· ÏÝüÔ8½ìÆÕé ÷h,½T+ P›6l°ºÕwÖ;O=" >j=Í߯Կö Ò -8‘7“IÕ~ðêx$šëpõé´$£O°;¿…&¥èN ï\‰±h¥L õ›Œð§’ˆ«X2ÍS~b±#=èfw3fÆs $gâö1y¨0`Ž©YYľX8Òyúý¤_;Øü¬¬0¯`ºávD·M ‘îÌW`øŽ¼?¢þó§îk9Tût¯ÀÈY2-*Ë>»Aoðù@.RâGUgIV~‰H_¤ÙPhzð`ëÐRôrÒ]¸ˆÀ$¬ŽsCRÎ G0ËÃÂp¦¢´!ËÑ6XÝ‘yufÍ÷ÔW‚‰µêd&øHÌNqJ±ÅîGgÙ«x!8O7£¸­ß‘£|C·ˆŽê‹¥Æìß´YEÏlJÜDª¦Ö'{z•­cßYa˜¯»äC£ªìÚ¸ÕS¿èAàx¸—~ûGŽ" ³× ›¢’ä{ÉlTÆòˆÅ.å»%±­ªÞÙfíò»g´_Rø©¨¶ ':@’¹|yÙìq7’–¶¸þ¸½â¸NÀ(3 ‹ ¿[õDµK9%ÊBÉí‚ŽÙÁjYØ“zŽ†AÕéà2ŸYÏ>u稃æXõNdod†¯éü¼¶-DïtÓlC²®$Q˜ŸÈ{ ·ƒ´à›l=¹¼/-Ûùàl$ 6¹„— »› ۠جê¯X Ûôù€ìšƒ{$5šM«(§ð “ÊŒ†¥êò<Îjw5}²¼ŽñªÄÚòÌ%ãÉã)ªˆ}Œ‹‡Ž¹ÝxŽ{Á\ÚË«#ò $¹c¸<©À¬xl®< åÒºœÏªJ¥°‹9Y€ìD¨h&TÖ¦#g?̶g0lêÞü o”ˉJÒPÇ–Nì'0eOÓ2»ÞUQ±›?Ÿùß ºiïR0¶Q‡¡;)“ktxÏ™šQ¼Âÿòü=”ÃĤ.¥/Š`d#9u嶃õÕý(¶þŒj$Ê æÞ¥ÿhìz´ã"ÕÄrj‘Æ{vÄÛä— Ž¼ÇIͼÚ:&ÕÐöB4C¼»g†\ Ñ»”±ÝÚôO„§ý/*û‰–’MØدXR|h£C˜Ê-u¤à‹¬4¸µÜ¢vt+GEZÿ+Ï)0ÖºŠ ÷[±“Ì^ßÐc½%ã4$ÕiA­]/7¢Y ÑЦԾÓGÓ•:ñƒ|qK®ÿϾzÍa«ZÊ/e©òa’ uÇaýÅrXM[ïë†ÀkŒ—I˜ã>";–- —+q‘ízz+¹I(ð÷ûû„ï0¬dh9¿sbŠ—i¬ûÉ÷š=§^°’¨Ís÷Ìm‚WÙ¾¡f‘ØÏBnø(yõ0ª¾'˜g_qRÅ+Ðú£î¬ú+  Ùd^¯&©á¥™`ÕoŽ•—jâøµÌ?)„FÐÁ(R;sоäÄ•oÁÕÏZ­µƒ 3u†ižßÞ´\ümqm¦¢¿ÞxGmö¹]Éýœôˆœ·+DÖ‹{´šfãO‹‹,Ï~;1~·ã2[-ß>±˜Ã@7Cya ﻽¯§ËÔˆXúˆQ—`ƒÇÆWØÎ"ä•y™[(†UçÏ+ê3‘ÅÄ"ì²ãöÌ¡ GlÙGm;¼¹í³'+`ÿßKŠyÅ„u¨¿Á¿Dé1ŽËØ)Š±žþ€ú¦©KLwͬ†Yú\CéW„ŽB› jsŠgý·¿Ø/XÎRÙ?U«4ÒmýÅ/O*6Þ/‹ý*þì$Ž› $Ë´á‘É} c<ñÃÔ¹ãËÊ«Úìú›š!7»°Œ…Íà[?QA}yîЯš2¶ø´¢‚3qD<`½þQΙðªd%8æ“wý:ŠÎK’`Jœõ¢YÎò¤ûgÙYÕWQÇlžB‘ [vØ,ŠiØC d^¦¼Ëkî3ÓìFh gÐ0u ^ñ£ ËS×~Brö¡MO¹ë,eŠD„Ô:¸Ý´úþÉ]®g[Öež*áÙF2i‘Ö5{@8ãL¼ÚÓaþŒ€Qc¦•}æªj¶åÜí6ñ3]dšsjâ~æ# QØtí×ÐW·ŸNým¥å¾œpmf¢Ñ³r™Ò?¯ñì'(½ß_¯š60^6¢ý |Ž_¶¨ âQ!«6¤ ‡±,ýñYÕ/ØE_æû;‰ëÌG—Y@øÎ5möWCE>ÏvçÔ®Ò¶:îó«7g3ú¿”}ZÓ*B'¨(œì,½ú›ã øÃâíû¸R/êÒ¾ó«DQìPµ_Z·Ì;¹Ž¶]hëtÕ|¥b/ì+Š£1U6&ï¦î)šOÜ°!w²oØ>!Dš÷kˆ°e½´jÊXbyëóÇ9Í_¦…ÅI’º²r½¥Éôþ÷9©<²ù;‚BO*]A©çÔýKÅob Œt³÷ns1Á8Æ\²¼£F —«ê•–@v.»!ºìE1_I1î- èXvk‡ãh„Ulv¿ÞÙí‡ÙOE\¢K6Ø’cBpÉAîÓ ¦,öدŒÁƒ=qtw*TΨ¥?µÙ•‚_DšJÏ´®cCŒÃSè>{xëe`¸0db ¾±Õnïðš¾4Ñ™àrîGæ‚ÀÓ£R®7j-:eµ®¡ÇN+ï+w s‚mÅÖÀåýÏ°ÓºÃ{:ûUÝD幋‘‘­d2¿·ì&Œ¸X0^$þêô¿v.= WhcÚÉ?‚îƒ 9_5OF¸ËPè€sÒܺïC…'iÊ ¿NˆøuL:¾ZÒsaæyµ+lš–Á„çêjó–åϙܔ{,’·›Š :o"Rp¿ünsá--6æöoÚèIf×)w=5ï¸ïhôQªØÈ©µ$öîGŸk¡þÌ`wùÓMÊ¢áWZh Õ‰‘Í{`隧>©¿—Dò.ÛÀ,~¯ûÄXj•¸ÉS-‹ÀMóW‡°]Ÿü'ü7÷E£˜@‰m;W‹"”œ‚a7Öñ@ÁÏn)éÞæì«Ì¡I{“¼ÓJ´5­ãÄiÙ¥D;Ø轌ßa޻ͬ¤a”ÄÆQŠ;ƒéʵ/‘ô5ÍDu ,'YgH™ýßå¨É’ÌeÚr_D*¨ù#Ç,añ뵃à~–öͦ{y¸Ú‘ýXzÏÃ,³}#ΩóÁ`ƒrt<ýq°†öæÎX9懰6gT¢ѨcýÚuèæ7Š­JÒé-ŒŒ‚9 S†”õèëÓ‰¨‘ÆØÍpÅåÑc‡ì‚â žph™½+Ó7ÇÊó"[+wÙ"ÉóãÌÎÚÐPµç#`&øÅMo5ËùM sP€aÄü·Ð_ÿxáAÑîÕ” Âó³à†(̸åA³iN"Š%IˆîÆ ÛÐbLh£5(ª#œ~5¹¦AÖ™Q0´t‡èí%æ  ãY¤×öäÔ‡Ô«ßµ—oJærDõ·×œ*‰†1`yW¿§]Å=pƒSžyYcrŠ Ï]*ôýÁXîH[šd™ ~žÁVjë 5O×S‹`öžÀ ØÓ¶ù€Ž„†P馦³|ÑéJömhX)½1äs a⣤”C_Á“Ñ‹»4ÙUô¢$B®¸Fæ‘]Uå˜Íš:i¾7ÁÑÞg¨¢Œþ°÷°ÕcÜL‹¿ÈûL¡<>GÌçN•µËnÿ|HlÚs~ƒÔüÑ1ëzé@m{öß i×Ë`BLMÝJN,ÁEÔ•È.Š$ë´p±‹ã° ¡øîƬRr–‰ÑÚNqˆLø€fWóûxŽžŒ¡'n`¥”xìjV+&ý«vA‹Üq°ZN˜ñu?·i•›§¡ô³Õ¯á¢~þI F%ã¦-Zö –ʼnŸ¨skÚlÕSør¡Éݲ‹Éc‘HHò¥tþK¥l¨¡Vr…Çã IDy2z:Ô¢àÏ(ÅPLr mkì+ÒŸW2c5©·YhÞ£f ÝmzÊã C¹M…IÒ»§Ž•Šd⇣†¡ýNþmhmWýÞªÛœå_4­Ÿ:ùèx·ïíÏûdàÓ}擶œëY§ä¢Á iÚ™ž¢íM-ë½Xê,±œûi¹õº¦á÷Å1eOÎ&‘:˼ó+ó;5 ™^…ìãj˜e±¹jtkËÕñºÓU\U-.@0Ž¼UïÐÛȵo¿T„œ=®…“ˆº7‹xnÞ“ìRsÝH»Á‡àò¤¾E¼o!177ÿycø{2M¼v°¨ë@KGš4¿¼(ŒÝ$’ñµ§O ÜÖ 4 ïÔ²¸h rÝ\œå´¡m68†oX¿é.¥ç’‡:Å\hZ÷ÌCgÆ{ú\þ´9q¡Pkäï…Æ e£k-ÐiB³ "eŸ¤W—Öž†'Ýqy¦8¶C 7®NQ2_ĵ‰ñÿNJ\N­0æzdqÇ“ð¾¨ÒRⵯ2±¸Óí{ÿÖw§€Åyoɬ ˜Àhd Ðb.*ûÚ?FM§–É—êãÅÐР\\‰†I}ìß^›±‰Ã=—+ 'ë:wÈ·µ÷æ‘&a”7àˆ‚FÇSœ`mãˆCƒÄ$µ<ƒÍç~šòU Ñ·þ?náM+/†\9,Ú€jµ%kš:À©ŠQlÖÕà/S¤“°æúR÷ÿ…‘#GÔ%2iÅ~¼ÝÙ¨´H8»Í±ùÖ'jâDØB'u¡ìÀùËxˆêÛåö¨¼ÛkáÓ^É=¤:†…¬±ùr{Å|™*]Ñ~r¦[OX Úh€_ø÷©Ñ}óÅ\”Kgu‹¬aaÂ_t«A§q˜>“Ÿ áe$Þ½¥¬‰]NE;ÄÎZªÔJ…©Æj2³*0\óVÿâq¾•ÞV{Ñw Ýá6 YáSÁž¹ù³Fž×¼N_`{‰¡§xø"œb<¸¶båÅ™þP‡\äí¡‡{+m+/ö §lûRËj=º#wU6gÍJ‘A¸\Sýþ(Î~K§ÈÂ0ßwŒ›ú‘7î¥ÞóÁÅÞ|¨´²WAÖ§±ü%­ù‡þ‹OŽ¸*^²@t;¾&£ Û¾R/[dÔö=¶5ÁDA3÷R0«ôƒ±ûØEj 5úZ‡Vr"»k¸@`–-’Ý“‹ôêXD™¨Ë¾ÏnýJãa)¢d~òKq®b»ÊjíÙÓÚúj¢X†AgŠ„l JÏÓP|…Ö÷}]ôKJñ»Ô%5—¦µå·&?cÑå¾7éÝ‚­.w[øÒr ÛßGQ@̓}¡Ú²÷c%›¦_vF–‚b™Ôµ½3oa6ã(ïWø´/¬ÚFP &èü¼ª-×^Ö½s¯pjÀá£9QfP¢DQFCª³5Ù\ìåŽM¹” ye†»ÆÎx‚žf\TF!õy§†Qãú¤-‘Ö"¢Ž#VÓ'%j6 <ÍJ âÄ_‚Mìph3$ªüšÿ’g[ÈÞ Ñž_ dÁ·*Žk·ÏEúð£+öaîÔÜîê1³îƒÀµ‚Rªýš‹[Œ®‹X¾L¾ùàUsÒ g€ƒÿÊ1âg¶¥›Â>ÖKzíT)¦îtOqì «#Ÿ‡œ6  µ§íؘí&Ôæ¶èQ—ͺÄ:u­T&lgEòýÇîÚÜhSó'§\lŠð€h¸ì½Ö§u p)¿^hTëUFàïÃõ,Eà…Óm˦©ÅpÖgŽ' ]®Âè‰À–ò<Ý:•*|š u3`ÀÄë@}3³+¾ „“­%&–\U Ǫô„ºÚƤaW¹ÍM±CøÌsVÞ Ñqóu°(}̺¬R{ ¥ÊÌqË Tå3‚?ƱÀ¯ë`ø JhÉ×Æ ±ÕÅÑ_üΙO2¡*¿TÇ£Üï/;æó.r63Ï—Îèç|8[äueÝ)ÏïÇ–NniVëte éjŽy»s Pòâ±ØÐä+Àí?ïŽÙõa î“vÀ’ý ÏÅÕ*?l¾Zä€X5Ø<Ôôny®|Pzž’{@uÎó$'§Ï0VŸIïg'@û;ÒöpÆÏn 'ˆÅ­Wxù'‘Xçˆa|À>íK æaÕI½2öâÅ`OÑ^Çì8ŒI·¤#bÈ6}¯ -<¯!&ÇáM Jw—M!' ³›÷H–g‹×[×g6±ßV¾–½Âyo_ô*%;íæÐrhæèÙ¾Bà:(@ná.Á%þpŒÎäÏRŒ,-ýG°^\E;ÉÑ;?1ÚV\kÃÉŽ±Öع5rú™¡»sБL«E¸[.®Fk§?6‘:Æ­Þ¹¾Âî›öFbirRê_µë¥j0ä”Û˜ÁbvŠd.x–fp±[VÆ{/ØMAlËÚt¯£,[hÃë nàÙ$oþºÛ¼+X†žÀØ!uý€qˆèöe“‚³fùËœÿ‰ùš1@¡->×Ðß/Pš÷(ײ_Z 5´(pVÒô…ç+üe´(mЂˆêÕLwÍž{Ö[ŸVú{Šiõ¬ªå§üZÅz#¤—E>¯]v)Þ rò$Ü;㹕› …Uú-+“;®š-®E:ËgZûþ•fº sü2ã6Ô#íô•ªKz0Æc²'ÛÏœNõéál:žxq?ŽÅJ€8=Æ@ê=V´ . h$ÕóÀ¡dq—¡‚å À’|·®Ÿ¾[:ÜŒì¾#HÀ$iEa¸<©Û¨%lœ£ÚÐòºˆÐi©áãÌ(Ÿ¼œ‰ f‹¼êØʨ“bû_„Žu;:« gRèþpþˆs-PŒ\qß…iè EÏzó`B ÄùM Ü)åW‡ 5è·§T«ÉÉ<¼ZÈ‘7÷aóˆ|'}¤¤_ävÑì ua±–Zµ¿/EçÊ–óH@UrðJ^´(–Xý>+üï¹Ïãé ±»ÉgàV:CÚ#Q¥´Ý€. Õ*EÇç_À¼¨†Jâ&©á!øhséµ»ï³h „ó¶3ëÐÞØ1Ò?¼âbIÙYö ­›¦²[áªGF_íãë›Áû¡¾!Çå¹ü±Ìæ݇¤S«‡äM.ïÐŒnèÖRÙ ,.Ú„‘‰"·¤Ä‰ïPìè2Ńk¦;"[ìNÒ‡9ZIVw3_K8&KLÈÝð£TËMX,ŸWoÛu¼J q>’ídÉù{§ø¾*æ“kÂÊvḏx¢ZÁFo'q:+â+ɹ.®ÿ9p)Šúî`%š`¶æ̧©Ø5‡äÔÕøÑš€´G>í¿`,5yÏW®s´šÝ'¨9¾ÆÁhÁ¾__y$›7Bev ü#˜\ÿT G\‘ó^ÝYdý‰tRyÀZ{Æ-wW¤;~Ñœ¶Él³s)ÆðWÈUZÝdA5êDkÐë” ÔêŠEwR º¶ôaÐåX†Ì¤n˶o†Ÿ6‘OãÕƒ«!û´$Gg¶Hu€2T ûñß`Œ¢n§K馢i¸.ø‘”=o×ÕyêÖê€ö¾ïƒÖÝ ÛÀD¼ °ˆå[ô‘æ“i¾Vþ|êZf,ÜgSžÂô—ãs¹”jŽ/©=¥È¸8Ä¡ONQwzÖ•bY³ÌÍ`F! ªÈß”ìë Yµ÷ßG!iïÃüÌç\×´`€–b9$ÛÃÙ­Ð7€1.2c§ p»<í^rµNAó‚äFq$WH õæœ&wÆöÊ]¦8âØ+å«q±¶ 2®q«ÖvyéšÎß‚ZF˜%ó䧼¬PÓ±œM©)ûì!·ÁL³¿PÚÚ§?×~}ˆßº±m§A²"ýLäSKl…í0÷Wh@”±ìƒÙFtÂ4É šv†ÏßfI”¢f€•ØuÞüñ„ ýG@r/ çŒ>™l\¸Ÿ _üs±DÓLÍ35 "¤ð7òBK{U—oVå¡Ñ¦cßJËÍ~»ˆð'>·~_Š]¬borª¥ôÎä”Òs >=4Âéý¹ˆd» Z–C@ž€ A*m:é*£q0¬1¨ö&×ÎD(™u«¿VVzU´4©¢N¥æSæ2•¤À”Ó‚üG§6 Th¾¡%꿯f^‹¸„ê¾Ò¸Ã–+«ÆùšbÏ`HVY€àF¸ÁÔ0ÊŒ³+£S­D2¿¸©„}“dn¤,mmf|÷Ô˜ú†òÜKIu‡&_Bfwõ¹|í±F!i_HŠ´‘0|±”ûµÉÜ]©cGˆPàçu·YÅ¢+’øñÔ…ÒÜ™ÕñÝ”$fl⃠Fä¼™g<õu!ýDä-ú(Å[èw[zKl^£/Ñ®h=¥Ó>€I•®Øª”¿^ÝPlÊ%àBJ¬ ¢1‰´ P­úhÀ^‹Ì2Õ¿, sÐóÎÐó–LŸ?÷ßC*õšwQŽÓé\-xŸGF˜+Äëz|+(*¥jG’Ãÿx&ŽžªUíùÑwé“G½ëO2Òïër©Kìšù.ߎ©IÄÀÎoÓ>y¿á æªæVÊ¥åñ¶}^³ª"ñÓ«$MOý.HFùFGAMœä:[%¸ã~Fº›"±s(Òc¾Þ›zú^Q}0þ‰®¡17TöB•ÖV²ÔÚ¹ëRñí„U™ÉÑîø¿;åé°t×òlý‰ÏPüu7Ø0ä=à><@³Ú`Ÿ)à€ÌŸÀÈ O¤GrØ÷#¹ì%ð¾þŨ3" Ê—ú?01ä»k}²ƒN¿(زž^¿î]Þ6x@KÍžSbƒÂ‹^Œ¥cm‡]ß_e`uÑÆ‘j¼pƒfÌ2MfòŽ<>mÎù=­1G.SUuzkørx[wâàÚd0ZªÂkìöBAГñ ÿ•ºÝº•Šø™444\P¿YRL½ùCAc¼­{uù>Á;ùårùé é„k*—¿“—…È«Ÿ‡ó»o–•L_ŽEœ}Ý R¬ ·²l{©ÔBã)'z!<2Té`mp”žCŸZ4Çoæ-(“œª¢;-){$UÀèº!\[„¾YŠêðˆNÕ$Õ{´_:¸5v@}Îlý'Õ‡¼HŠ£`.™H$c‰¯sM‰‘8Eä°ñÁÁ®×tF vò¿$—1QÈߨ¦ÂI DÊùƒ\ªj"}[v€÷ðNQ(-œ¤~=ü·°K„®ñ<´2 Qaû˜zŽçPHpfR3ÊÈ×òUr‰b£ÕG®sü»]ûe¶ªc¨®W–üáW¾y²g•c% ´¾+r»ª½gg`×?û_ž=Ï3.¡$»†ù®x?–kPo t³îó--—VöÄúÈ ÐÍÑ“¶ßÛá(åR1<âï{\è& ‚àFŽÞâÀøV‡§Ýv±7 órõݼÑV|ÌÝ*õ7šXZ°©£?4ne¢¿lNÇ‚¥[óª*öºÙL£±¤†½Gª÷k?á²°]~KÆpf¯7%–ö¢AÑf;ÂŒa ÛÊs7‘HÎÅ$ãJ¯Æ4&E-ÃXÉ|šªs’˜|nl¬ Á¥£mùˆý­?FÞ DÑ(Lï½üÔu ù¬ÄìÑOaK¶ÝÀ5½1LBá%G©’;Û/<ü Ý0 ·þ½û’º“Ÿž?=>¥|¦j;»Äá¬Âù0:1w¨¶öÕ$æ>ãüþ„CDFlÍË'Kò>]}%VÓÃÚ7½G9q_Œòua¥)˜q‚8_Ø — 2/Þ€*:—¤?€ “øæÕnÅ}5GÞwËYOñ$²¡w©½"SN³ìøî ôgt Z²šÈå:šN)s©"Š0—KâÙ%Øf”÷ <* Y–ITnAµn^’6ÕÈ,êÑóf™ŠÍ5æqêõþI× »|-ºW¦8{#Yæt(s{þ­aM·ØãŒivZ<8+Õ]ñ$qMÒðú•ˆ‡îeüoFî¿øƒƒ–è|÷k`áI^™Uî z»„§Óœæ¶ÞªŠ™#þãVQãµÏÁæ>ö­YºX{Á˜oLÈe§d0¢t÷îù1À2ÚÖ)äž]ž”K´K¯BWêÜ>,ǵ}o þ¹ž…fÖçÔ½tñ£ÍóqõC9Ad*ðý š~ƒ‚ùzò)Åß.ú§GñshèaÀ”odLTœ-O7à©oF/Q¹¹Wq,cŠ>*BV¿#¨è“xWžŒ‡bXL`ÆLÉöË [½¥líVJP µ8íÖ²1XÁ¯qD‚ý…ûò`üæû1ñ¿»HZºágé>·!<§rsPeî6„¬Üà ©–©Ù-Ý´\Þ2˜´vÉù´|F<–3}ë!_ÌΙ% äf ›h88a)Å.þÕ³Ç2Q¡ÑwJ,×AYÝ;âþCã;ß ï—åµwµ®sY:—ž $7ÝÀÜön¥CÿæuqzƒqObí AE—gO(Çíé†×H-„kÌæ¾±8t“¶WOÑ<…ñÁ°êr¦4fÝwSÜ6‚} æÙ6L$&•ŸÏ÷"ƒw+› ë»AìPê(Ö@w* 7{¨ŽÐð‹x¥bŸÞNì‘h_¼UZnQ³Nœ`™X9Œ™Ò×Xï õâÖíJ$©€©¾lŽõÙáNŠ?é³+KËO[{ˆÙÈÀ²«fûž|$vh"ØžÑx› èPÃà§C: àUb$?Œ´Œ,ZûÿÔêJPH°ç@ií I“.zFªº³Â >åy]2½¤w¦µ3¡i÷;”WÍ`ÈË…}9ñeµøaœ|ÒìKØ@] D 8¡¬=©;qN}ýhwÀÂÄkï¾.F:vl+À‘ÓF½¨!èàmNÒÛHI'ËåB•øŒH&8aÁ"ëTJ¾Ë£ªÅ¹~!œá¶åοÔDGÿÁfWÞwÁåzÐó@ýµ24‰L~8ƒ*)è{—»kžZ¦“!°{„ä>/f-Y Þ1#Æ$yJ,}™c¹›ÜóaoÄ4.ÛÐï\¶p|[cÉßäí­òÉb¼Òž;ýnúwj87;ö Ĝٸ•Ê9fíÈ+8ŽÖÑŸM¯>Dc°çMm-$0!TL… :bZÆAÖ²ðX~á¿NŽwô,#_ éÝ·Å6L@œŒÚCõª‘#˜\qôamê>²ŠûÙ/GµÁ’w·ßîèl¦ð7P“ŒO^|Û;=‚Ç«n/À¢3×{öRÅèYøSûŠós}"Œ˜ÃwsÍø€‘B[ >yø°)ë¿üÊö#µûù{fãR?+¸Îõ¤ÕyÛÓ²—r·HªI1Ôe'€ø–$#ÓPÙÃ^_G ÏÈÕKEÅÁoä…I<Ç ò×éä„qòy‰Bsoæt)¼\Øk yqÄÑÝœÓ.– .üÝX®·€ÄB Ô[/¬ÄàH}#¢žþ!®]ºr•Y—‘ÜL(´y­–/Ëc.‰¼zkµ…[K›vh—…"§KG®ÕYÕovò¼væÊe¯» **ö°OùbÆ×ÆçàaDíù‚y¼Ìß5K–Ó^)üõâeüL¹ h}™x³]hxÖè[dP9깧<'¤—ýíô¾P­’Z;œW"¾—Ųr<ÎL»`´:FkÛ7/úõS2YõnM`·S,^€DÈäÿ•Ëafk ŸÒD~v w~™N;T꡸¦$Xjk'¦¨7²šT~üVÇÀÞhoc¢b¼(–ƒ@Ê{#eí:CYW5û“©²ëK¸.εøŠ¢¶Û~©±.Ø*ÄLRº¶˜\÷i‡ðêrH ¯Z–¸SÃNò ÇLÓ3áTö,gI7˜fb@".ipoËœ ~î1d,½njé&u×àÕÛƒ¢·ÏsRåBvp­4Œþ2ö“Ø"ÖǺlƒÚû‡Œdµp­zñ$cbéJ“W½k>ʼ‘¥éóoôÅ {]E4ïSÅnã"·²0ùظOà¯.ê[_ ­5+ëKÕB>Ûý2Ù»*Æý'Âjm»!TÍ»fM¹JJ¼õñù@>…´eU¿¨ßܳ*J ,1ßKª ¼P™µ2¹MGTi[a?• ÅÓ·®á0._7îï,¤¶W–}ã°¢:‚pŸùrïÚýw8èi süøx ò\}'5ƒ<ª“‘Ôv¹ $†Ò­ ¶MŠ<ÕÓs5X ½Cã–N2«{´½/…ÛÃv‚=EŽðå@rm+.eçÁWPÛ,êÈç|Q­B7²¬DaºÇÓ1S;m°Å+Ôà/ÇlË…õãs%ä¿xœ5eL–刯júÉÇŠÒĸp«h.I4† >Mä«$_¦dî bU"Qýe´HhŠûfJÊOâr³. $x‹ã‘Ym[=àtÀ^¢¨tÒù)•Æ'N¥#z†ÛäˆEˆïæê6`Ì®òØÓ„_¦ÂfY±¥˜õæ,á**qX®ð€?*ÌH>ËÏ…š"Ó‹é,ÉÉVÂ_èµ?¹Y°ÙŸpbË_/‹qÎÞÿš‡ÚººŠ……‡¢å”ÆÁñV–cÜ7]nç«šgmð¦Û§¿ ܨÂëºhÈ(oãÏÉ?¦½·Ï£1±d'à‡Lve|V±¯Â/ÏòmÔ ›!ƒ÷UÝPW{º«2-߶Ò™õdê.Æc‚…'ݲâcÚfÑë+Ó)q¡¾,îÈ­ÔÔT‰Õ»åL0ÇùÆlï¯)‘hC…}I ÃI³üZ+›Àh endstream endobj 1177 0 obj << /Type /FontDescriptor /FontName /OZLTXQ+Sabon-Bold /Flags 4 /FontBBox [-163 -228 1000 968] /Ascent 750 /CapHeight 710 /Descent -222 /ItalicAngle 0 /StemV 93 /XHeight 448 /CharSet (/A/B/C/E/F/G/H/I/K/L/M/P/S/T/U/V/X/a/b/c/d/e/eight/f/five/four/g/h/i/k/l/m/n/o/one/p/period/quoteright/r/s/seven/six/t/three/two/u/v/w/x/y) /FontFile 1176 0 R >> endobj 1178 0 obj << /Length1 2421 /Length2 28567 /Length3 0 /Length 29575 /Filter /FlateDecode >> stream xÚ¬wtdݶn'«bÛ¶;阧bÛ¶m§ƒŽ­Žm›ÛöKÿçûî9ïÜ7.Þ¨5öškîoÍoâÛ»H åi ­õbÖV´Œt \E=}k+ZkK=+lj†”Tبç`jm%¢çäˆÙ™dôìŒ F6..V&#'燣ŠŒ£½žñ‡3;+3€…™•“† dš›:èYX¸€VÖ†@C€¥ž‰©ðn¨§oØ[98ëÙFÖv;kCGS+c€ƒ ð†« ÐHÏh°05ZÙ8X\­¦ökW;Sc‡+…å_áþ¢PtµwZÚÓ-, |ì?Ѐö@;' !ÝGl÷ÿÛÉ×Çý9Ühçà °6úG€äGðv6ÖvzŽÒ³2üÀ2uø{LÖvö4ŒÔ\VÖý?ìþÆhHp´ÿókhjoc¡çúçÒÒÚÐÔÈôãêæÃl`aý‡×{;=+{# ÝÇÒÙÔÁÄÚÑá¯À€.6v@{{€³©ƒÐ  góî¤gñoÒýçÓýiñÇ`¨ç ÷ÿ%匀¯¦¬?îŠÿÉý¨é_ù¶wÔ·754Õ³3þC)þ¹Åò/Šðï€?Ðþô?ÿWRÿòoÉÿ§þì_3ÿGÞnÍׂzÑ›~ôÔŸc>Ž4ZêÙ™ÿÕÿ‰‘ñ#hƒ?,M­`èÿ °¤•‘õŸ™üËnèhó÷=' ýÇì(þ÷D‘ñÃñãó÷+­µ‹;-#+€–‰‰ãCh?˜q²°xþƒ§ãGõ­þ’à^_™~$tÀ,Í[PI ø´¹ZŽÈM#8dj¥+xX-_i̺#Mðê‰Çv[9)˜hYP©J¼¨üd›ÓÑW\('À ‰šÎõæü­(¾ïÖK4ê¶xchœ–‘ǹEDaHèo{ãØw_¢õp”4W»¼ÕÁFFÈä² p¬¢vXÚѸ莥‘ršá i÷w²¨&N½¼õ& =t›£™Ôž¢¥äIч<Âæ&Óµ Â¼¯ BRGÛ·´¸×6õñÖù¨Ã±Žg‚ p ­o¥O UGª{†GÏo'<}liÊC•[qmVØ0Ð]ŽÅ) š=~y ½\ݶ¼û$Ù•;¯h4Õ…kß@:öJœ®<¯š?¶±ÖØ™2„•ÖPÖ®ÝYÎŽžò^»ò7™÷ob‹æág¸_tŠƒ‚a’ê¸QÙ¦Š«œ?- @;бËÓÞª‚! ¾Cœ$•i%æR z» é)O‹Sð¸óÐy–ùœ,x%ìq†“¿p—Ê™ô*ݨ¬Î³f‘E4GÛ2ö%ê×ádâÓ²„X“ð£S^Jsð4³DŒÊÚÊ Ð¿ÒìŽá–'K[Ø1ç4WÚÐj¦BörÚÖr"côÉÞQŽ¡¿VîgK.Ý©/äHÔzíì;½Cw¸òÏKæñÖ9/Kny³Ú¤òÆŠlÙߘö‚öÇU¬ ¸þ¶?B  M²à½ç•„ÀÖÞÚs—¿nåCú¬™J"@×€dq*}c¯°76ÖU<¯A—ÝJG°ñ¡ÕAÁIÞë ¾¼r“ú*p˜Ã¬ºž[ž{©VìñaëÕ8þÂÃÐæ ¾…î›^Bv:h2ó4*þU®[m’ ã@ë³T'‹*w„Š¥*nðˆŸ*b§}”ÎB%ß¡¨.(èä‚w—”±€«}h&øÜ!–¤ðXþ{¯¢d§´èç╱Òüb©ôÀvçGWÂqÅí|ÒGzHú•Ù¿*{ÎöÃ9øzFRçÂÅOB%IC–oÃiÃrÉZ¼:©qòÝ›»Ù# êÛr”âL¦Ý[öLi<·¾4íÉÈâøç °¬¢o>ÝU ™!MM“`6¯Xý/f«[7K`ÝÒC1ž½;HÀ´G¨(n\€™Nö8¡'kgQG5·Õf,!£•1(ð%ºòéî\7–㺱^à¬ð7fQ†ÒJéÙ¾Tß^Qº°è ô¦4Ÿë ÊX¼7°<Ô”ì[îÊÔÙ}UÎõ…`QF/¸^dß8ŸÃISþ¯Ø3·õm=ñ|(»ýçfï„LúuÄ€YO"éËÒý9ÆÖ®²²ž­å~L)óg}ÌAW'N nÎ^LÇo‰ÆkÇÇ1#Âðb½Ds÷º²x !| œ8JÊ r”e6±ó¡ôKÜPf°íHŽÛ½áìQ¥<'q | [·èô“Ǻì•]Q·bþHY ãøõiõQ·(ót%ybNc +*<íá¨i„™ü_ôÎÙ٤ȯ\ŸOs9_xž–ýÛ8^Gh…ã_ÙY‹yì˜J¨|Kƒ‡¤íÛňÝÇÛŠ…XC ;‚û¼2XÏü«OÐ#~Qú^~JeR”Çñ;9âˆF£´aW·>9åî…˜b\“2©©Ã|.RÅaOòS¡¦ô{º“ÏãÞÖtJ*mâÏ(@^vÖ?`ȳ–´ÆØ ½HÓ éU8Ø«cT>œlø#ž˜{ÍÝóìÝÁÏhmÕõ1è~u©äiP÷¼þ`[„õñmO’vC•n9ö®æŽ– -ð®%K “6è~d%“èêúN ]Z$TO§£4*‹³Ä$÷wî5~~C «G :Nž7­©¹n•ÏSu¦Óñš`ø› qx-¤ÑvÅi‰TobãŽêNÉÓã¼IS‘7÷7«/5ãKÞ¤ÝÂbR nð>(«½ãÔˆ!?}¶ôÚ‡\³Y„e–À<4ý֥ߖ©Eu5rûÖKòk ë5%JW¨‡e_¥÷9»•Yàqî< TqóòPýXh¾§‡€é³åV\A¶†™M•²ñçN›Ã5©UŽ…Œª!äf³OÃ7¬Pç»ÆL(ÜMS— ó»úîî'œ¯€`VØh“D‡óˆe^>û %@,©+b’^då÷0C/4×ÉzVDhÎÿN 7@Jÿ)a89¼aµÊ-W–õ%W˜J<]DzôKp¼’­¡à¼bYö;?BX)î£Ûæø7øê1¹t%ÆëV6a3NëþüO¦—©--»pºZöÍ\A9 Þ¼öäp¿ LP¤Œ3$À@ž Ä…LîÆS¸÷ã,_FüH,TÚÇ¡Eû¹Ö Ó\‰»gºg¨·Ê ܨZßìN%ˆ·¾Ð¥ìLO[¾êj u#€¿:Áwh±¿>²à^˜U­:Fª²ï*/¦ñõŸ‰Ån}“‘dÅH…Û¼=>ðâBe(õ¹³–z{`wÞQ3[` ì祗Ÿ—ÆIýšua‹8!"Å+˜?Š +$ø F‡á¬ðÊD*FÆøÚm€}Æsžg Æž<Š{º;Ûæ†ïT!-øDûláo5 ÆGŠKTÞTû¡²s>»×5ùí··óÐô~w¬OêÖ9‚¶¦øåfÁr®w+{¨£¶ë&ož(½TäËá·ÿį3ÈZäè\ã¸$ëìé[÷Bî„™ôþ,2tB¹%*#Œ;þcªa.´“”ü±=šæÌ÷ A×ÛY䙬u¬í»h ‘V›•ÒàïîÜõ±Ý‚ô9×É™†©7ŠŽ i° 1vOwñ˜Èf×°Òq  ƒ³dÚËõöãM°ÔôÆEn½ê´˜ŽÓæš‘ÜöÜ\bïÏ ßè5…+©{”d_è8f«€˜_ÊIBª¥“"ü¤”*/¥ðZ¿>cÅ\Ú¢Üâ”ÿy?$¦ŠéÚõKW_²Âj]Bé8%+ƒîOÛ¹Zu£Pý‘XõKøä-é3AøNcáf*5ª•INzŽX^Nšæâä'ÙXRÿÐ褷ª“æð0ˆSë¦@õ¯/`=;n&<(`;É7ù¡zÂ.”ï¦T¥úe¯ ¦ùÚÉÈihb›yI¶‡"”ÌÇìP˜¯”¨¸;!A†Z…~»øçå—8’ƒXñŸ¢I³§…ŸM¾9GVþê,L p ‡Ï†1 ëð”¸Q& ÿ.µ}>(ðík@TFÿõ·5…MœdÞnšù-ÞÜ AêKÆìKãWö%Ù½VžPñ Y»êšT]Ä(YõpYŒá¢Œ*ê='«Ã¬ÕøioœñC ÓŸñöœ˜‹>Ó÷KÅ×Ïîc·Ä[ kÈŸx†ˆ<¢9СƒøÇr®‘»X§oÜÈÛíÙP_"–šÆYNžÝèqy§&î›Å¿c¶Å–( eå ê@fç`-ÔϾ2DÙ ë¯(F{UCJT:aþBÌëJçç¯xÔ¤Ž½wdeH&†zŠqgk;Zø\¥æÅE¤w{Pç2 PëæB³Ú^+zyòòæCqÝÔÒH!£a\žó´?R¨zD•¬êzõdý5½,®C%´Îˆ õ´«Žœ×¦%J¨ß÷PMU³8–z¨Z¸4v+j˼P»%@„JŠš•_ESœFAY‰×)5ð¢R;´j¢ó9oIdŠ=ÐS€ Âõ[ƒ…W…!„G¶0G¸Wê|‚—Õü@íýž“ýѼ~ÅÒüŠ?Z«ò^£ï`Z¶ ¨:ÿlé–ešÕêõ$6Áøé aF6»ªrPÞaÀŽãMчPÎgNkÞŽš ×ß’9(pz/h-{Bêæ„bèýétzaéK?ö»w£· ¦] žqb¿´*Ç͉¬`9Ù#ηD㺠é°L9ÉôûÎU¥ÀÊWH>¬þ=;[¦OÛC#®Ýž52±"goÐ PGsAÞ Ìm’06êmvÙd”ƒqM–i!?”ÕXIƒ-Â(ˆw¼\*¾º¤õ§Q%…"ãGO™yt2ÝÙþ]°l»‚HÁz«¸+D؉Ó^ù´ÒÅ]ÑP´ý¾‰n+¼Û£O}—Ó}*ãëú©W(&|KÊá©j¡ãY#¦l×ÉŽG‘CÔØ+5vûëRÂþI  AL-ĬÊ7f†$$Ýa â÷ÐJôKëÚ3õÇi]]1èx×ææ1m¿xþê]+3±¢›€'Z¬O(Ö}üSâaö¡6Ñjc䜀#h“–‰VܤH ž·]hŠ¨žcSß ¾Ë>™W¾…d vô´R¨µ—Ûm“|bì)ä0„O™Úá9sߨ‹žnV¬eL‘²ÕK()c’G)„­ÏEvÐvzoÞ?Á¢KúáÍ;ë[:Ý ßRYt[AòÅì‡Ú,&CÂ*Ùz6‚˜u6ÿÑËP;ž¹¶F•¤/˜–;%°- Ô‚rêqzõP’ГeEJt•J%ÉÈ#º~š~r(‡Ô÷Á„ ¤óë;å«­€XNÎóÞÈGŸ;"…(b=•m3¶¬³FyˆwÆ0ŒÏ^ú[¹Ÿè~YSÇ1¡5¤­+¾­÷¿íÝ娧ÝâD·,ìê‘”À —À®ÃñU`jBn…—ÄÀÕ‹Ô`†s1eævÙ„¥U"þØ»ówK`¬ vØã-#›ÆØ|óçG7Ê›4Sܨ)Š+qÃ@ð ¿ ͽÆâ*¢>1Ë7{š_µÿ¥¸ó³˜º°òVZ*×½ªùì½+*-#ÝÙOÒr˺‘–Í·ö³ÏÈí@¯½*„añ6=!lšWÈ+Ö$¤£¤Ÿµ¼KíG|ZKâÒ#D6°î|1 úp»7¬nVÇx&¹tÆ §îŸKø͇¢çÒ…¬–…YƒHðÆ‹9µ ÓFU°0Jçîð–b#ת)¼Ûxë]ŒK'=?7Û8Œºּ׌ùFõ‰?çå2ùzû¹-ÐœsƒºÐmî ßÿVZÝ×ÒŠ‰p,@û‰)yOôþú=`øYa0“l7¢Y˜£s»Ê£¡QFŸ‚HÕÃB]¹Ë ŨÊGt\¼QïÜ0P¼¨ýé¤0× ÛmsɱÖà`çsâ;âWµ+‹õUñY~)„*"LéÀÃœß.T ø€²pT TåAÃòDšÏÄG§‡ÓÈ”DÀ/®7IêÂgwX&3'ûþghÛµÃiémσ˜N%@òë®õëþòÝsT–‡×›¼Ö‚kî·Š¯•þGIö%£r1^Míø”#qp¦jFÉ_ÛT>?„åšÜÒvŽ‘¯áÄ©ß­ç$ ùÖ‰Ä?§·Ú*Ÿ™Æcu‘PÎéãu 2.w7òÔÈO˜ìf[×ϸú‡ÕÓämM%ƒ¿¶™60Úý^±Š% ÌÏN¹BéÊÖì¶å)³‹[ˆ¬œóúñµ` ä·lÎÁÜ£¿Q„bˆ祱|&¾ß†• ’H ›ßMb’Ûé¿K)Ç],°Ð;Ëx"ÒHü!9[3VÜ3ΕG>v‘vŒø]¯yØÐZ|íT ›kÓ‘Tä+M$¾:vâZÒR8¸¹4zyÅUç±!®ü¶Îª•ŒÞq²Fk¤#rVLÖ™®“gáÅŽåK@,ˆÃ°Àøa©/úÚ“#Bõá쑉ÖæWÍA ©r,™ÄŽpDËÚO߶Äøš$óJÞáé]g[¶î/»2äˆÃ¶­¿ãWÌšô÷¦;•BÌ 2´"IÌ5Ò;.Àžï5øP@ÔÞÂ'PI_àJÉT†ñ A›õàQ Õ›z«~éQžA$IJÀb™óz'\œÖ)À}ûLdaÂ~Í™µ–TßEZÈ)Á¥ùd´ñzSØH²ø­’¢Â&Îßo ÎãÔ!˜±¸ktÚ«;ˆ+³¿•ß98¡P­Ìt@EÐ#ƒ^ò‘†‰fzUt"¼ip|µm³jZj…B(Êþ”û’Àm0ïÇvc_é4=Pœ±Æ IÕ¿­{/ª!¬šwAMð»ÓŠ"¶êÄ´wŸÁ.‹‰#^Ñ©©äfuVA%²E=k^ið tÀ|•ß”lfÂ$Hé Ü„wÅ´«î¯Ô¥–üÄæ%×|Å1¡'ú2‡M dÿ{p9F(YoÞ8’P7çB°\<1À;Š¤*'”y6ÉÇþnNÀˆâ9ƒ×ñŒJÒÍÇë(Úv&’µ+•z‹¯ø0î­^:é}ð„PÁi¥P)Æs뢄 ¯Ó«9Œiú€´ñW$3 Ú‡uçéX„·8oÊà¹1ˆ> ®Z¸ž(Üĵ‘ƒZbÍ…² «7–žx ´FŠÁ6êéi ¡/Úð*§h#¬5áÉ×Êý#ئ{c»ÐY²ü@…lý¡n+äê¾ÍèÑôÄ }É»*‹hÔ˜#rÞ /4ˆ¡eá¸KÑÙO®×/"F|ú¤ëÙqÓâñn0P³N(ûµ…»uBü¢õ©Á?8ºŽz/TGeºÍu½„¡hAN ¹ô«å$Þ½ƒúÄèR@1°ØŠÑwÔChh– ß{nR—!B7Œ òܳÏzAo0(§Ös¾ýEg÷†SäecEÍ NGª·×R™V,iª®hB¬WAÔ6iI­öÕÍ"èïn0u½ä7?ó\íÞÀfåfÔRÐïGú]ä×Jζœ_VÉ‚'1ˆo‚tÛ~;çVž‘§»µ×¢Ø+YH>Éœ3ü¦Ô ð%7ì‰U8$*\…sP{2ÿ»h%‡iÒ½…ns°:S ¨0‡M´…uŒú|&l&éºbrð‹“ñª ícFôRþ}‘,Z–pU0ΰ ë\à¶èOJ|°i_Õf’#‹mEjÜÞÒUJüÛ¬õX2HIid|ú¢aö”#ä²(P5¥ÆLµ¨¾„’õÔ$0 ,ÄO9pun ®Jˆò¾_˜–ûiü­ ç·òwùPù›–-e@S !LÞ¦G󆲗ÌúOO-g+FºkÏÏ⃨ö´gð;rt„@–ƒ›µñ!$pd|£Â·2£êí¥£ÈJôdL¨rû"•*›•àI1ó ;güYc­bÍf3ÜCÇNÈõ¼$3ƒf%BY_ꨀÌ9æ«ØgI@Û!{Ü»É<>rB)ˆÞ]JMÆü”¸ 3åÚ½îlá³®C˜>²`8aÞ `,6.÷DZ–U²Gè"¨ˆ ”‡Æd¿|ýƒžõïº;ÿ*\¶D$+×s”Xîþ$û匔R³¤ËÆ×ý†JRžì– Fñ’ÎÕ ï"Úg9Ú+ mTË­ÏïCÙ£á6{ à J‰ñÐ%¿ç²ˆ pºQ'FÞšç^~³b†Ã‘¤Ûˆ<ô´pâ> )‚ú"¨ DÄVmÇŽñàËEc3<gÊÁA.Âóæ‚/¾°s2Ö´ò^2¦˜~áH´ŒRQM;½<ã&Ýú’›Å¢°(Ùµr±ÛðŒÖ0ïzÃåYÍ 5O‡p2#*èß× £Ã¡PÔº‡¿JT à‚¢½Ø|;¥^t#vöEÈFç>’ZU :ôÌ… G†.»éö:˜wµž½ãn4æD®ä÷PÇ9c—†"¢$Ù¼Û¥´ÙJEvôqbÚѱmjuòŒõõ]ˆ¤§¤’9{åG]»'úv/²}|c}¬’æÕl¥P(ÅÃbº>6MbÖ~NçŸEbŸzHT<Ö¿Ålÿn·éÈ@^Œo-ÁaiýÞ`vZô³ñpÞE}Äxm¦ãpúU1ßý.¢~]iÅK1‡ò*ò1^+Â]¾TrÝÅëdüð7Î —ÄKîíéµPL@ANY»'Ç…á÷ÑitÁÇOë_‹„9Kæ¿A Ü~’õ4“Ðc³7›†sg°ìš­Í–èÏikàó$…VÏ~tïtDo†>»Ÿ„^땾þЗ«3ýÁˆ!¡£¨¶¼bï[ $Lt˜{Ö±Wê·V´^Ó˜G:]À†×Aýƒnf“uû:³ ‚AühBÆÐ’ÕCÈî[ZKg†¼¬0…Å£ŒRó/¢£[¹Ìç˜áLx3=><"™æ φÄoÈßr ¦AÍîÑ}B6ÜÑ`*5kµ_béP=Ö)˜¤box1"˜o©ú™\>ùáUDLË hO³ÈÒÄ¥M”ïH>ú_1råšqeXt÷OMJ*@B8ôÀ—|“[ßØ{È lB+Õ­!(7MSPŒÞfP&?$¾é2ý ¶lwÑYŸ²$ª“mtL™­æ%ç@Mw8 èº=ÉÁLè6¤{-5«g+8ñKAãÂʼnÂ˼l4Ç„bW@¾u¡#­W•‘×–NÝzçŽÉEDÎr²~šgv/4^¦0ñ™€hš¬o¬ LÏ—Y€öü{œè‰­zö8®ßo¥Ãþ†jÛð%«‘lwR;PøG$j&%oçTþC}iXÍ4âññ/ÅŸ[§H¸T¥^ûu'0 ½!´¸í2l¡ì”Z+˜¹Îô‰¬9’8e'ÅÐ(Cte5ýnc{7[†íÄ^‹MàÙ\էў+ÀØsÏFÞÙÔ¬»eUÕz3[×602wÑ&Òåƒq(ŸvÒJ/~)6=£| >L⽕[lb¦t¾8 ÃTç®2’'`žìSB½Ð‡»f$]A±X'a“oUt£õ3• ×ökv ;„%Q*9ÇJ¬æƒµóbСNZÞ™_”e·DQ$®,z4àè¾ø÷±Ç^—Õ¨ÈÀDo ¬¡~¥½h0Á‹²"Pkãâ!`rviðÈ%ðJ í©£˜ì±Í”þàt`—™¼Ujš”'®É>UîÙǠòÃùøG`QL~…ë‹H™i…O¥ù¯®ÕŽÍË3$]¨èÔtÙЪNJo±K•¢ô[âØîvÄee[^F ªIDÕ$Q±îhÊS¡jH{er!ûtr!‹ó“I2âMÚ¼=Þ²ìø¯+u„ºÝ“:¿ýï ÑA KÂà¦2i1=U8).Ë°žndzªq÷ú?ZBÊ喙ؤö4ÛP*©‰‘²aÔ59ö~ÓuË¿7f¦ƒÙsC<‹Œ[* ZwtѬŽ¥V’äp(‡ïHV¼Ûq¸ZWvò‚FŒåt=nƒäNç¯^øxå쨼ˎ $ˆrhø½Yó‰®KŠüœpCa"r¿Ù_ùK¸Öô¼7Ú—Yñ;0|¾NbM)õ KÂ0»±Q…œØýWŸæÒ;áP»I-uãT8øýÃeÉçEð F- UWኒ¼~€êýK¢``„&z1ô²õsu«:–„~N†âô8¸Y†Ó­ºÃÜwÿâß!cy²+‰û2žè!çÜw]fC{¿îX7‡4hvÆÉŸ2å†{çë´%"cãÃ2&Oå‚ö Ô–<‰x\¼öÞýNœ4,7} •°%³AÑ ¢÷žšëjœ¥ì ÇG4ÌOiáµ&¦Ÿ‰4:€£Ts^äéüÚ)¦^„¡ü‡íš¨5ËB*ë'9ótÏUú’uÎ?Зõª?–F™ÆŽ ÍØ÷ߨ8k±-QIq¼´ÕƒJ`¨ÒÁjsDåa8¼è±²2M4ûŸ{›Ê¥Æ±$T³çóã ÓëkUµx0úÉhq§„Eˆû ¿¦ÿl·vK¤ß”/½|熛™ê)QÁÃâKs†“ˆ<ê'gK.ùɺŸ˜w1°õ{Ï€†ì—Ïn+ ;D2kœý¢ù/ªÆPÃJ7Ò>?Š žÎnù1yd^ËúÌWSyô4¿Na38ª`“Ë1†/¥Æß– —eÓÌ€öºC5Ç‹)/HªA¥Èr¹Å…ý(Ã"Eœ´Q®xëxƒ_¤IúMÃQD ŽJŽí·ÇÔŠ0jŒf´<‹ešQU ,~yU¸ø¯Jº¼ôöÏ]Éf²I>ßdzS?¿aÅ“Ï°Bbd~íÂä[tÏ\[–YÉæu˜s 2ÁÊ-V\!d|úÄ/hõF¼¤­*ê.G0ñ‡[@SrÞ^@y¬½Tú§?ÍFcý£…~ÿôœ hô!%qßtñó×àIõ 1—¹X,¤¶¼ 3×jâ3S,öª•Ý)ŸÚK˜öÐÛeÑ´±ÓÄV…†z•Ç@õï ïëJB.±Œ?ìè‚ iËà2¨+¦0Þ»QåãP)…~T‰O¨¦RaáìZ“¼ÔÙüÝÑ·r|Lž¾ÞaÐÝÍMòx¬2Ç /D|&Fýl¨hœ&¨3·J '¬’_L‡v´m]8-`«Ï ±['¶n.ဦ_[¡ôì.Áõ§m [Æ:_5Eè´”œ¢Ž>$IèG$/yÙÜB)TyŒ¹©>Œ”‹ø+ž¯æ8B}O,‡cK€ªû­žhÔ ä4¬aKt}Yñ{,Õkl‡LÎåZxùòÎñƒDÑàÈK iÑñeUˆõq"byqMô·RìwOâÇcŠp6Žäm˜÷ŒØÅ¿k‹{7ñ:R\Rð=”¤º™‹¦û« g…Ÿwô7Å}yTüJJ{a Ct~õ–݆»eÛ?ªÏ"Óº}›%RQP`IÀ>mð‚ ä‹õûìdàõªCݵ¸‘LÜ°òöËo¡kå{]Þ•™AÆ]OÐÞ¢§Q„€Ì¨½Æ¦bˆaµòç®PGîhöa—væž¹~£iácçKÔcüÔùÖxó²A˜¡Ós1 $I`ŸEÜ÷ÉWÂrºõ ªXÑ—Wê\œk7b·¤š•)çuLµ$Žù=¦+;;Ns(k5-žˆØá&ÿ“Šjƒ¤­ áh1^Y ±€!N韌üۣƷ6SÔ-ˆ-óó•n6d¾^bžM+mJÝ7ƒÊqƒ¼9•Œ àÆC1N·Ç(üšv“7Ö«”“‘ 8í|„ð-´GV­Cc3Xd…YTиé¡zè.{Oˆ»ú ‰iy´QB+À0dŒ1Ñ£WoN„íjb”[VÇ+fX¹K éez»Ñ™{VµLdS©ópï2¹nòi¼<›£D “¨fÆ:Ë“Z ûÖDfÇhR=ÚwÇ”ùFÏÚ|‚ z°Û{A —pÕ¦‰ñ[?8ëž"oå8"§#mEx“ØL³µÃ,×ý»ZÎÏŒ‘5 oejÂPÞMÛ¬!\N¸…?÷nádÃçA¿)“Ù¸ÂÏŸQ¾•523Ï×Kâx~#ß^8%GÉ8K­žK"šÇò“ÀGc¬ª§¿ÐYTK{ _ûaÛ è{/šb™Ô˜—^™y%yk²R\Kõ¨Cadƒ@¬úpº9Q]*I$ˆ‡7É*e âZÑ7¹”s&mæs_õDíZæ´¨vý‡;ÈõY¦VµùD¦ÃºtmzçWÅOÝìøü™žÝ-©eß–œ廉ž‡¨ø9~û认¤]ßrLÀ{ê" î:éC*Mÿxn»œ¨+TF„ ¥Û:±g´;-ÕÒ1ýÎÇ[yìÙ‹Ùp òî·ðW €Úå¯Lwte-ƒQ¢x‘Ð4àõÉ­‚(8)½…¼ó[ZUu^®wÃÿ> ýéòòó˜÷È£F#ÌWÜpj•Ls&1Ùñ؉…&\ÆtŒ§OYÜ¿þÂÒȵnç"óp_æPU-õuôÖ¬·æÎ!‡Ü´Šs›ÇŸ¸T0ûù³¶CôâjxìÎ"+¸ƒdfbUUzKžnqŸêóýñ×’ ³"íènƒíÏ\ Ôš0£f{ªtÑb¶øqp:AÐé­aGö†ç+«‹ü0‚ŸÏ–‡š™ü¼B;ÃrºžßÆ aª±T/°f»évpò½Ó×ü+w…NOâäƒ6Y'§»&þP#›sá4´ž[YΘaÁ¤G ‰ˆaF³vð+§¿ ’c¿ä \ µXJ²hµQ¿­<«Ë"ŽêSW囎‚qÈWqU¶T^*^åWáÔU{¢óõ½ÕWº÷D©×jU íÛ¯ÀÅ »J=â;ØÞÀäc&õ¯S)MÏôV¸DÎÀÚökq©G;c=YÄ5ÃEsÆI4žÁ3Í©L]¸]m¡Â™^Æå…tºç7¨¢­ðMBYœP†ï>í<-h’KR¥¼dïIùŠŒ§‚*  —¸fkÕ§$Ðçžú¥ßqs|†{·ÿ«u7!å„Ãt“I["B=ñj81•=N1GTrž_63‹™~/WötlÐ|›ÁÖþ¦XöÙëJ¼'¶ˆÅÞº¬è®®È®áœ oaíµºáøÔ7>’U(DçœRœ¢Ë¶C>k/ *ÏÆ*£&"dM¬èÄ)¥àµÎä;åBi_ê1’¹~Œšº¼mv&„Œ£ÇÉ­–³Š˜XU,ìÏMûW"¢×‹”ð÷x~¢1£˜7Íï{OñÂ]"EdŸ¤Ý:UÛëh‹/a8ñE–ZìâP~ñó¾‰yYOÕSâÜÅ Xè1Ñ cê´§¹½ÐFYóÌ) Æù “ç¿ç,‘™÷ÞÿÒ.Ø sJB¿î¯âO[ŒæžòÝ!î»^ rœ†¼ Y†Õ…4ÒAL‡sÖˆó¦;ÕÑ•Ù%ûa>¬ÓÏE\>ÉêÒŸ I¿ O©1F„H9?;O™éM@ÐÖõåù³«Œo”5¿5;µ­ç>MÍ" ‰ÉaiúK!±Þ:BÕÊØ.åžl„éLhÔ¾¿ëµáˆ ÝøæM¿îHó¥[Éäb~nߥà}3½C-RÐËüLcZea´« Þ¢N15t#nýª)S¾Ák•+ç®—÷U°±Mjheì=Du˜QÁY^YM¨•†O0|&¾örؤoïä²Íž¿©î%%Ö*yâ·âc©f$&§š(|¤üT‡X®ÇôÈé±=¶ƒÖÙ©3Ê]½¶.;f³fVrUi>$±¼wAåùÎM¤L+-#uæ"útKJlEl–8ŠxOêä´‚…‰/"3Ÿ¶@Æ%ûŠ’¢ØBÞN]-%½+’!{ÍÒf¾Þ7Q Ç»ÐûÅ­³²‹yµ¤Š®d ~G§"v·S„'ŒÝ·4‰uFxødYÄ1ÙW½ƒ/„ÕÕß…~ÔÅŸß+žKÐ_Pÿ”Ž5ÊHy¨’;Ú£wäàEiíek›ä­ûÙSÿ-8<@‰7 Ž8­)J&™7-²3¢Ä+6áð o5eË'¯§ieóÐý#^”uW~|Ð R?]‰ÿ=õkW+¹÷Žn=Ú_}”\Wt:óËìxƒÅíHé_£±Þˆ’:‚3D¦¦ß„¼4PÝñrͬÜg¬†“räS É5C{_ã¢N‚ Há¤Êªqë-ƒ?b,»´aúnjªÙP€õr*’²SgÒ‹c3£ÙfKq,K߉Ñ÷#)iÍSð¹Z3ÔëÚ‰Û'¼+ØŸ˜Ekû‰,´öŸ:m 7VƒÉS1ÐðýÅ&«Øì±?uýFvš•É%¼!(¸ïð…Ëï?OÌ ;U#B³ë¥BÛJðú¦ãåuÀA[ìï-°q´:R”_íè'®ÿñEdÕzñøŽC¶®X»òí®ç‹^^OÛ!ÿiH*~b=j’€†ôq9h ?éônr$øáX°ŠLx[ÏZWí»Àü¢?tžAÐnökÈï’z3Th:Z°L[cæÏÒª½ÀR"®ØcŠê|ñÂuñ¾wr}ª‹½è´U"mnê`ȚΦˆ· ìZ¹l™ÄƬEÚ#:0ˆÙ\VmùhÄÞWW°–±gz.–òOG~Á¦äŒ®¼Ý?â'77_˜ü +ÜÊ@,´4‡ã4ØÏvB^´E@ó¦:Bï*¿ƒä…±‡˜=žŸÆ<¿iÖQÙÊÙ{»gG#À~ÑpÍ=¨Há º:w¼®jqç¬9KпJ|&"ù|»ÅfŒBéãrªüéRîno‡ #U²\tð›…hÜé>6%ùÅj¼+σ[ž!g`ˆðÂaŠf™)$ R¨YW„„é¬fðÜ9^Ô¢|Ùì鋈ÓÿLù&ó½ÿ,@í-ï­N]³rËÆgõúÞ ûcÃEÅ„=x#M%×çÐï³2j‰T™ïì"„—{°ÀÂ¥ßf§y¦:Q7õ¹W±ÇɦÓïaH(ˆÛœõB^Èg•8<ÿ*Xápãäb©‰Oª%¬Ñð ±<9X—ï%CÎEÕÄò ¬ëÜöFJ¨Œß9å å‚âƒÈ'ªNæ}ŠA>‹ï 1´]^læê8C¬@¹àÂÏv-g¿(Y…f„æ!¯û†Qk©T€É3; Þ+JUD©£2†ð^'}5dÖÊÐ ‹–? ÐP•ëæWs¬‘ížîë¦q· kÏgÄè‡äO¿°3ΕpúúEâ3=ß¾ä‰qîç´ŸEi/gå5P˜ÛÓZZï 2ÈÎÓí¤‹e ø]¥Õ]½#A)Æeha¡Ì;„šÔ;öñÁÒéßG ͸‡Pª/{ùA)C7 6¹©úMÌ™Gµ|‡¿¨’Ðc@@Sf™’.€×Ød"w½ yéø÷ σD€íé`úá¥SÞ} dåú†;š_.òPF¦§æ£·gÞ]üzXUó€ÆÞoržW´x2›ÙRØB‹då-GÂNV´Ä¹®íÊÞ’¾Ìƒ$Aê/û©]‘o"ŸAE ÇUö|ͬݿ^ Ó.×G3yn’(é„~èò§Éäa×8VIx¨ôÆëgmi¯È—|Ð,CJÖÀÕÁGèLÉM]"ÑÚ¬ìˆ`ð:XšÌÔ€îº]="_Xáÿç/´WXWO×î–S ?A¿JŽØ8là>™¼ì‚HE‰,®¹U%ü¶ƒ¾AǧÐeŒúª°—cq>þÝ…ø·Ê°õäå¡|ÅÁ­œh*½‰Èo”Šµ‰ºûâ O|íÅ>¥ ÌŒ±æÙ†HiÀN·ÝNðÐê÷„y{waņ¶fsŽ#¡ ‰ªH™^HƒH¦z¸®K¿Arݳþ^§ˆãÕÞ„šªä:+·ïxåèIµ,‚†`4OûQdŸ½Dð¥,\zýéU^ð§`Ÿ8Å9< °k/½´SáØ»¯QqðžoÜÊét¢o0~L8ÆÕkØX¹Ç~z‰Rÿt²È~I®ÕÓKh‰AƒWâ~Ãe*rŒO^n#ÉÖ ,h4dCÒÓ5Lñôa/;\Êçbq¾‰ÕD¾ÈžÊ…×|õ)>áÒ¬Õ)_ÞD«¸#[ið(°®ô'jÖoZ»ˆwºT"‡ÉKF‹®1q ƒÒMçÜ«%‚˜¥çЂn0wÃáŒ$%ç¨(S,§úŠA`û%O°FÿÕˆò½8ìó~¨ÏŒjáv`ƒ%™TU¨Iõ&‰e@¨PlgB¼âÛMH»0Èú4ˆÛ“DàK8>J£Ç9A¬#®¥æÙ¿ø‹wæm«[Ž=åoó]0Økï‘øµ}ÃìuSðg¢„†DŸG£zbšýY†}ª÷ý ('“ç>ÆÕ¶iæÈ_÷Xüöî¾JŽ3*·èEª!#è•IàÃZ¯2• ú« sxB‹1‡ôAÑ3òÑ éä6øò vZ’FΓȄßæ«ÚÒz£X\E¤lÝ l·o¸{Üê[ÆþžÒ@ûoÉQûl‹>C«†éZé9kZ+cD+AltVÔ ë ešãš’ý§Â9 æ~ŽŸ.›÷YÑ/¶:¬?;yŠp„Dãˆo^ÐÝu8šÆ±Ï`[¿ÖE¤äáò2˜`/¦¬Wšúq¡¢æ….XþUû4¿²}ôº$«¢ÑS‘WçLŽï À[RÑiÌFh‰‰êbÈGÒ4ó Ôû·’/_þí4wuň±ò2L¶9 TV×­›È/t[ºÙ[ý#ˆSÂI×d-kÂ=wÅh'Ÿ×~¸ðæ|èc´”ÒÌ0‰LH1¦ Œœïz)üPæ7¿N«@¢˜é¨1¢@ªòp2ûîgØX­¢õ€èÎÚ]òD¤š£¶½{¥øñŽU}.ß¾Ÿ1xk°”ó(×ø dXGíÎý¢?›÷¦n ïÙh„òe;WÆƒË Û·¶ºmËÝYØxÊš—Öº¾´‡4;#ëb ZW²bîoUGcìŸA¯{Õš¤—؇Lд l@2‘ pLÈhH[€4¤fÌÀ·ÖU¦¨¯îD¸µÞ ŠÎ)'ÆúK®ÆÓ!®ú¯alWÂÈì=ÛR¼ß³œ¨ò²¥Á¤NŠØîw³å‚¥²Ä­z/7UuN&†ÜDqU7ôÙ›ã¿M#”D* wd9Ñ›–`#ä)™uH¤S:öWòËØMëüc˜¢[À:ØHìF=7ˆG¡À;né&îÙb†~wk8¯}R5[9h\Ãý&ôüð ÜOY– !¨’b]xÖɧu¨HŠ»u>k[î½þpÍH/°­Z\Ò^†ç=³žHƒÜþdlíã•áŒ1ÇJ—&6Ô’r¹ïGi–Sòpìß»·`ÁëKïÎ¥àP“¶æ—È{Ö³S1J ù;J®Fÿ+2æÍoQ!ZT d$»<Ž˜OôÄ-uëqäYFÏ‹¼À4¾ÏÀ§óŠX&ô ž'ؼ{5EüY+u"Žp~àXº,>û9“&çüLÕlÎ͛ȭó1Ö4Ìîq¬±<ÌðšÇóè )ˆ[Ø™<ÙItúÃAhÇØ6y W&d‚dMžV§söFúq|jœùùm‡à6¦®Î¦SG~­Ásóÿ¿Y d@Ťâ¯oK½YúÅ/”RŠ.]îñ;ÊÈðREÄ¥‹¦ÈÄ°‘ö/v¼ ÙÓ ˜¾#w[xسq£™šŠöU°uõ¢]ô—Íy‚&(¶­€nm3û°HÕ£ 'œö´´|}ÚUJ®Ú0œ©ldÀvKÙË웺‚t KÜ|ä&ÔÊ’üp‹/­—w0[:ªTßZ‰À«:7²xnböc’{Ïì^vX¡F¢Û"@”™(d§€Vý!{cã.Ÿn«Ô–Xþ ·ÄK‰ÖO”•8Ü×g´à C8AèIêK7·¡EÓèÅæÐ(ûs…òâyU“dš*”õøã )hé¤ÿR#žÅÖ}—–";¼kÚzGRxå}á_“z³úû‹]Í¢/ <s ´šm;Or£u3™›Õû®¬—R eMqÄÌÏâàûâî‡TÐT]kį+öãjT`–L—OÛO='H” °S»Ù«‰›²ùÜ_¹q9-Kƒ{2kÊÔLßB«ý2þ¥ÃötÎcã } ¤J„ádè1É’Ñk»³qÕq½l€p;e’h"ÈéJ0mho¸vÄÊí÷Bzt$o€ ˜Q5qcË(*›2ÚÃE–óÒ4J»o½q0·”†AL’¿O”IuqDØNä…ÿ˜_Ì¢õ÷ë$O¶—a&¡™ z6^w~/ÿo‹ÐÝÚ$ù¡Ô õ¢ÛIä÷Ò“lš Æ_@€>›¾ÔÃ6-¡˜#ìtã*)ŠÊ/ÄaC"œpäÓ‡«ë¡rÖÒ„r`«Dn]h÷ûW¤¹Í-‚‚È#u´{ßÓÞÉh999F÷”ÆARvTÑÉååÒsP¬ùáù—Et)§ó@{…¥ï‚7Ês!ùö±j ÍT¿µMsÅI­fÓì“0K?‡ÆÀ÷åþöÊþ0V:Ø™uC…àK8{µœ¤Êlvƒ×ä(u»D 7,éR˜ÞñãH`4¼§tEWœ;7Î>ðѾ)“ˆÿ: «Æ¾0Ñø‚¦v‚ _ävkñ¿õ•Ïçv<ÁZºÿ‰VXeBJb8”åø@(æ‚Á§3`¶Á ¶gØã¥* œ{!5ž'êg9RÜ×¼ŠALu®HPÆýÖ¹ âG‚HÏ!˜măÌ=FŒAæ=hÎÉ2û pÙÖvhBT½(×'l÷ET šc°ôZ/¿s xuIïÜé1ÈCe‚ µ‰¬ÐþAcX*£|ÖhÞy ¶HGäÐFg]£/£bã(JRóh‘•œç•q® 9ùaL”H»P™¬gSïZœÌ¹™Ó}"£™¯_dI“ Ø°Íè+ ÙRÐAµòÜà˜é ¬qÕŽ så‡Í§à?ƒ…`„%œ4›e ÕÇ4Ç[-„˜K#eÇ®g~ûë°ßo± `ª°^çMŸµ\É}Œp^uÝ*K‘nÂXu%]¢.Ÿ©øút^<>’ÚQGŽî ͽ6uâ²(Ë™K·ûµ(Å¿‚bƒ/;'ÉC&A)Ÿg`zFÓÀ†p¢Ÿm ‘ÚÝ_boTFQóæÅK¥­/Ü}‰8´(¢ö½>Ÿ2¢³2.üa í@ZZØQ•{¢!]ÉØñÚðÞ‚6é¾ÎÁM›'.)Œþ SÐ@»R 戦白 eú&SHÐF6ö^ >~­48oë^_²̹D»zó.VÆÞ3iãY)£`ý‰]ÈÏfœÃ¡[ë×âôäÞ›, MÉ5É˼àå¾mèx2=ºìIæÇKÓFm¯¶Ä¼$F±œfeCÕo€ŸÖûÆH@‡ÿx ©ÉÄsë2½ Á°iöÙ3ù[dßjoŒ×¾pqË×øMÀä #Æë¯ã¨ÿÝe’aÂì†k‚ñ\òÿ—çÐt…+³͘Ýhø¾'“˜Ûx|2ÂrArã0Ã}u¾Aoꊉö‘ÃL¦X´N£&ú,/ì÷Þ›šˆ¦ål¢—¢ÕÊðÍXѾÄè ½eÅÐhˆ±ép;Ç6iÕBQ¶xÖº•‰‡¦¹pFß\Ñë€è0=¯ÁÇûru0ªÍðq>QõÄÎè×C§HxdÌÒXf>]#S±¨îõ˜}›©$/§Ã¿êp‰m¿€Y定yc™ìÒáàŠf&Ú@´^K²o°•ÅGÿOyÌíŸdäÜH_#A%mç‘Æ€{hZdÏÛEãMPJƒÝG²É};¿Ó·]RœÚ¿[<æïÊÁ‘zƒfžê6Åf–»“#Ȅ똳i2î”* ±pôujyÓ°>Ä:ȨMаR×;-UJã ÔŸ*;ç± pÄyW)â¼üáïâ? efÇ·’}™I;P Uº_KÝ<ˆÃªÊ㼸šíÕÞ°2óæÑ'£ñƒÕ0yn}ﬣÐ;’¼(¤·Ú>UPÊUí¸+Ûöb÷º¹ž£Š¸¯èé7²~ƒØe™Oå€MºY)Cû ]`eÅ(a,gøŸœôÂi 1 ëÜ™QîÄýɲM²áögü(7¦WŸz¡uU¬¿xæÉßûy‘„zI6MÌ“‚£T¿3‡œ¹þïÌÿ¥ÈVÛñ£³¨ÅÁ9¸¯A:ü@Z­X†ýVź ø¾Â•„²–`w¸™²¢qZVê ) M’e!ñfñ¿ýJ=zŒ«EÄÞ˜£GöåE÷Ù±hL’˜ `67)¦=A Éi(œÉÞÊhà¡/Àlv”/O§:ºJ$}Fðy¢‹ãñ¬)r/ÍlÃþ*âÓF+c$›Ä9ª·Ð‚|ÉD}©0©iùØ §ËGúœqB¼°6è Ž£|áS°V±ÑRjé·ìé° á£+âàÀ‡»ö„#ØÁB?Õd<%‚ HUsjý òʼ¶êt;Eü^:ÖSû~tåÖ«ù©‚DBÆÆk÷»$âväKÇéO…ïv‹;^ãÀ= ¯~§µñ oÏ•vŽ}ô¸Æapº‚oý%ã¹­ô_i?3‚c‚^Ë´Ð'ߌ¥Ýs[‹&Þ×ÁÿÞ'L÷‡Êj¡ÒŒY aÍŽVõ{àÞÖÏU¡BÉàá›Ëþ?"+×û|®¡ZHbÈ ÚÌ3-#Ý<ûИáWÛÀì⸒èÏD3u–}9bž`G„XQ¥SÕiWh®•ïªp¤Ú'+Ç•îï:t»’•w°äd‹M1v¡Â²/}wàk@KMnÕ&o!Ë1_Š–Å´‰Ãƒ‚;7w¤žo¹¨J¸Ï˯uùCüóiôÚE¾ ÄOr™ˆ‘´Ê³­øÃÀé\Öè6þ¿†û•sá¨X쪫ÿâZ|»Ä„¨ûÛaP™­§Ó¤FÚû™î—™ÉàµÌÂ蹶D€­‚›­ÀQ¾ÏÞ©ì¾4¯Úõ{ íØ¥lÒ¯˜Ñ¦,Úç'‡Üôè·Â³k“2 ªA—«©w§JNWjJ¨öÊfаήNù’ëKŽº€G’Z8XEŸhB’•møŽ/&°ÕýYŸ%þ4è¢eV1çìߦúLy :ªWFÕ´Vü Dð‡z‡7m­š –ñJD§_íD¬Þ‡òzÿ³>F{s¯b 8žœ9¨b'õ²ïΕ­Pu•^{§¡¹¸™_¡Ê8Hþ‡gn*GQìëßᘱYÜñ=&¤AËc÷õ¢ý=…(¢ˆ¥J²?}M˜»ÚŒL×7-™<<íÎƱфnGp¹)o&8°näšCþ¥Ãùð‚BÑ]šÅ熼—¨'?c^6ô§^;UÇ'Bðÿdóc:6–³ ‚xg3…G}Â®é £2R é6o<ýóï…F|VÏ; 'oƒPó•°øTHV‡¶>;Bf¨®´w¡ý»äõ¹e×O.¼ªè“Ð*̢ŋ±Ù;±^Aå°á»²å©ÇßµÊ*˹ŠÊÊÌ=gðÕË/7¦©- N„§Þ·>nøáC˶_)Q¦Ðš9w x‰å8|ÉìdÒasj®fâlvíûÆW‡EÒýáñl)°ïõ›Í‚£RF”lwo¥ïSv"®n=û1 ã(Œ"üCe ×únŠA_Š»Þé²ç¾ÔÝáZ·™©l˜§dªŸR@‚Y5·¯C™?‡ˆ‰"‰s^,vsO O¶}¯J¹3…©‡ ™Y͹ K=ã¼>¢ ÜÖ“2@zRÇ-íë£éÞÍS¹LÑ._‹Š FF®†ëµgÇË,Nu¼Ø½?¹šÑ·»n¦“Á/J†ä&{ƒcCf¢ì¼ºŸõ­øtìGðò+PÀÍÒS˜ôx]bw×îU'{÷ò1LÕ™FÑ÷]0ú%>¦+úY r°ý­5n³90±V‹ŸL‘è4Üy‹ðzP‰ÏÁ'Þ#Ú³~tóÛÛK;¹Æ¤«£ûåï¼Ô¦‹:8K<ëû¨ªNöàß;´ôæC´Iú¡""×®Ý/gÀÓTÆÂ*¹8*„©ÒLa3q9w/eæ‘ÞI>z•ˆÝEõØÙò¶,¶§~ìm{øõ¯±7ÒÇ#ÄÍWQO!Ðé}í€L\n‚½r’£`ë"CoÙ ‰‡`:• ˜híÍØdÝèv K|Ÿ†øŒGjÓaâ‡%Þ»}á÷í62©öÓ €eÖ܇:CÎÛÊós¡\m¦x›áîèú}ÜN ÚÍYBe—¤{c²cvÙšiô÷»KZׄ+}V¹‰hû#<¯ ä/•6h-‘L"V;ÂTxñ®ºyàj"‡ø’ùfÀŸ9eR?îO­äûÙ*ûðnƙڪË'ζÏ™ÇE7DÇÃ=ú÷C–ÌmÄX²Ã'Fpýº’™iýÊÊ™Aæ㙲í¯ž6Ì™D™RA§ë9L™jçBÁ‚q…˜Ü^&¡õ¬p»›yüf‹Yi^·„™ñùT}>­{¡[óhÇ=OJ lÏ5hšËÁÓE5;ª¦ØsdPˆ$Q‰Ôö9G¨÷ñ³8QBY ŒL”±î‡tÖ[Óê;ÆÛ 4œ³"Ú>±Ì}seãoò貟’;ÀšCÏm‰<°¢wÔhý@K¬è$ÁeìYW¶­“š·½*½X`‹€Õˆ‘D!jÎÆ~ñåz [*ÑäÎóôSö›ˆ[×&~/Þ ÍóRÊ@½ØÙÄíz­³€Çn ¾«¡¾*Uóp‚s_±YŒ ]çô‰(¦°™‡þetïk±s ‘¾¿£ÔW¯<²§¶=ãð uV:¨®l l5¯ÁÀIãt2uLbDˆƒõÖÈ{4r xçx&´ íÂî¬?~@Òc®bÅH0ç}dPÔ^$ÑöÝ€‰Œ‰è9,¸ÁÙ#í$® š„R!G\Må¥E»29[7ó?‘ â*ÍVÝïBPáñÊ$FG°Z†Šë´T…Ì*4ijc´GÂCõrÛüæÎÁéÿô¢©fªö£7¬U{uºe(xÇ”¤TlLÖöâR([†¸ÑµŽ„ÊQµ‘ÐÖ˜tfÏ‘S Ú©éµA'|Ïǹ§vÏ°<'[ˆ®Ý¯­3èkQ×7$ˆ;w8dÀ÷ÂÙB¥… uçÁ$ÿý±¹O=ïs˜O5⤈oÌýw®ýOp–«Øó;Þs¯ðÎòˆ‰I|gÿ(‚o®¶^wçˆûÚèiGÜV¹ùj\C9½ÅJ5Ö½ª±ÓN‹e°2Ì{ǬÐ÷ržÑE_i…mÚ‰šŒI‚ä ­ä ËL½»Ó )â„ }Ç|pšøŒrYå(àøá tcô÷Ûºó]&˜£5t17;Gñ0|½ê·ÆÂ&Ló»S"ŒÓìjÊë½y¬^qv¾ûš.r¨ VNFO³è—“­BIõm’¯Ààjd¬Ieˆ°ïïMÁaê=v“¡(ÂHXǺUž˜ÁùÜs`÷!æhy*eÝ&vü<6‚â+=Ì8O“ñà1β¹‚¼™ƒr àGr[ðÀ7Þ—õG WÙRÅä>DÿÓúÜ‚¶¬p¬ç³I9¤ü]Âÿ¡[ýŽŸüHHÏ…ªiäû GÈÌa,²úôR×àt‘ðá‰íÖÑÞæGWóôÔ)ÖeÇS·e3úJµ0›þvR(Y‡Ús—´Íòm~·ƒ£@^Dzç ìåÀÙû3O(Í5ºÇNáX AÒO GÊ.k`ÿ‹åtcär5Ë$S>"u´ úˆ­=¹W¯Ýô”– cC;äEàǨ)Nógô“ÀMÉôÈ­͉1ÿ&Ú}±¬ŒÎ^ mëz+ÿ ù•9C?r\>ÚWKèÜp5ØcRB[@ EóÞ‚Ý ñExÛo­3Åz©73â]ü£˜Æ‹Ëµ‹HI/ž”HLA…ëþæ춊„qêû›|Ô.\æ½i옱v\þÅX”3JM;pDÆ‘RB€¤~tjÄ„ y°ö¹`1žõ:&ÍÔagjÏëøõCVì‰NÏ}‰)&Š~ó ‘Ö †Ý3¤ tû&÷áë—in0{Ê™Z»G¯‡ð2–+JY" ž;—åžÝ_9ttÉR…îæɧÎR2į5yØ?]2`H.ÅïF >}ô|ΉDD_ÀNßM˜üD‡GÙ“’Ÿè2Oí-¾ÔgzmÔûƒº³.ŸÊ¤™Ž¦öØ€Á‘åΙ¢J‚ @‰ÑH%Ž;闠‰Î]Ýzsõátçò&ãj7ú*ëe°¯‡bC½gyœ/Y)ü’ÉpbëX*"±·D—FÞîzÖöÚAuµÎÒ]ÿ¿K¯Î zKä|þ|ª¥ -Åü;K£„“`Ëì S€MQQÁÊ ?zâÝr âšL kwMŽ?>“6àt´:¥Ì |5x!e’ZFß#âQžH¥Š…áyö‚ÔyJM #(3÷Ó‡ÿVéÚкÅÇç5£ª«DçQúW±TâÇ£ØR‰ƒ‡DÔ=¢Xõ÷‚a¡‡µkÔJµ RÊïÇ—ôV/y'˜é³dWí¹¯‚hƒ°¹bõ–êâ<šËß‹F¬ƒaw¶|ý&:ql]¹ô0, nÞv:ˆz4¿`oDôÊ5çºsÚœ¹:G: ºlýhz¾ Z±T°I;†ÝS\™®Ãd:O¾´…DžÎv‡ª«AŒ‡¡z CõµÇ-í×ã’ÿ±Œs«ì y§êæõÀnÒËším¢Ë}¤Y¶v+T²^ñú gŽtíˆ[øsäíãNµÇ|0k+ñåµÈƒcÞadÅK •ÕÔw\XyÖ4ój¨¨8³Sö|Ô¼9¼»˜ž÷È*Š@wÉW‘ËÄó!Ru ’€öECÅÜ,ª7ÌþŒÀKwó¸A’!šœŸl&ÜjÕ¾xñþ¯‘Ÿ$ïCÜU¦*®†ÏÓèç9g^éYO&ª³úH+RO«gk=ÔääYú×R™ª[˜P{kd4¥#¡$ªLÉͧ»Ðò°ÚÃ[Pÿm½Õ3Š|{¹Ã>õô±/úæºL¿ík˜l:ß®D‚Ÿy¨Rî¨æw•šµ[£:Q“†žžöéÎkKžu¹ yö-¥¯×›Ð+|ª\Œdó Æ(Zø+t¡|K?>Nú÷!놯¦˜1µ¿ìûèLáv “¿N2í²Ê‘@uZ*¯ BFèÀ’`š¥X@>lƒÔ»“y·t$ð°ÄFO¸·ê~}¨âñ#(*”†ï0D}Ûz›Œ9ÆÒkn½8ZÛ³™º_ç˜2DÜT+´Î ñõE¼· .Ñt ÔWE{^Tgoì¸Àß>B÷Žt_¨SèØ< C»^ön‚äâq¶ +r¢.醱-A|âG7Ú‘ü°zÝ”¸bïÓÛc¿µ²É‘ãÑîR³6RPÞþÚʉûÎïž1{òÃ×ÌѸe4‹Ëk 9K1G6?]|’G¨cžæÚÉ,IDí Õ(ù?TÉ$¨‡?ZyÂÚ4ü#¯CVww°·¿Òxò5VÆ{&=¤8—Ä¢¯<",ñ–@! m¯ƒt3‘"À%Ø£Åù×£«„˜; [OÕ-.êô¥fè3ˆÎ¸³ t¾YUÙJÁ÷/Ó@”Qx·ÄgD|jI8 _5PŸ]©oÄb«X&C¥Uˆ{Ú‹ÓŸ•.]Å Iã:B§›_>€µÜ&ηÜÁ:P!ÔÈÍîx¾8ø:×›–1ŸÖ— 6Úˆéï~O=MV9.¤+và(6ÿÄů¸A©7J/4/š”RÇžã6.†°¡¼¶¼Ìv¹–ääê7¹£? ð¨sZ™¤æ·æ)U ])èÈÕÝÝkáÆyžÑx:7ëzË¥4ÚêH2‰QyÓ­Iæ`züÔu%ø,ËÉsßþº³û›â$×.Üõ²Þù®ˆ±¸$¤l•N|Îj$õ]hQZÜ^‹;xu-æ4vþfO8XgÆ_Ùä‘Ï×±f^ÊË^Ó‰±ú#ÉþvªMÚûtëRœy%ámnp×ôpGaHÄÿOˆ?´ùMlלëb®c¡wKJ°`˜”ƒUHÚÀÈç+ÿãð¦ËúÿÃfFýMI¼'ð;`´^Ž§¸¶ÕlÅ}Ùtí]¶Ÿ‹@«ÛndaíD Ÿ¤›®88FC¤5C~ëì»wcùž`ùv¥_Ǹ$÷ÍÚR´8ýœwK[ctá0Ð.ÈVþP7§úSCö¤¼méôÌè§NFX7Ÿ Òº$aå4Ca³ï§£ã§˜{J­Nnšneñ¨+WF ™ÊüõVC[X•üÞŒ!\– äØ®ôaÐåXŒâº"ê9bˆ„,–)XçXD*v^Ð"£ííSHD©»N^(hRnsÆó#í=‚´Š*JüU•4ïCv§´h+  r×ÎËçCš†!­¬š5º ¡~²ÓiÕL8¢ÿŠ#áüb@Vj8áÉ ÑJ ýšgïCgòu÷]óä³û±¿&5/é{߸gp+ºJÃéO oˆë­í¬l^-ñˆ­Æ#aZÚ_d67TI‡PÕ̆†¿b4`•Ê#ÏËÛ`¿±iâÔüjP.ÈŒõ…´¨w HËÎ#`AÕA/»Ïo¹2“‘©ˆÃq%yæbÚçPÚªRøãË89þpÓ/-lM8f¨0øì´@ äÆAº€¥ÿôn©Ê|iƒ¢¡Í2½ÆÕÒ Ù͹Úyµ§_ï&P?LÈœË*Á%óEƒÁJW@U_@©4E×A±•m3®I@AL?§z=ÈñÓ0)àÝÈ0‹¸Ëu*:tnÀÍÕkÍ´€ƒÉ¾SýÔèd›pÿÓJÃÔ46òÆû÷Ó½¨”ayáw¡ò10{L¬ÎÅWMoÑ„?Âų+Án+¡ÄK kã»0èÑ…²¶­”ÿùñ–…×já:étw¨¤ªÍÁ ºzŸQ'þYëbg=ØLJ-GÉÓé1j0~9™ˆ)ÚeÙ¡"9d•Ó$â՜ևµócOõ¬Â+Åü­(/–¿Ål–thwZr™w¡yw¢%0Ž¼–Td{1¢‹Ç׿\Dbq¾.ý3tÄó8Y%v«Ùó²ÂxfÖn2oYŸ—zàÙ ÎIœ‚T£V«¸Ë\7£QF4ÿ…åœ÷hm\zò‰7‹•pnûUw’͉>°jÕæFÊúð¼ìL oœ}•säQŽí¿¸+owþ¨(ÙS;.¨‚ |¸^ & ×™)›¡1n¾‘ÑÂĪt¾ïÐǹAQ½pån¾œ3 Ø&¡) â6M¾ÅhØ3J4€‚C(Þ5͈q“kFP»&÷Öú2Ú) k‹¯æ0C‘K® ³h×X$ã€+[y¶ÕY£dj.ú^æ z4§k(¿~0èGëDl—Ã…%¸²~Ä_ˆ”«=EçŠ &Æ!›Ÿ`›ë$økïK q¯j›Ž¿Ð&¦2¹]@ Y$""8A”I­ É>1Q^¾(Ι‰Ø…X°Œ˜N=ì%$0BDNÌ13I*7@ &ÉÏÛ/Àé»+ú0S{gòÓX‹D7¡%4ÒÑà~ý° ñoÑS ÇÈ™\ÔÑc66/,“P;¬+YÄ]ÊG»_Á@àºì‚eKlôij[݆¿šóÈIk'!š¢ìTz¯êÖ­¹óÅÏ£$Ù²Ì,Ç„n}: jÐþ2biÁù%ƒÚïÒ>-'˜Œ(„Ìç·pß¿¼¤'m†Òõû%Ü ²$̆ŸAf€èÏ#t„nq²Emü ¡Ñ;ÁˆÜ `ò~Ñ÷i5Ø”w `ušCø:j©ðs­wL¤Ø/Žªºivô‘‰þýo›2š°àaþ&:y¿_´©­¥B 2Ä_•W/¥ª»£%g¦´بE:lÉÿÊ }0*ʳ•í/ç2z '>#ª+0@ñoëþî…•ùˆÞ·LŒb4AEÝì|iV,¶Ò>zƒÜ T]·ÈImš!¾³š™-0ÐñOcÊß"¶'ó »ÿÐL.ú‡×¡§ªèrE¡,K• Q€i¯tÌe)6…ò¶EÞ¶ö¥!™_‡! ËÆßߊ‡‡¤‹¹\HxO Aúú•èæÛ\שÆsXîóôæŽT+ÚúZF›âªmÈ [CÀÙù€ô"{À]ê;„*O4?Ñæ‘À» ×Òšéœ=,¶¼’©Ç{&êKá L&Žš0ÍPÃÆÑ=\pÀâXײWs‹<ÕKk£¬Y]›Ã”õô®ïpÛÒÚDW=Ð&}–­´ ¼þ?’vûI8ýE@è=|3×£e=² #²ÊÞ)jvJ¡é»‰”Q5ϵw˜Ç1€5}!kFTÏùµÍfïmŒˆ¶xU«].1BRԪОÃ0ú«û—XÛ­›ŸÎžº¥ïœWG­„* Ÿ–®4¾ß\G0q4Í„Q5?Iôró‹'öÏn®h´¶ÅIXÉÇÁãž½uÚíÏ7…;î ÷„¦‚q«Ö&½îL<(K½ð†–UfZ9P¬- F,ò]¦4p&ê-e,EÅè“uG²J2LòÛ¤ËÑq‰v;„Vž½»™´¿¥Í|ÅŒoDf[”µûHNWÀ,Ü ßbüʼD)©9ºc0˜«š~?ãO´›³Å›„©jܱJG½} ì¥5½] H…”‚êjÓäõvYïàÔ ÜÆ‹‘ã;Àkäß—÷½:¢°p8i‡¨;ûZ‘4„„ÿ‘ÛÈègéõPZ-¦øÔySÁ±Ÿ_ îàk´ÄJ†À•yWÝ¢v£’…|ÿ Iôˆ–($ëØ++~œ£ïý`M•­Êë{í"¸ À#&´H]½¡BcÒDuµ]΋€‘å9ƽ] <±ƒGš’—ü |í1œ"®UA–vbGD,˜V~ÏfÞTBƒaÙÈìáXuÈ+³aÀõ´a¢¹"?n¸ªœ¢§ai2áÉL¿?„¯»M4`xváˆc. 'ýäƽ©ÉöY“E„˜M=€¨ÐNù·BÀ‘Ù´‡ ÞÜ¿:Ýw=ø¤ÕÜ.O×P—º'P[‰ßmNDp„ˆrܱRsªõã‹œ“žÐ–ÛWü@ùMBmuÎÙùÿø\„ZwóÖúÉ4û¦,™2_j'„¥ý¢ÃL!“v‡çª&‹ÅZ9À ¦±ìýüv¼­Q„AÒ*+RŽ&…Õ®Àzÿ@D\k]öÌ(g£ñ_¹ 4ÿÌÓYIpëU’Ì£I¯oz˜Úu%Ï +[°=¢5|—ܨËã?o¼ÜÕl9EJޮͅZ˜§Í™×£6` ææiÓ±ãøÁÝþ—ömÊ‹8éûª\̓×.a/Y”¦uãžÍi <å4èrŽ€‡r¶p}O¤I+1p»¹h´¤{[ÅÈSzÀ†z·Asvaá#ÿ¸ÅØ}Ï™KäîÞ"šIɶ,º’áH#Íp+e¿?r¬¨œ…hÌ—0>-`˜J*Ë™2~¨kJ*©FßΦޱæåÙ*t?g.` bÂ'7>Qcc¬ÚÊV–ÎbùØÊÿD :ýs˜áеÉ7OjÂM |YÝóà5†hÂZ¸…Ën@qygøš½-D'tŠÜ¯Õ+‡\ÕÍ©Y-³öâAÁ4夥,àÅ´¤¸Î¢§ò3©Ù²>&–ôƒ?¨·wG >3µ©ô§­LæŸk\·$·È€.€)‰˜zÂ¥_H²ÿ4ÝÙt°Ki1Ì”©ÌÒ)t_Z§ÂJ.^iCAÒ:6F¹`ìånJPå9vÆæé ï˜d¾ëæ!wêsœ¾1¸‚2pÁíPf³¸ÖÐçP͹å|ã+^ ³ 8÷åø”àF†Ã4¸Ô¶¶pcÄ°A¾ D6ûcf–qÃØØpÙ ?íÒûu¤Ä(íóÕγ{£G¢2ùoðÁv‡¡ÓYœyk›`íÆ{Ó¥ÈØlü˜»öCHkJw¤{ Ö©‰¤ tÊ2.å_Šç;T/ƒ‚”‰$és)Fn$nI~37ÅÌÔŸ«f™ ñȲ‚&õš`™¢E~ÜDðMy({3-¿kÒŽS’žò61+ðÇ¢S·H;–šgTÝc´:ÉÚžNUG4ÀÀ¤ñŽ“…FE *DYsŒ˜¯ÒŒd2þêêK¸—Wü‹ßÞt•fwÞ@º¦Â íï·¡=‘™}:f«Ä"ª´¶6nŸ!xÙrMûþž¹Í†ÄhpfÈÉç´çŽœªäÑ'Ûåv±’„<#ðô4® ªqbÝI»Ï(ô? ¹ÞúXÐA߉FÏz‚ßÀqþ ðÎû)Ðå»ËH»äOôCZ¾ôÓ,’+;çÅÖÖ@$E¶å”¯89rªÒÁÒÃtåÒ]=k?u2[­H÷ÊŠÔ)ëÂî»ztè&.Øêä}Â~~S* 8ýˆ®ß^ë»Ã{HvìM´’ÊñɛЩ5d£;YÈ0+_'ߥ%tJt~ce{Œš` ×BxñàÈ ŠÖÓ $‘|ùÕÏtËu¿çx]ña'òÓ[þsبÓne±=oçÏ·ö>ÍJЪÿØ(eœÐ„Üpˆ¹x¼ S2¯¹¢7g=w²¦«Uÿ–d`B)™ p´Ç5;Ð ãä •æÍì ¶Ÿ¶Ùô&¢rg˜Cæ endstream endobj 1179 0 obj << /Type /FontDescriptor /FontName /UIRQAT+Sabon-Roman /Flags 4 /FontBBox [-165 -228 1000 944] /Ascent 776 /CapHeight 708 /Descent -220 /ItalicAngle 0 /StemV 93 /XHeight 442 /CharSet (/A/B/C/D/E/F/G/H/I/J/L/M/N/O/P/R/S/T/U/V/W/a/ampersand/b/backslash/bullet/c/colon/comma/d/e/eacute/eight/f/fi/five/fl/four/g/h/hyphen/i/j/k/l/m/n/nine/o/one/p/parenleft/parenright/period/q/quoteright/r/s/semicolon/seven/six/t/three/tilde/two/twosuperior/u/v/w/x/y/z/zero) /FontFile 1178 0 R >> endobj 1180 0 obj << /Length1 2426 /Length2 25141 /Length3 0 /Length 26171 /Filter /FlateDecode >> stream xÚ¬·PeÛ–%ŠCb‰["‡ÄÝ]2qww8¸»»»»»»;$î;Ÿ{_Wu¿×UÕÕ?NĉµæškL{oBy>+= °•¥= -='@^WÏÊ’æ—½®¹‰>€žþSFOKB"` Ôµ7±²Ôµr„mM’º¶z'3' #€ƒãSQIÒÁN×èS‡‰ƒƒÀÌÊÌÂKP0 LŒL>qÍ]@K}+ ÀBWßØÄø7ÐÕ3ì¬ ítmC+[€µ­•ƒ¾‰¥ÀÞø†‹5ÐPWhøô hi÷ `op±r˜Øô­¬]lMŒŒí?¥äú»ø;6€¼‹=ÐÂŽÀgnûKÇîMh´uÐ~úöyÿß-®?ÍýeÜhkï°2ügÀ¯Oçm­­luÿ2¥kið‰ebÿo>YÙÚQÿ%ü Í`ieÐû+ºD4 8Øýõo`bgm®ëò×ÒÂÊÀÄÐäsõ ó)Ö7·ú+®Ïèímu-í ¶¶Ÿ['{c+û¿:[ÛíìN¶&öö@K€®õ'º£®ù¿;Jû_O÷'¦ù_]{Ýÿ_RÎ0ùŒúó€Oä¯<Ð}‚šüo;=;][à?•â_ ñ·/ÿAþàO´ÿ ýOÀÿ7©ÿDù÷äÿ?§þì?ŽüŸãþTû{Àþ P÷Ó;#“ÏžúȨ̈I …®­Ù¿„úŸ#10|:­ÿW”F&–°tMð/KC«¿fòo¹ƒõ¿9mí>g@þ?æ™âï°²ül  !,”•ýguäÿVÐÿd‚þ©õÿãþÿÝ¿zø¿xöÿF%°Õñ¿I$°ÿ‘ü_Ðìÿ™FþÏ$û_#‘ÿ´aÿ/)äÿ™@þ[‰þ'ú€ýÏšã?L?ì•<þ³žƒýo±ö?%ö¿ÃÙÿ¥ˆÿu„ÌÍ¥t->g–ñŒB øÏl&F&ÆÿM_×ÂÄÜå7þVÿWeà?¦KÎÊB÷;ü2Ÿ¥ÑgÅi> þ-4±6qȘØë uÍí€ÿ+Zmÿj +;“¿^>¯ÐÓÿËÙg±ôÍ,ÿJ Ë?Ž€ŸÉÿ—? ío‡éøÉÉH*SýÓ›Éß*Bµå_+yûÏÌéÚü»àïc]Kû¿Z@ÿ?!ÿÞ3üϽ¤®½­‰3@þ“>?ÿ¶Òü<âç·rv£a`fÐ02²’- €ƒ‘ÍãŸ4õ>;ÀÒþoþ ìßö†&Ÿùú°Ë Vú”¢(/•!Þ.£23ˆöÄZirî–+×jsnH“<º"1½–ŽòÆZ†æ”Êb§¯J%¬óÚzò‹¸Á‘39^›ò"®}ßÇ\—níŒRÓs9¶¿“úÙÜ:ô?”j>'Î×­lw±’’0:ïk…]–G¶ÔÎ:ãã©$¦¸üZÝÌÊ “$@OO>ÝÉBw–(&†G•ç¨'1YŒa÷Ðùɧ4-D˜P¯ÂÅà”±Îm‡üVדºæ~žzoUx–‰ °Ÿ‡üÇÛiÅ“Ã5ªaQF³Å€“Þ6 Ô!Š]íàþoP2CQ$Û'θÏÂ_/8Ïî_­ëÞö‰„$éOVEß5úË|oÃîÙS+´! D·3íø å7UT uà 9¡ÔF{zð—Ñú {šs@Á5…M9„ÛÁ£a¿Ð£µ[Ã_¥&%ÝWõÎA$&p;:d¹FâRÅ•jÁ= +¾Méô¨Ð© „¯÷\£– Žø>Ø»"ÕÜ‚‚ä1k¶É)l>9{zý³œOº›Õ%%1•Vé]}ßÒÁDÍ̆~¦Þ8ªd&yè,èÌ2,ü©U3ú–¯5`n¬«*mÊÖÈ9ëh9óOo,î~þ zX%ƒÍi|W‹ŒJkƒEö2Ôœºd´Ïk.MvÇþñý4²L láb`£´IUï#ntñ‹a‡ú3Š’dûQƲLo>1{w‰þs/, ¿°]úÑeÓ½ÖNœ>×™!œeÙŠO4ÂLU0:ÖEIÊ5W 6Ê[ÿ 9\·•×+Z7 ¾©£Ž-’m‘%øQŽLäÀqkÇêj¯¾HǹC,ãcöÄšÿƒ<ÞU¯GÓäx¸¢È¬î¶ü{u¥EÄ,µ3™•ßèûòtŒHz\AåîVÛé)?U5,ÂeÇ´KÂÃøp{ Û3RëLlîúhlÌà—§Ð\4-eß•CQÂgìB×]Eð˜çù´Õýßñá#½‹G>&-™ñ5$2"èYfó7}Híd üÉrñc¯+¥)ÉHï­”œ¯ ‚ "'÷’°^N¢±ºHærK' é­&ÐúƒÜdî¼ö,ºÖq"•zVãÐQB¾æ&›jÔZ½ÈÙ‹›T":yušÕy_€lž]=P4ÏÁ“·yH‹u??öà‚šTç[“ª°Ïö_d’+q ×ÆmXø_.q‹“y¥çz³ùLp¶¹PüÑs‹OôÉ4º$žû^ÓÖx̨)&á0cýï”ë5\²(ôõÍŒg(ýj½%4°ç²´wôBaíá Jï«#Pmû½øÞg±&1£Ð áIsstýVF™>Æýî=ýZ5;i[Pè.Wð 4{v„p3o’0»ãHñZ^5»ÑÙqá$ŽíJKüÐ…öFé¨èí3º¾¡ÃWVl„—3º›üããž Ö!$ŠNCÔ/´×&÷£¡½pÝ¥)izYi>{QbJîÞ2AeÕ»*(›-=Z=½'–;h%á8Ä9Š” ³ ž †ãæ3ûÚëš$Cû:=gωÌã*E÷ÅƉ%Ò= ² ůÐqþy]µ8&Fe86ézÈ s<ï;b‡>@j¯>ë)„øØÓúU9}™uX}ö­(ÖµrÂÝ‘7Óg)hëõFÏ‚i’fž»‡Yæ¾Öº¶vá}ìBkK8Ñ WLD4a3³¯‘)Û?ÞĪ½DŸ°TÙ ÓÐ\AcPt5—_õ–l%XðÙ.@jÊdPòb0 ð&FÇïgG×lN”#zl¬¬°"µŒ‹#8ùÌú¸”Ç>dqÄÇÁÉ‚HÉ`jxÇfΘÇ"•.HXö'œnœ)´Û~jB¢Aë…d˅܇æfk¿ 3haë óÜMÐ1Ft3&‘Ëò<Z8jG`me¤ð<«¨÷£-me³‘-PÕ ˆÝ#,H@ŠeQZc’Grkp:X0ÔÞ3u:,¹à’å¶CµË5C_¼WhƲ´+>RŽ?$¹øÀ—œÜ-£ëÏ .›lLÍxêZn›±䈵±‰¦† ìvZ„{Œx´!£‹âÜs~wÌ›½ð¦ÑIÜ>~à¶ÿò¬À¶á<yÉš=m˜ù7‰’&¿Ï° Ó«‹Â5 S—î°W¨;èæ  kÙT,QÎÐ7’”FV,ûTqrmnèeÆ7³Óž¿Sè?]ƒ–…H¡`}Y¡ Lõ\àû‘8¥J2׈Å!Ô³Âmo tL!¡ä pu"L;^[‡.*5˜Bsö3±ô=J¨Úþ$Ç?_@0Å‘Œˆ_úÔ¢ðÉ9#~ $@_›ŸYÒ+k÷ »€èw}yôùÛ9ã*”XÎdî¡Ï 4Ús›EöØ¡qîîâõáõË}ì²ÿö¦`¼"’Ût›žÎI³s;¹0ä‘py2¯ÉÈñÇþµëcÀ£ѯJËäÅ °}Ü–›¢ Ç71ó½Wë„@’1Ù ^ƒ$º®'KuDQö€Ž&B^¨Yø.¯ÝˆÚ˜©t<^“X Gµ4³ÝÓ«@”/6$xª%>ž÷PD AQŒáµöZzFVrWÄT²=Þê Ïä0»|ôÒÅPlݯ¿QÃÌHÙ ¥ò=ÝfÕðæh»Š¡`“: “ãjW?'Ê‹ÎD—ô×Ý$ôÞÑd¼_£Ôs¬-LTWPAe)_CŠdd‹‹yŒ9›áˆ6æ N»ïÙeq¿ÂþB´–ÂàÛ€+¤JXáü1®¨b¤¯=AJ–é3²h¡ùÓS»l]0ÆëîTnñ†¦Ó˜á ¿ä8“ßîeÈqÔ©ìŽmøÛ¾![Ú¨òô²}ïÉäƒsãºaÁê£Õ¦ÌêPv)Ê°A†*  ´DjárÊÇ!6pAÐA5«g SÑ—Ÿ U´Õü.šJvüù扽À¢z!†(u2‘YSÑŽ{–û[C•œ«`cýâ±²ÊS÷ds-¾Ùˆô( …\ò-¹pˆK¥M°ýÍÉ<º€‹ÇÂñb d4yæ=þ>ê[;5„äŽ0ZRÁׇ}2g3ê?åqXæ?è÷ÜÂe]ŸñR{‚.Κï±p)L”P}çp›zq7Çy¿VÊ$tͦž® t³cŽ50ÂdÑ×ô§âák”ºæ|>TH^Ya™bÍ|Ú>íjM†$Òw'tMó¨€y…a0ÞoÜ31`´›©®y § 잃Óø&d#xj$æoà¦ÜŪ̸‰í¿¥Ÿ¾gÕgõö(@a®Í }‰ÆÐxâÆ °À¬ÎÐ k!ÙkÏHHÆöɺ0È)ªËAAlˆènæÚì¹pG¾Ã¤s­p5ÓzvãB3‚Êb7WÌ)‹.°ÝÔr›cÔšLçݳsâW Ò]» í#F†(·ŽúÞR3k)éBs{·_ܦKaŽ5’çÚf&Û,»2<“Ä¢ ;ù¼µé¹–ÊeÈŠhÏHp¦€lj¡¦\¨¥UY 0´ðEiÁ噫ŸŠåWÝa”Yµ²BK¬ÍÀÕãÕ~„¬DìzR4Aö,í¢¹‘ïÙâyI‚.‘V3;ŒoÎ{m6\fÉÌƘÓê¢2·aeÊ–îW2bR0H²–KU¦ìÖcþ/jæe:Aý]‡ó»ÔºÍ0õ–l$®ÄRU,œn–³™Mût‘^Yu”εq,SBå}“=~ÿ·%ñ"áoÅ¡w‹~¤-+?˜Â7çDs5wý®S\L>\Ø‚q݃šï¥°¡^í:x´?ž,tK|†àxªŠ€¥nÓ ÐAö°§QF£­æ_ܪï þ|ó?8ÔÓÀ±5ä_÷J™¼ÂGyv#µHå jÕz7?Y÷5ïV¾ñòIt‹Ç\éÚ3Ź¸:=%ZÐñÖHåýùæhï]¹¿Ù̽Ýiz³× Û$’‡a,ÏÖã%h%?Å”úè•Àħ|dÜ“¸Ó·-Ÿ ^Tâð$Úýš´•ãÙw½y§L³¼0z2'µÏ¥#¾"Kü(9±«€#„-¡5ÐLJQ³ªUcâ-'” C¿JõºŒmÒ”fwÉUS«:“€nHdÖ}áÇ ™ª_ &Zsbf0nNþÎÛ07ÛȆ›û‘Ë_¬’r õ"ôÅ‹Ý?ìêMÖ-tÅ¡©è)þ<^= "übç¥ÃE›;E-,_RëÒÃè;Æ]#“Y…@¸Á‘éÔÃ@yÄÛ€úèø‘1hEþ=¦ÏOuøC ¾§sÛ¦Eý× (¯éÇÙIPÑÔFƒÜ¥¾®áÙ¾"GÐ8±¬êí§%#µJîž“áÁzŒ®Ïe)_±‡<¨x –°C“åa­ýG›Íè==ˆT…py‚²ŸÙÄm÷[éƒA©Q<ÓY«¾¤¼ß9Uî΢,®î¨h=§¼eçÁÍ6KXõÓáMxžÉi ,ÃßÑUt<)mÀ\AïdCLàO=óüµI0/\ tŒ´¾.ÏÍcyÍ8ºƒ¼0‘Nßqðl½æ W`‹*¤ïËaŽ±ü½Û†U<”ÑXáƒ:КÎ(4¸·+ûVž´ÿ•vñÊbœ¶¶»ÌXõ‹÷*Fn@èkPéþ|ÀÛK9ØAéÍ‚}˜Q70¶Âï¬ægüa²'—Ѹ{PFÅGºCðÛªjà €*> Tà^²½4ûNË™Bqòûd¤D=WlõO9dùâã!/ôJ뛌ÅÉá^ó5üßM«»·ºì´Ððá\~¡§ÒÛl¬Š²KœÐ Çö1aÝùÀV¤Öøâyõ¨wS– ا£=ý`>™e‰™J2£‘̾Éån[¹ã@)r¨dßœQ|Èj½ø™0Köjÿ;‹eç·–r3€S6rÇ°2«ÅÝ)1q¨ú´Á7< Ü~–ž\‘îJ4:q»Wž¿6žúX¾3/}ÜäÆžF)¦7æ<ª3sàtoω·âÕÂ<-+>°·0€lÑ­B¨‡²…‚ÜTæä‰O-Å£]ñlvF]½¸YbÉ“)§ ñÞ¢Æ$m‡ÛËœ;ÄÏVj½­k'²›$l¡qÐn|;pÝ•Ž Ÿ…_=˜»þŽš2 f1dÒ‡ëÛo·gù›ÂçâÝ™¢ Ô.m§½ºê*¹UzðI~[%ÈyZ@wÊ› ¸Ø¶y ä§æ ôÉùVJø15»õdÞé)X˜•µ›‡ÄÂy>ý=…sÓÍký€h—Y‡‹q q.ÄûFaE´›ÖVrm!6Ò;rÐÒÄŠPO”øVFˆncíñ|Š$´ÊHaÓ¡z馱éÂlÝ’XýÈÆ™¡à§þµŠ‹>³›"»~Ȳ†é´l’ôŸòð3$:l•âoÒ j8‘ :ÞF, ½Wçi©E¥Ù}Þ åŽÂò¦|F±¿.»x¶Kà%Û$‚+ÀîÇ”ù&Œ“7\Ä Øþ< °û­î˜>«±m·“,DA^þ1I«ÿÕõTÈ)¦,~ÄÕùÞÖ¢™éáÊúƒê¨õ,Ægéº2ŽxÛ4X©Ä?hÍ!@°`F×Ó÷XS¯@ßF%}N²!ÿBHÖ³¬ë4bf¨å°BŽ$‡Q܈·³ö3Ò\?5fßÉÏ7_øv Ñoåâ—{åi±uëÎo|¤|¬Ø²"W¡°"â´o!p¾Å¾Ö@µa .s¸aÜA—`J@õI€ôgÝ —¦cÃï .»-GûÛì÷$Ĺ’ÎÍgçßV.ƒw†No]F6”ùp*ѯxŠ´øe¬wû4¡ì››V~«ÄÈ'E-tz•R½Ù¨åj?Ñ!º*É&+ÕL„ ù]¯O™ïã8WlÒ¡« Z©·›±šq>˜«Ì:Ä¿ë &ÒIÄÆ$w8ŠDðhÞbsA$œÃÈfi$x¹êo&J€ˆžj1Ìᄳ*ôͳ·ÀT¦ÉkŒAîŸëFª:„´ÃCoPƒl~æs¹jÍz«‚LT¢,sPG‹_*þVçÓ÷}Àc£‘¯øífQá÷ÃØbjËŠ¬âHÝû…¨õ›¾ò0.Íafß.ÌeΙ׳ â%¿u÷à&¿D³Ü€-]߯v<üÉeÉÜÍ4öÇ7û82æh¸Ó±>DzFOÖh™¬6ÓjëåN^ô ¤“ ãÀsÎET ѬôY8ÊD¤}¾ïÜýI¬–‘ÛWò¡„iˆ‚ðÇ¥±ƒ/ÄÛŠî4iSF97AR=€ƒl û7N‚«þ4š*¸²|ëå¼/¥)©¸Š6PY5Ü8½ l‘ëUSØüý_pàyÓ,Û»7M¯¸6 #fo¦æõZ[ShB1ûãŠ?=~*v§ñKmbÚaדx®`pä$´¨&ª ÇyjM¡Ñõ{í% ÏŒütX2a@3gŽ¥ëÄÖ`ayeµo.lõa¿²,ðs1"Q¾ÏœãO g‘ðòf¨§°OÌ™:[W­:À†.O7Ps´²wˆ\ÿv;5J zø&ھЉ’­›˜°Á~«RiEÿTkIE¨ÃF£î.é§ßßäÔ»ð3Üt›¤@,ô {ÖHb&ŸýÃêæC*ðHãë\†pMº—)ÁWʇ±V+ÝZarØ„Ó—¤ô† Ù¥…Ja+NQ,f›äè|Û Æ=]ôž&¿{¦¯ÎÊ$lBÙ•nɱuN¤%"p<°`¤Ëé³ÐìƒBSÒ—¿wüž=Áí˹Ë'ÀG^;ܯ>¡! þ=GRTLôOÌ.Wm6™ð)|»ÅoeÙ1BRí´&Ö$©pob_Nz5í™éôyû²0^ç`cÁ—(›’sZ‡†Ú/ÁëeKÑ„J¨pÆæXî‚S±S­j?A•ùýÄ“PJ!å7R.iÁ‹êÓ¬˜ °¥„rßØx¢*•nSxkÃØz½sþÜ÷lK«Ÿ< 6ÐE;y(6å,±ž†°ìQ“Z*Χ6@Ì .é€ø“Cwh³6ŒÉø¢ªHÉÙ$ýa·#s’vZ?qÌ/ãûû·âf´«eÅ7³å1ù°j4gKâüÊÇYo›c„v®žoœoÕ6¼IͨÖàœ~×$«vb$Içè +CÐNT6QWëÒ1ìXA$¹Ì_<Ÿú[©S ?ä¿(MÚ*äùú‘5áKbÓ†YJpám©ßoø©û®|‡5™Û¯,~ߺ×Õ%ß¿J!ÐÅ=#Å#Ú~E¡^ar³†%v,%›@p%€¤ñᦤBgXŠÕ.òI'Ÿq¸tîº~iú¾yJE»¦¼—‚Øü›™¬eCÅã@T!©Ö0×ÿµe¥{½ÐZ’MTÿ¦hu¾e©.k ‚ÀU 1ZÏÀ˜Bhʉ ñg^Ð'o°q÷YñãìÉ~ L!Œ•E2ûm'ê!å€ö«€iµQJE`ÚÐ>yØ–§ òè#l¢£);wþ&º˜©¼®hgðrN#‹\˜>¿0¢¶LÀ©‚4 `LÈ-ÓŸxµÌ墫/fbUœJ‘Nò:ˆÒü†LSt§S¨¶Ç4Oí,d{'é„3XÐÇ0XšCeŽ¶ ²6dñ³8µš³ŸûfÓч,tÚá)oûýòOi=Ì—ˆw«0ÓÖLƒ•ŸküšÒNÜ"¯\ A.ž;…×(]ÇH»`mf‚™âO¿)«Ü¶Sx&ýíæžÀ~wÿLé¬28Œòý6kîh÷½Ð}§ãý¤ö`{™\ßÉx!î¡Zë^]åu“Èp6÷£éÅÃ4Òo@©÷ŽSìüâ;Ùof{†/¶X±—­œI¬bäuÖèªYt?˶#×,›°`Ìt¾(¾Gížéª,ûw­Pßþúδ^ï¦÷IÈ8 X?–|vÛMŽZín'6ȇ?Ç ûK¸®ñkÙ/žÝÞ)Ë>>_„?nNŽk?.Ô°pÙ ª0¿—›rºàÖG¹Ù¨ÿ}Y³&XÇ+ÔÈÐáeuÕ•`à ÊL‘ËNïÉ)Ü5W8aˆsœÁÛêˆã¦ïŸ5‹}¥Ìk{%˜ 7))'âùµ>ù°÷Ò6å´ûÆ Äƒ;OŸœªm@3 ‚y*o£õö×rëf±!–4lŽ¸¦ÕÁíªìq8~yÛTÄmz§„ë¨=6¬0†v'ð'ð½@l«i*¾qšÖ)äôÑPºÐS“æ§Ëg¨œŠ¼:yµúÄŸ$½«¥è§EI²ÌÃalÈóÎHb5ÊvB<;—T4ˆÎRà eé'C{—Ó£¥×T v?¤X{ÚÜß¾>,$:qî_¡òtU0Uñ¥L›{-j©øð³èÁýSÝõ²€µÜ½K@~ê†ÛXÑô R©?ëjÊ‚½5gË%ßC‚tþ] ä7CˆGªDÙ‹Í‘ÂZ”—êR¨dèg¯Ñ˜‡ÍùöxˆÂD€sñ OEöö½§ö¤©™©DÆ[Gâº..ÉùvôxècPS“×؆:ŒPTöèhnÀÔÆkBíüŸA…õnLëâ@¬MgzB¢€ýÓ¼ÂCÍÙ؃ÉÔ&^q¨å2Ï¥û…»‘(ŽFJá/±Ó”C®‘ºQÔ?ʇ}2Þ$‚ónmpìjÐUƒ°ú2”ø1£"èÀl9éÃEco#{Æ(¼qÂÕ_­S‡iu꥟síÛ #µH×ñ¹–ä²1Q£$Ú± ÄÈ\EÄÍå²û¿{ØUý#½L”™çE-ŽðÂ0ׂÉdøV´1V«°k&[¹û‹éúçÏì®WÙìN4mxvãkè®1Ak«‹Ç³ëå~¦‡;jY˜Wäß;ïÓrgÕM]yb°¬Ã–Ùd™ùÿiWq¦å½b‡æÓ¿Œ¦ÀC&sžÕEÀÄG8²@¾ÄO‘._‡&á÷Ø.æHòÞß»·¿/7zYeUðerJ‹ÏUôN¸Å—(’©6…™Ny%gð`O~šDÛoçÇ.·Uð7þ˜"4|b2'^&ȤDNlÐ<‰ºqó쫳§¯ Î{”e\<ó¹·Ú.Ó·"ªµH³«©¸\N€ÅÖ6xcü2p›øÃâ‹ d7}"¨ $·©FsMÒÁ·aLBþÎMe„—‹?”ûî7qËû”¾ìbòŽ :M ¬OHöPݸc…B¸IbPD©fxœqi)'cß@,÷ˆ&k±™m2]:Ò æˆfËÁçðx0Ü]U;'RVÑ„´”9? ï{E%ˆkug¨°÷J«UÝ”u’¯2GmJãË[p`@,ò• ×…BU7`5:{¾^(Ç{%`Ï]¸>¼~<øélk›i ºîþ‘Ä„‚ˆƒÿæi ÍOŠœÿ„XÜq6ß©¢i®œóðè¸Nü]«*ÓŽ(Œ×è+øU 湤· F=ÞÑwðÒ½X2ø"§€}¯ßhà7ÓÂEŠ¿X¬V˜{È üR2óBÎù0ï›=tf¤V3Œ‡¨s‡¦‚¡bš#‹¢E‹Å=]~„‰3Î&ç Ó&¯Î8à\'Ø£Øm.D—HOu=6Ñã3þÑ/{£h©¢% ­›Ë);¤…Á`7ÍTO´S™¾ ÄþUŽ¿~?ùiS 5ÿâÜTrð¸7œÝ>(7 9IÑ1ëv¹!‚ÐŒ“!TÓh Ú6ßøÇIfO³Ä OvP~¥á%陧¦E'̉m{[Àfá'ß8ÛP{ô÷ô`NìlL­÷·FPñÜÎÚe;åú‚dÅ«ÄéPG|=ô‚1™M“Þˆ8CH—¾7RV:óÜx;s4aG/–1•ÑîĬ—†DKÇ^7,*—{Cwdï–qdkð³ø¤Ü“²°L/‚EOöG¸O¶[êi"©?Qû¥„¸—Oˆˆ“¡:U2Ϙ1™Ê¬ž‰×ÞµC—*œ•ßp»ÊüO´ÖûkÔaí¼ 7ÍlE‘€U.ø!²Fè#ꬿ’ÒžÓR9û¼ÒºS¾­¡¶ãv˜*2ëƈe “$ÃLðÊãe½ýGñHï…w ¹ÉŸëí÷:öѧ}eBNÔ§@ŽyÔý­fR´ço9rŽDîãµòÖ÷»„‘ãZSH%iAiC[wÊ¿†ÞI{`ÂW„ÆQYSm™M~zÏWþþŠŸ½U°‰¥fSYü. Â)÷V»í´ÁZ_°6™¯]ýÎ6R›äH×Ð'çÄÁˆ^Œ…k%ò,x9æIþžZÙÓ塤äkæx£I¸ó%âœâ¤ø!:cLô‡A[ºZIËfhæ¸^ÁZ2â|Ù:¶¥ =hLLNš3Û銓íwæ4 W5‡ÈuFhqnŸU;¬âxuµjFŒmXnl?|0K±}„Rg³d\C³ÿKm†á`vLLôë=^õŽøð–â˜@5#@ÇÛN_˜w£Œ>š E<¯Ø·½ØDJµeˆ`8ZU<®_¾:Hý) 6ËZbn¡cè®·p€ì!Ih\Ñ&ÈMNà ªÍÕ†8EˆÓ™âÂFº)¼P3[‘”F@+Ï{„ߺáãÚ”Nا?‹Yk_²‡Ós¸Ü¾¹—ÀȽ”0Ò4Ü6t˜ Y4}ã°ý;7¿Ï4ÄŒø—~U‹˜;HhùŽ‰Äòa’ð\¨25${Έ¤©ºi:™îÁ竧ÿgyDœ 4¶hӯσþ–Ò8ûéèQ{©/+J] ªd|@3ÓѶŒçV»Ð+j}Aç£?0%CE˜Sy{¼ùX\L²õž¦ÐÓ¹š®ç¡Gè(ÍJCþ9f‰M-“«NâòÇtá1êË™$Ë%‹ü¨4¦XŠÆÀŽR‹ êáD::Œá¨V'ÏWâó¿“ÅQRü“·pN°{À³JTƒ.~ÀÛ!ÜÖzNšÜ»Ùª§Ý¡u .#°’±jO=°Zûz´ƒg??·¿X8Ÿgt·xE$mÕ#ÖhRÿLnq÷‡–”’I(ò*,/-bí!oÍÏp‹a²Ÿ«p¯ÏŽ²©käíã¢g}ù®N:ïö»B¤Ò¡ÍNÛϾ¶*²½€•YÊ¿q,j]jÕ/x¶O“̼·I…C ºŒ:ÈŽ[¾ÆCÒí§Âºô~DKòà5c_3ËÇY• øö°6 ÄîʸvH²žO$¸Ç±¿dz¢×üàÆëùÚa²=É Ë£ùë°Ÿ7Úż\Ý9œ]Z™"5‹8~£½¡ká)&U¤ÌdÊ·¢´Æóó 3­ðÑNiËúøX8X‘KOG!èí¯o]ëùÐIÌÔ]œg¼»’Ôe)ê~šw÷ãQ-TLtÖ7ÕçñX:P5 ºN ½h»Ój¯îKìa¬Ì¸\š0Üðv³î…¥[ p8ÎEú€ ±z‚¨,å 6Èô1ÚtþC*ô¥^ˆSÌãúº…HHÀÑ+ÜæL%™AÂiOªô$K„ׄ‡@ÐÀ6Ñx˜õ¾îL¶q_;§Eå Ø7ÏŠ¬!î°e|2ï5ƒLÅmŸÀàý‡» EÆß ¦)œ ÞpaìDsSÂ:ß›,1©‘û=R›-fD »žábÈFYT”Úq1/àÑUSÆXGÆu}—ÅBÕ«r-ShŸW¢e‚±ü¤a/ñÄʼ÷ä§ëd8¿Ž´üz Ž7&ÿÙ`jÕ±YÜÆ Q E9®q4m… =g›fF›ã48ødümÞiZ£àëOxÏD®íh…òDJìßAF”’HøßV÷ÉÛï=¢Z3êצ¨Î+ø¼üI£É²Öú æ~ÖFÕÅZÓ³ZÅV,ûͺÞLÕËizœo#åuTˆÄ è¯iUâ,BÍéü|ŽX‘H»zîÚތ٠w9¢¹²1AæüøI&í<\oRjp÷3' G6u^eÁ†²`þÛc2¾=‚Ž½Ó·o´ðÙª‘£¸ítÖ5ù8É9Þ0â"ôȬñ¡Ûª±>@Ø@¤hTÂÖwª*)躼oÎf5¥ÊÖ,ÌÑH÷#ƒRÑçPºGä¿’ÃâÑé•EÕ3„tÁ*Ç°³¹\y«7¼W•¶#ÙšsöŽqˆf)†^¡R×Tìì(²õýløµ¡çî2”ÞõUÛ#ó{ÞAÓ~Äl¤}¿¬Ž«‘Xñ%ì˜9ç6E•Dò?ÎÔ:ÿèìifeÂ3ê¸dâÀ†™Y{V8£½¥LŽ°Íÿ ha&î²ÌSÜ-Ñy¶ÀÛ3g‘ÅiŒg¨ÄÚá éÿjbT÷Ë4M]$0F$ævFUÙœÊK—}pÚ}þj±ÿ‘¶BSbè‹8fy¥w:7Ï»&þ)q€«iûlw:Î:‹ÈôusÅ䌜 D"üþ°dS1ï¯ëì*ó7¹ósÆq®CGÉe%/u¤™YGö­ý]ÇgQu™NïZ¬|¡ƒŠšQ¥ â‹Tâj†IËd>aÊJð0Q_Ik7pôËha–$«š3@»ÉAˆ]Œ•ž”Ѧ5Hq8O°Ür^I#÷tÀi?úÃ+øf©Ë~8êÇÙçßÈ2Âk'uYjz‹}ŠõD¶!äj2ôƒ0–7.C"OþCÂdNòNxÚ°# HÜwSÕÏüéy›ßU#$ÝåȪ¼à}Ö±>Ü­2’â Òâô_“¹ÅVŽÓÍVù%4 èÂü“5}»TÀ¶¨CÎÍ”–w®ó»åVü©P§#ßð¸µî» ³ãL—}ƒò² ’t=¨Xtë ß A"9 s1ß°ëÕ«ÞÖðãA‹´’{èc^<Ú8d¸£ ï?³>X+DQY–r±„Ÿì=Å •VHG)ÈÚeâÂqfqñ¢ÿzuÝdž©<ôLj¿1Nã‡)¡œ{ë˜yVñ1R­9‡nØ ‰ÄgËFbnJö›åq½þ«êÁH˜[7Œe„¦7¦ÁðÀ·œéø‡!«¤&VK7ìä—ðE!²2zu#hY,ÔÃìÌ.çÄßjFÔ~\åUUšUßú/qÿÌ溥_Õå¸ÂKZ´¨gá)Y{¡·Ñ2Øà†¥è°½yÁx9Û¿Èã*å„Ü/>}oV5Œ[_’ÁœÙ‚Þy5ô·§JQðŒ gòËå´ÒŠbÿ(/*€8ÁÇŽ%$Ž*Ø¥ž»«:­öƒm¬aà6+Qºå¬¬ŠiÕ¢ó°K«žÚɪ›Âg}^¤¢° Á)»¥ÿê+å"?¼´ÙžŠž×óËlQ?Dñ *Š9˜ä c 9‘ù(þó·±ÜésÑß$’©|$p2M‹‰ôZT#€fSÃÅÉ%ð$Ïs¡ØN‹)‡¸~TP›_ñ:X÷­ßŘÈÚÆ#”(Ô5理NàµÃì`å¿åg€ôðÉ•æ5|pCXe;o×b=åkh¼sæ'ÌHІý×¥‹Ûð0½Ó}ær­š¶º3ç¾ä9°½£~¢Ã¶Kĸ¨4¯ÎýÄ°à:ÍøñBª2Yˆî“VüPº•s„aàV}a]¸ÙÝÀK?¡ªö•ÚírߢäLRfšÓ³U¡uŠîA£&¸Ÿ×úRò“vé8vÚy{é®|¶öÆýgµÂÝsÒK9©K×V…dÜÒçtwþS«ýó½Äw5ߟ$ dx»ØóÎo\Ò¢¨Lä :øçxo—‚`a©Æ_Ðú$öô\bc🢉 ƒÃ€Îžïrc¬BmOr>Ý>*ÉWı6 *a™ §i”gØz¤ë¬`è Œ£ö.} ÌíÉCš,Ï~úk¦¾(Œ”Ю÷(“Um¥ð‰TŒ'xøñŒ=dÜÒýZZ˜0©ÕËË¡Ï ÷éh^$ÔVCaäPÒ!¿kõÑYiç[ÎÈ/²a’,H?ç ™dŨL¶gò8t½›˜JÀœ¤Ä-«#E84k'Âx²ÌÐÔÁªëÜÑÓ³w•'Ru«Ó[É]9‘!Ò‘š(ZqE4{žþ!“_«Ù·D]ÊzеÉl/ĤlÌê‹ôæ#úÃ:Ý’,½tZiV–!Z¦Zb‹4ͼ2JVz¬O`©¤àŒÒ#ñ±ÊPõñø[ê67ÂÊ=è ŹSTZÿ\É…ôÙîÕK˜fÈGÁx>m¥JäÔ&LØ[è"ƒ"ïöØ?m£ÖËɶÇl>ý ß-þgºn…Ë®sºÁo}£‚­¢Ôk_3œœLëœeQ%Íê¼P¥|>j.¸ªóøAhظàÛ„Vß…LÚF-ë¿Að3i B¸•Ÿ8ðs@Ó(â))¦ã{ÌÖSR¾ÆÓeeB‚èR©é·íQÔË'Ôd<gê¢Aæ†F¢ÄÆy€; é€Š!\R·&î áF%àB¥Bå{† <Šë£k ÛÛœ ™}_ð„sžCÝâàó9x¢¡v×î¾ Þ·0B¬j›º'89Æ¿O¤Óñ¨ƒ×}¤v °Êàõž2 -_v‡Xª•ç¸ ~u+t$ªN ˆ¹ÌþµV¥¨¡~µ~àã ö í…Âxñ7¯™CÒ<—S¬àt­Ää—™æ¸ÓJÁÌGòifeü3$z¢p±È rƜᘂÍ,Ò0«±©ç^¤ÕªwÇŽ?ªöKéÝoP¤›—JçÞ úZGÍ€]‚N.1à xíq;ða蟀«É Z«›à˜íZ%cƒ20§Ê1̯l&SrŒ…ñN/¡Ñõ8\©Æ\A¦§ÑuK¶Ýb‡"9õå\‚ùi…rªÆç´²ÑzƒüDѬ’YTQX€}^—Næp':©dV²q-ÞP}¤1§ È½v’CÚ³z¹†!§eñHô\H¡•…¿Ž5zýŒ#'‰_k ãå£h”{î[}Ës2~Y+!ò^¹YÙÖ¡ÆàÃ~’Õk6›Ù(ò §mê2PÀ_I»„– `Yè¸SGäÔèÒ¾ÔŠäbÂisÇœ77š}¨£“O΢å&p4¬ÿž¶v£@<á­lôGŠ¯çÌ|,·¨‚D C“LBç,xs’Œª'b(˜S6rÝ¸ê­ ï¾b'VÀ(Îç­ uõé%¦tïêïåp¢‚Ýšó&OWØã«ø:ýê«ù3îsg²,XÁ ð´¸ .ýìçJsŠÓ¡S¦&¸ÉÉ>¼j¦02wìó Ôa‰#tª&.aMSd4KŽ­äùóš¬jkµ«¨-s„ÕÄO+Ýv” ¼sþµ™HîN««ï»Ï äÆÁ„ìÔè»Hwþªþ‡¸~Ý&<À%}Çw gì÷uŒH ³vÅõ¨œFw›WÖ¥Õ¸]"ÖlQ>gp{¶¡—ݤ`Õ»€I™cq2=XŠôÊŠ­TiÂ8]ËëCð½PKxÊrZâ¬ü™FÍ…èÂuv0žÀI:¼æÒÃþÙò*ræóƒ±C/P>ù ©7DrX9ŠÛxKyÝ G>%‚PdbÛܺšüÍÊHÁéjº6ÞWÄg<˜7Ì!¦£DP:yPkaÇ‹ÉÕ­¨I’ßRøm×zÞ¸] ÓŸ§5ŠUhK­I‹¿Èر¥ÎS¸ë&žyçAŽ•âñkÌ…¸,·™Ìϯ‘¿H€IžrJõìa›]²á?\R)h«ºÜêdë4Ú&¹ú!/m°_££±é0i [é…ÙˆR¿,{bXúf{ôÓÓ˜rö@>ÜÒÜ­gF#ð¨o3²×ŽHú#~Ç—£é3íúvÈüÕÞ™-µ Gé1—…ÒÉ>˜²é¨UƒØ\ƒån²_±»÷bÈ/rËßp~D"[ˆåGüjœ¿¿È·ó>¡].X8š°A¥†Çã95`⤈)Œá±Kr•ZX…fFû0O"~ËOÒ^j9΋–x—®kß— (`\OÂaÕ»oEï¢ ¢*i©³”’ê…®\JðnûIwËÁê²’M–\Ç=[Òi±§¾Ê è…©ÓQ4¨ —ø“c‰÷—‘=ù<Ì#r8_{Í–0)ufùo‘èE;À|kŸ¢ŽãÚYJ)²„d“C­ÒIùNV¸Ǒ_•Nl˜#ÆÏJªÈÓÒüÔãrgׯ”å”#Èþt–Ö¦§ã<»‚—e~ß·–ŽßZ} R¿Gc8`p÷*×å8‚U°,ð7ögxCóFª%…J¬•=ÞÛV·ñ9µ¾ñ?PûïïkóuPlM|é–5Y½GÕ™“«W]Ròýüœ š:›3}5Ê öU6­ttb;„”‘vªºÉ=CøØéóM–hvÙæYQÑZt%óʲíögä¯å€"Ò7x>"O|·«iøéÅ•ží¦áz y`›æS^oØû\T­T¸‚º, Ç'gÓÆ©œ{¼œáIñÛŽ” —lFeOJÒ©«Àym}è‚~™`•²?@APXtrzM) Ž¬…(¿Ô`OÝ¢ö²K ^I±å ¶E2î„·á,)â[¥ÎÔâo>„ª^ÀÝAíPóy\[hö–ó©W…3kQâÙF²]3¢MŽzÞ#¡l5î= G?²=»Ä„U‘eF6àÉÊôP'ÐÑàiLa™‡2:»wÿ±\·ÿÓ–‹ŒA¯´`;ZI»¥ÈŸ:}ÔƒÄÜäwp¢%çÇ»½`,D°elJ E’§xâ@"­{”~´Wòë˜÷»Û]°lRkw¢Êù-ÝkÏçJMVD$)d%½§•ç~Ú ç+î³hGÜ~rSÅÕ±ÈÉYçàA‡®P;ß9ô¦H…ô%P9¿L…0B"AtYèñ ™û©k’J¤P>y…ÐwYCè]¹”ü˜Ù’}›ô-óê$ú íÔNNÕ<¨¾$¹ÔG: 3T^î9g¹ýÈѬÜùa»s ÃWú:À>‰ÙµÙ$–uÑ•‚šL Š{ðîU¥½že;“w¸Õµ²“1obG¸¤{«s*s²RSkW劭f¾…9ÍBš¯{ØÙ»FŽ@ùbÝK¡_âUCjÑŒÓ\øzˑ˱Œ)§çQDŠ|5/šˆG¶Ô)ðJsm‰Œ` ³KžYÒ¿‚]:Ë€yãäÓQß´¢ê1^Á÷U./Þ=òjäkµ’ÛM¤Àë«vÿxß+­L, ¬\Þ<Œú$3ì,÷ð‡ãþF‡)Ù®€úÚŒ¤þDØ«'’áI42ªü&XG-N³ÎÖ\X‡ÆÿÑ$.Û@wµçÁ¼G®Uñò7ð6¬®Pî=ÁlE ¢>~ûa%$ûéÞ¤dµ~C2óÜ]^GåeSÊŠŒw^£dt°xî>hVMÙvqZ³Ø{¨.#’‘ó¡`âMy…°Ð]Ô`âéç4g8÷úŽãJKMåvào” °:à¼ë‘¿$@–æ¼’þ-[ë4cIC<ŸÌ¦7‚ÛÊ*¥!q˜Â’wv,˜µRªÐ ²¯.aÔˆcþâFPk˜L%Ùû\("¹¢Ì³sbMŒÕ’,žZ„(š&Ç8Ñèx-g ýÏëã`NßµÂcšýsè—®Lè ÐÒÚýÆ4“عï3qG¶{E1Öè¨N=mÌ™BÍu©Óªò ¥œ¥Z‡{#ð±8(W”­7¿+1KT|µÜUùéÅAV‚2Î÷7?çs¤ýx´Ør|Š›ÔæKj3Ü’Æy´öýãòUÑÅ­¢iÍèõ|ò•Ñä'›ej:¢‚c˜ìn„ m‹r»ƒ±ú«\ìs\é’8Ãþå|Ô‚¢Kã1\‚Ј&ch[N`¦Ænµ'¯ŠÚ-ØØ qd“Z[†€ƒ¸g.»u$ÂÙÞ”ƒ?«‰ õÀaa65Ì Ñ7ñ‚øYõøùS¬åß{[-ŽŒ±x”‰$žE§›ëU¬GOjY­£H»²ýçkX¯ ÖC à!_œš–œ 2†¡~¡­ï7|xHxr|'üÞμG¶Xk®éo®Ð>GËÊc¿¨} -y-aÒ•`êT«ˆ‰Ý0Ð>‚ÚþY„+šÃ=²Ï‡rÖ“4>Únõ¤.%Ðò1Œ–ZÒðåo”‚¸“" 5£3EÇ0ÿìÚMR¶ÄqÈó0°ï4k¹ª×¬ØÉ€ÿ­Ä;êok˶1jÙõâ»@p$ í[Oøàv]Xü¼¥w È¢É×NŸg…s!§ Ül^dMœè)G]BWœ~Áje &?œ°"¾2³¿Ù¯™qçD}l›TŸXnB†¦Z_sÔïßÍïEö¯Î®ˆHwвD{]ëô_ý.¯ºÞ žÃU Ñ0BÛþòÞ@/Ы€ *›Y¬“ô…õ _×¾ôÏ°{ú£¯½ëâ_÷vñîŸ7Ëšð¡óÚ%ÿ‡ /úm8³N ûîZ¬oÒ³? ÷ž²ên».óHX)Öß¼to-ÑÉ­Ö½Yì©#=»Ùi$¢ÎÙe/Êg#T«ÏܤL{´Úƒô§d¥E÷*³fÃÑÔ>Œ{ÞA,´ù@ü,dô™æPGÕ£|ÃOáe¥lCq¼R†aع+ÛA³… Ê) 0J«µÎvèðúØ¿Ñ[&¦ükË^°ÖÕ~ë¡™ÌXM |Ó›§ÏTƒ×õhŽy+1¾Ñ{/šð%Ž •)­xÁÙ‰´Âÿ­?ôd¹ÿ[´R”î™ àÌbÖˆ÷É[ÈFIiñ#‰ñ¬Pæº^<õÖ’[éír”¡`Ô a5ÑvwÇè2»/}Àƒmé:Þü›Až0ÓZ‡À?ʵkLÓ.çT”Ÿb-s…åèg„YÆT×v*DS½ZdªKOÍcäZ Õ|n3‘}]‡ÂõþÝ+µ)‰ÙNi4Àïº} œ[8•¾·™_¬+®ôyÄzò@lë™ïëÁ«àjØŠÌâ2FYìpüˆÎU”¹™ÄlZ}½¡ŠîÎâ/Ã[^}BIQÜÈì^](­g¼(2UdûÜ&ͨóŸîÚIbZ-r£ñÐà¿ >Ž/tQp†©ï7§õ¼~Uú°>s®¦@2N¼€Ÿú>X ž˜{Äν\|+ü"ÿpƒî)7yÎ\`B™ âl£ÔøHpèH·’Y¹¡XyüøÀæ>Ô¶Ù×!®E$ }~ü ‡Š:oxbƒ§ª± a1Òª4ZýWòœÄFcËÑéÞ3õ+,žMœO¶4ž\/Û§ËJÓ˜(Õâ1[Ç1 ÝÉÏÚ-;&'æäg0Í×ë¾syÐŒëûÆ€¤Ç9.!z¨Ý!J07:ž~|éKðð×áù5¶f|Œ QQÕo¯àÑítÉÿ®n]Ë=ÌÜë\<­…‘#Ùâ'È@#Glúò~s…?•¢&·rαˆ<Žpºò p(]Æ. ;emGj‹”ïö)Üø6áJãU•2íÚ9åGý·ÎÜïÉåFÀÁ/LI6YG¦äD}Þìhý*Ï)a"$„þÕõú[Y'—¶¬¬­ÈØLT«óú`R¡¾u†çTHæѳQ båi9YŠ#–µ8A~‡¾V*×0ˆÙs{"ôÇ!q4 ·Ø™*ÎÞb/ÅÔ>Á ŸÁ_[¯(´ê±F„Ó‘Â6W»~ú^&{yg^ܤR­""9møž^ E4s£Ì {¿$2±­+žOÛ¿@]×ìD²™ˆZS´Ø}of$ó¾4vƒ³Æ¨¿Ó¹Ãߩø™D-Àô%olŒ®+Yãn9eÌ Ü1‰øTÚ+òÉKA\ñ¾it(/mEø°s–#_Y{àÊÎ9ù9ËñÈ1¢u½PðÆÊ–Ö£µó– ðà­¯¢Ae1u2'ëP}§÷í[*[}®)"h!¼cÁ”¾ L{ÛbFè¢Uüv¹¥[±JÄ,{~ÏÀîÒªd’¥9¼‰Ì:ß•;D†RkŒwž½ŸäËNÛÞæ¦x[VU  –†bÛ´p;§Ä"l…Å%¡8ö¿ŸôeÝø²ÛåØœ\Ë Â^a Ÿ*|W_˜8áÞ$eÞª¾á1t­HÄŸ•@È çÎÙ[… Ñ=++ñ=¤,h¢P@Ûv¦ÚÄ3è_$^‘­˜Ç©@ûì‰á¥L­hüÛœ ßðˆ+#­2–ÐéœHª¶Ìk,¼*™§Z‹º‹föçæšq:*›ÊTÒ¢îm2\mšx5ÊaÙNé0–Èœ ^³@%ÚîœÑ€öÖ<¸€GÝ’\˜ù!Fd_UÙ¿ø[_çK#»€’\ ¹É'’ð0Wõ'\ð;Ùöøðµ ÒsºçAÙ“}ݵL›Çs׊i¯ºÝ)›QM¿Í*!¿æšƒG Šcìªö0åõæoMn é‹SK¡ 7â¿•!Í­=b„<¯>„ŒÊÂ?°)²,“Ç€*V?hrü|òí+†´îòñÃM¤ mÐ]×¼TÐØÌ)lÌòýÏF‘±-¢ì·7 Û2ê³bðòì»>3VãXŒl¦zB UýÙrw¡ø‚aÈE,’ŽåÒ F‘ÇÜÓ`qV} Óðh`§MóO¶ ©\K¾[ ÄÖÙ'B¡ q¤8aYL4¯m&цoBþ(¿×x;·£)FÝ Å¿ð[ô¥·z{Õ-¸û`‰TNu+o/DsŠ® ¶ÒÒ1°ÖNÿq>×f"…K,¾sP›´nÈ-sš$ 3¸½IWÂõ…_ŒDÉŒ½KRኻ8úãî ¬X딀ͱ@0:Íeäò+_¸œãè€zª{ûvæ_.áºcyµÞf!XÙ1åtdÅßñ:“G•fˆvSëóâ-y[2ìwq)ÌåK<¦=6aˆrÄh¥Yj\–…Ã]Ïzªû>Ôº=ö¢3yé6`@”PÆÁ.a†2D7'Ô ÐÅhGŸÙ£ì ׬¿‘[¹˜}© £[L‘;fï$ß.H¶ëxøKç)„&FÙÏX±2“aŒTj¨¼•H*%u&Údc9ëJyzwtÈÀ¼KnÝ'4ì-"ÿô²V±–34UÏöƒk+ |ß´Góá>åv¢Ý͉1†Ôj“pÍ’›Ž† >Eº„‘þô«hòÎ솾n¢à‡«Ï ¬ (rbº k6-–üâ@”&äí¤°‚ŸWF•À+¯W^Üã~ˆå„ü:*,®bÊ3È¡ÑLx (Èéç=Ø» É>Â%!ö6ÿÚä)ûñaãQD퇴ƒ¹a¼øIú2'’¿:µ£¢iývY ¨¶t|œxS”^[BJ9)~[Œ+ñÈ•‹w4tlúI±ùX‰°†IëgÈñƒâMâצxn˜g6ß?Wž»Q u£'mÌGG(éñ”¡1¾ óשÂr·¢£I‘æ&‹ ~îˆ1.–_¸aJÿžÉŠ!‚ ±„ÔáýWS†Y‡\PQ?WÀ"ï Ú CŠ!4v|³ß i&ã«u¨»ÀÃÏ2,1|úã ×±ËlÝx~[ò•lšÃ^®v˜¨‘Õzpê élGXòÆ7]ÚntvïZ„©p½Ù+~w5ËiZ1x~”d¸ÆÀÛÛþ÷VŠü?Q‡·eÙØÏæ¾bµCho‘¼ Eô“Ç×¥AÏuƒ±Q*_¬ÂJâ,§=- ©6_DQjÈ ˆrƉ³/µ€uAsD‘TÁxmúšÓÍšÞÚEdJ¾3!üþ_§ÌÄBô×j8!??0HÔ –~:µ¾?VÒ1Z]‰›j­AËÝuÀÇçT­/É.Dä¯s­Úo]‡Ó®3ÈÁÛ™´[æÐÒ_o)ñ:±çÏX•kûŒ¯F™MÇ8Øÿíãø”5¡„¸)#öD7!Ñv· \cÒ‡tM»5$’è)½þP=Ñö›ÉàéâO™Ÿè±°7W•«‚ ±Ípæ>N"`UyG³Óïs¾áåÝ=[·æ³–F„0õ›WëÃ÷÷O,´ ~Ú' •¡D¢pZ]ÇÖù»g£B1Ödm}ª¶óNNu*™cëtóI*צ: iï¸ìåÕ+a>¾zš-àIâw6³v "[ÁÉ©á0H°¨þƹ|m•éÎ JÄ ³Ögúu¨Î&ËTÚd½½  jœ¡˜‰³å‰üQ<“Çj/ÏA•yÅBg2FãjWžuÏ ìSb½ 3²Ä˜Þ±!¿±/ðë/>襒ÙH?4ìœLÁ˜(3ÄzºSÃí=‘¦‚`‚Þöás—ùéE¿GÞec4úzýIVþ#`Œ•šlã.Ún¿c>øÎ+âöz ½£ŽL¤3œ£ ÷)øp–¼—4¢¹ ôíœA¤x:zEB ˆ@d[kî£ 5½ÛÆfG¥æ@[tl¤JX’šµO;×âÓü#°áðNîG¹v 3¸› GGRÔ®ôíF¼Fgâèì²€²¦I«ÇpzÖðÉZ»Ì±¡”£”ç ñò2Ï_ÅÂ:ºýú yLð*‚Õ]ù;&I'.è:Ê&øC-v2ƒqH`x_:“.2f S^¿ š¢2Ù"ªô…üwíLˆ4±ßrQ –5ËØïû­Y“`ñ×=ñFîB#GD`]qïÖ“@²tíTbºê£¬” —›Ÿ¸§‚TRÌblÏf:„6©ý8gÉhíe{¡Å#ˆ"3Âñ}¬†e7©œ½b¶%ëC)Z~‹£’J¥îãOׇ¢ªw”Æà=1}襊ŒÆ´¥”nÑõ×)0(ÐæÔ$SÓjv~»T+õ«ÊÝv!]i±ò‹Bý Sž–¬RyºÄPf.Ý­{p( …$JZn(åæ_b}ÄùÆäÁàÊ[õ T8Ÿ¤‹3Û$ø.ïÞ&(×ɵx«s dDþûH¢¤5h½(ÚkÑã4c6Ú C›_Æ Å!€ŸR ÉàÞmM¡JL’+vO"¨ŸKþó Oæ;¶˜ã¸“Ôl¨ oöa!†)S•Ó¢Ú2¯¡Èn½´sYqGdJ·Se>k¦$\«Ù*p#ò1Ã/Χk¿¢±3siwÌŸ8:êzTúÙJÚ›"ŸŠ{¢Ê¦`r‡ ˆ/ñÁÛ&o¸h¥nn~ç&Š/ÀdÿsD8©Åm˜—Á_%ÂsD>}¬rFѲ<ÃäHÚQFQkÿ7VnÕ]ý:pÑ \’Fÿš=LÒ†ý%¹Ã ú5,¬»ÓÞuÇ]ßV`x±Ô¥rº½™ìdñüýç@±E½;?9J²[ä\©Õ1“Ú‰™ÆyÉË–Ðj]ìES¦?aØ–˜wÇÎmoÖ¥ KÏæ5(9¼»áìI„OøÀGlHð'ÝZ<£„B…u£ef¿¨š-hÈ›ñÇÅ—¤ŸF…ˆî¾æ¯ ›ÚB3Þy𠆟C·ôºGií·öŸôìÛËØÂH«]ñù"ü’†sF Š€Î…!Șö ¼€–õïsÑÛ[ ¡·•P¥HÑYm ÎÔòæx>[aQ磟­K,üªm¡E’LeË1÷´Pé§ñ(F<²­Ó{L±Æ-Û¿#Y…2Ç…W<%Ÿì‘+eÝÀ‘aŽX} ¼Œtsù ÁÆ ¡= Xä劻lJ—/‚ba/ ëK½ÇBGw×-µù­Ã”:É‚ÿøÀ=toàæ\@÷I 1æ‘òò›ÏßT|q>‹ÀÑšî½Cý_nº$z“ ˆgM-R@D>úÔ;ÿ¦·J³ °–øÁ°1„2]ƒ?DI›äãov€³³ØA7jñü{<€š¦ñT~âúSÆßQTx“ØÏ\r¨É´­YW÷ÐÒÇ©åJà$ÕYZ|ÐJ/‘1!·š†â𱯠±C‚K+‚…ýiÁ§‰ƒ“Ükût³ÈȇŪï ÒB¯E:ŽëÄH~NK½Á'ÓÕU)Í98!Ã=´•`³)}Çœ£6ÇgF·2{¿rÓÑ5\ùŸ]ÛD @ãI2…¡ð&ˆV<˜ã„&pN†Éà,SE9^ÖÉyýK6áÕ€ŽpœÈ0í…H10}¤µu7”Hj|å̜ޣ+|×jÜðR0sâÔÑÉ&^7™_êe´©ÍO2àP½y#ÇùS‚÷4o‰zûƒéÂöÙÿ*úh´?„Âé‡ùÂZ>]C{ýÜê·({Ê?‘µÝКØ*Z6ÀulmõåwN ()èVDöà~ìÓãk]09á3PÂî÷åM½3ÜßSMæ‚šmþ´84—ãžø)OXÖA¤+]dù%ëEëw8†k)È'%`Ûæ¡wQß| P4#õb—»Eì ¿wĹ=³³ ìp‰`´\¤T`t c\§—JF…ׄ·§Œ'& ìï](ÆZÐe-¢ê\ëJr½ðtåä]u1Óf™.eÀ÷$ç2AkàØn¡ÿÈÍÝ÷?õœ$cGñèž‚8"¶˜ªJ:ÕwØñ[çþÔ|åQà]+œ5ƒ¬"-Ön> ./»üéè^PÖN'…–2Ü0hyçå ÝQEŒyñ]¢˜PØLœ·%4s|6é66j¥½FHþò×Ïî=Õ©~îfn=ÔxߎUûq Rœœ:~ÓšÕÂ|´ ô”d¯Æ[€„Þ©¾:•ÙíÂNþX'ùŒ9rÕñ„qÌöm<þ¬Ô‰#v‰°›ÏË»zŸl"i"ŒE»$&†à;¾õ2CÐT¹!ûîVÍLO¸N{¹ GÉä"¼Û›©N{k®T[R ê\¯µ½:ñ±VݦŽ ~k˜š+8„‡9€,ı>Ú¤Ø)™8déÊÎnȽç:‹ød²ÖA´íÓU†LÄ*PQ/fœë$& ”jø ŸúxR¥9ÿâÍ6|0§ ¯­ô²ñI(q#aµô+ûûbxXÎ?¿Eu=/ð±Ú›Æ=ŒÖùlºÅ¿ìÓîO¸µW#ê¨O 7f“žû£`™új÷âV‘{êZ£Bâ^àxË„:Ru0èþý·/plà;ÿ7÷Y&žü”¾¼6Xò%n© ÉCè„Å˦“ØܯÆÚ”vñy"3‡o¤XûD/¡èÀ›æ""n Öí%­/JurÑ¿FZÑÿéóìÚaµÉ­¨^ rJkçu[ÿ°]!û9 iŽÀ ŠÜdÔ⡷д8:m²þ'a~M Þ`cŒÿ Z2¶BxO™“×uÍÉ ´Iq#”¡ÌVºS™úi™Àú6Í,ÕÜE !¦S)È^Ž¦OüãZÑÊÙ·¾ÙЮÌ ¦Z®m{î ;КÕþü…>¶Rç$‰D ùf¥ØÎÏU|ÿÚ´¦dé+œï€½=nÊLêÕH tÝÜ|ÈD™t¸Ç ¶2Ø/±1Ò¨ ¦ó= û¶ÇJ"lï¥%'Y•¥2 QZ»§{»pZ1‘Ž˜Ã‘“ôJò¨Œt­¬÷ô¯ƒ[f&‚Š%1¯*•}z½¼Ð›ç90tJúà®X^!] éŸVÐÓYËýxs2-·ýöhb««nC8¸=ž£FK 5V@³äó{¡mÕr£½²ê9~|Œj[ Õ½äÒ ª‹¦öÐðrÓNñÈçó¢H2%ü#œt¸0R É7!ðÒL[¥$©NP#õ³…!–†\†çÿ‘aZM:Ãͨ÷Ó²vÉ$âÄO:^¼—¹¡É`Jã®D’¿›LΑ$‡ä·¼-ÕKÑßoìHS˜$þÑ )˜Aõ— ‰¨j²mƒ²côGĈ3Æ ïH|8l[Wি(B²–yþÍ€ Âç:ñjõײoí=+ŒD¨ªwþèëè4}E¬P§k{ß sŸ•ÿ­¢€€*ø‹uîë†rz^Ë*ÂQAÓ}õ¹£„MúéYv]£0øAƒÃïÄÆšì4öÛžI öú ¶wÍ”3ªP5öÁµû=19¦Y×úJ:º{§¤Yw~ˆ<λz>´.ï 7ëyŽvS5±B“ÐNUxǵRæ}­÷BºŽôåpˆw¥YÃ’‰·²­Îe°E.Ï—ÆÚ¨BÅøiÛ²„e»JLþÖû²Xضöïß-® ŠF(ñœ~e®ËÇßýðä©°ºcÅä7à×dDùæåʼn¥4ºÀ·ðþîkðC/ÍTŽÒýt³ˆ^Ñï:b²+íª:J9YTq Õ ¬OMŸ ]ðhŸï!¤T­#-ô³D:êÂzóûèYLëG ÛKµƒm ¢Æù)Ð =u­ƒ#Ü÷ª™ *Ü ” >7Çé“€{±:‰Ðи–Œ ÏãG-Æ¿2‡½boñvÐÌýc"+($µŠç²b9cyÝuQŠ Èlƒ¸Ea‡bÐg»•,ƒ¹^'"_/º€?µíZWçÕ)Çga çpéµ6g?û_vÉ#oô^=TÃúm.K…†–)ÄXQSJvºmÂ?ßB4æÏ,”ãÇL͘.Z–uÈÕê*u@&øØÏz;m¾«‰ÍHïþé™DJ7´èã–&PÙžÐ# B‰jß,4¨¸p; m@_ !ŸùµÛTžoõ:ç•ü Þ‘¸ä>ßeùVN—|ð}‘Ùü"Ò¨à¹æ€‘*P󢚫Ë:âí!œ’ aà-zøÃú­ÉÞÈú‹ß»1n«BÆ#vÛ·>9Å©X~aL~ÿm¦¤ÏϱMÆRQ¦/.\B”ú¥º Õy¸¸^[_Oâì…µÓº/ê°}xWMx÷Û7È·aäEÇ$ÌÍò¥‚á 7³v]öФo#-u¾ÚÛÁ¹Ž¬iù=wÔX=tåBZI„ÎøtÅ5${:1s¼ÇŠ(4•.B3#î•Ëôj 7"*«×¡Æ¤ïU®dc65}³­ÝžẢQ'máÝÎa®“ Œ,Bô,?EV´èa¯¿ó™UÐM?s“DélÓo¬A…CX9ýŸ}ÄìCÆÅæçµrø>< _ƒÅpã&ÿøŽžë¾Õv¿öÿýBn1.?\±CÅþ笾{ôÂwUø´>¿@0U¥Åe•X@<[—t.Ω|©n}Úå•Ê3PaXûÞw‚½ã¾ÛJ4Xõº-JÊõ¡œžà:c_çïæáyã“ÊK˜¡\Ø-1âÚY\ܘ: ¹\Û [Fߨ—Þß¾*'å½]–ÈuƒÌ‹îùƒm×u“‘‡ ê½VØöÆ~|HÀø]Œ}y\¼@b_Bn¹@ߨ¡"á!¸/øn’š0ä³Ù¿j]ÜçÞÚý³½7•£G žÞ#ƒ(!„é—žBw RÓúì…c¾1ž^5ÙÓš¬]‡Õ!jfr€‡gz^+¥£R8AK¶G=Vî!Þ€l¸:¸ùmÁLLû´À8^B¶S·ì‰Å˜ì«;“°ö¹@Ì—'Ž³˜;ª?v7ÔîsÓÏìÈMÝtÂ#+zå½Á¯¼ð â:·HX‰)ßy¨ýÊ„—âäM3êØV+Ô­vµœádÛ¯Xžò‘**{xZhéÚ!«s›:ŠÈw—~“…î][¿æ^è‹îÒ@gQïl¹ÔÙro£ÄA 'fr¸ƒé)ÀmðÚÉ®<Ÿ~â´­ƒ‹Jq ‘xM9 endstream endobj 1181 0 obj << /Type /FontDescriptor /FontName /BIRPMW+Sabon-Italic /Flags 4 /FontBBox [-146 -228 1059 927] /Ascent 747 /CapHeight 706 /Descent -220 /ItalicAngle -12 /StemV 93 /XHeight 457 /CharSet (/A/B/C/E/F/G/I/L/M/N/P/S/V/a/b/c/d/e/f/fi/g/h/hyphen/i/k/l/m/n/o/p/period/quoteright/r/s/slash/t/u/v/w/x/y/z) /FontFile 1180 0 R >> endobj 1182 0 obj << /Length1 725 /Length2 1025 /Length3 0 /Length 1586 /Filter /FlateDecode >> stream xÚmR{4Tk¿©¨«QŸWÑ„“ä•Ëd˜BƒñÊ«™9s˜9GgΈ¨{{7¡•ÕS/e´RÒ“n·§ô’D7žÊ›ÜåÖ=ˆÖ]ë®ïŸ½û·×þíß·Í98Fò“â!ÓŠiÍp>Ÿ±ˆÁ k–¶¹9›€ű%:‚`ˆ€˜¶À–Á`j›6ŸD 1bXˆ,‡A$ ¨%@.ÁP‘8'$$ü’ “[òÅÖTS„€CJ `/ç…r}=…§o ð„$À“ %¨x£"ˆÉ %ˆÆ I€ÇtH“ÌzX– ’MàRàãÁwã,÷å¶ Ÿ ¼½¨vŒ”9Rl("É¡íær,’Æ"áX4V%G"km& ¨ˆBƒbÚ6C¾q±h,yüh‰$£ô ÊCK€ÀhŠ-—H|R,ظ4^NBøà$00äýj¥Qª@ŠJ’þ—üÁ%”nX ecBe4"<”¤ü& 9ƒáÈßø@•KGÚ!å õ«ŽI’¾O¤VžgêÁ㯠ÁpÙáŠÅ€’òS@ cÀp™'@G.hTËèEæÜG@h"cXcèF?Xîîxb²•-“¬Xv€Ér°‹ì˜kÿ#W$'ˆ‘æS»ŒæÃÇa"i7Ôá"§M±ùç6­ó(|xbýáý•ô:¦ÊÓ{Ýazèu1õºÓ op`µŽÑƒõB—–Žfn_™‰ ê¼½[iœQg N…~í¥/•š­²ªýè\5õ'¦¬˜æ3…ƒÜÈÒl£Éò°pÿQ ~nļnn¦b|àùyõòîèûJ³·üDË~÷5â«çà–BŽ±Šzð78¿ßPmÇßQ‚@û­µs®¥gâ~Š¼­;ƒ_€çÎ7rf½G ˜oÑjÝËÔõù{ž<ì×µUXå¦ÜŠ> ¾²ül›ž^ÿ à|Ü-•?½G!±ïå¦T…èá¯ßÑAênisñþUüYK3µÇEÙò×»#ü ³ÀM¿ßùЬðîíþRKzÅ4Ù_~¿ó­œëéò\aà½0/»Þí7éq+š÷é¢ÂÙñŠ]N=ªe¹¥Ê|­ùØ»— nú9×ñϾŽ+·öM;†ðÜJ·˜¦^½¦8E‹®Ê)j9|%õÒÑ›9 =!¦]HÝë¯ÕÕH§p2…eÖ”>hx²oõꤓojh>oxgÊvIË‹ÛU#Ÿ–L2]Y9ËùÉÜÝ–Ýû›Î+4’瞃ôØO²â üšv'ãü¢Ëú}×5{¦šmÌŠ]l¥Î)’ÆaÆ…Û•ˆÅ!š9Sn‡ìM»³m±+¿êÉý‰úï 7>"VEÂÍôW/ßËÔWWS);èß;C¿f­s“Pìòª=v|dÿžÀ¥±f´ðÁsÂÊO9„½Ë2Š+>~1û±óúöÖe—y)Ód4µºæ-Éíƒ?é¬áæ9,ÿ8`÷Öao)‘- ÿÔæøbâ òõ^½úÏWØÏJÖrK×ÕœfÐpÄvÃÍ ÿ\ö½bS¢µæA‡š~ôÓ” ÚAÕ­d¡{ü¶–æÒíƒa1 ðÑ3osÚcuAº_B_ê]æëiÔš=S#/ƒF¨Ì¢.»MŸ¾<®±™sèîÔÛµ˜Ü@UC›æÂóú~%¤æ}AŽ.wšý1m›‰³ô:ªSèìNö.ÚÕX½¬Âéî@bZ£qcŸÛ§Ëg5™UOr[×n¿%n–|qﯵ=cu¿ <ÅÄäØ=¿˜ê,æé endstream endobj 1183 0 obj << /Type /FontDescriptor /FontName /SYEPTZ+SFTT0800 /Flags 4 /FontBBox [-213 -359 1395 851] /Ascent 0 /CapHeight 0 /Descent 0 /ItalicAngle 0 /StemV 50 /XHeight 430 /CharSet (/twosuperior) /FontFile 1182 0 R >> endobj 1184 0 obj << /Length1 725 /Length2 40170 /Length3 0 /Length 40661 /Filter /FlateDecode >> stream xÚl»Stnݶ-Ûv¾$3¶mÛ¶msÆ™±mÛ¶m۶̜­µ÷>wŸrËxi,­¶Ú[½AJ$foç¢âé`ÊHËHÇÀPSQad``0Ð1Ã’ ;™ºXÚÛ‰º˜rÔMMʦF&# )@ØÞÁÓÉÒÜÂ@aLùo#@ÍÆÐÄÒÖÒ  focïfilàqsspsv¥srå£û'IÙÔàba 0³´1Ë+hJʉ(ÄåTâ¦v¦N†6W#Kc€Œ¥±©³)%ÀÌÞ `ó`logbù/LÎtÿ`çfêäò03'{[€¬¨Š ˜¼œ @T˜^E`hg‘ø'ÝÎÅ™ëŸhSc—uGpùÉÆð$£ÿ‘þÇëäñ‰†‘`biì025·´ƒ¡ÿo’vfö¶ÿ˜M\þÛõ çð(þá`bjöO´«œ¡­)€BØÞÖÁÕÅÔ kobêdø÷îN–ÿXþ;ÔÐÖÒÆóÿ7ø¿"$] ÿ!CÐÎüúþc²t³ô05Q°tù‡o'WÓÿ˜ÕMÿ³6²¦&–®¶ÿI7ý‡“VÕÄÞÎÆó¿*þÓÊ¿ëÑkª‹Ë QÿÏüÛ-jglobigPvù‡OC'“ÿ1üÛ­`hùŸ úo,ÿ=QÆÿ«Ëº8Yz´èþ5" ÿzþ[Òý¿QBBöÞ´L Zf6#3; €ƒ™Ã÷Á5vur2µsù7éÿôòßú¿‡ÉÔÔÃÔf}ÅÞ˜;Ø*µ%´ÜO´p®œznÆ@•z…xuºçVÙrÈÂDÿèçA'Ï6ÀˆwïjWò¡™Ø0Öîdr*Šèî Î[1I…¹_Íö‡bþ{ý½05l[c9°Ûð{—ôõ@ª)–Ìœ‹)ÇR^dqÆ(ƒ²Ž~p¤¸µÅ';qJ]r” ­¿p¦ÔÕHs!7Füoûë ayA{p·¤ØúåGaZùõ×VÂÚš óŒ]˜¹Ú‡ØØ"Û¡d;nß1C[‚×Ìûi[î‚+G:Ïyô# à·<ò[;þü²øòw\YK_Ò•<ª±Ûϼ,¾áG{³n¯~3êê:äŽþ.3äº;R ¢Ä0l°GÞ£¶û\9ÒË.2Ž›Þœ›ãkê(Tøå–'sñú[Ä•‘m‰ 7NçÎؼ6»¤BJi.›&r—l v`›&¹v§p2Âg¡ÐcgmmR påÌ@O„ôƒcæz °v±üñ‹·¦3îQœÍάÉäâc–ÌM1ƒÞ‚¬ü‚‰«l*f `¤ô—¾´kÈ—òðÔå)áçG#UOI|·tµÔÍb Ol²Òþ¼Š¡DnÁ5»ÜÆõ¢踞¤AuÈÉ3øA=‹Î莓û¿Ý¯S‚Ðü`}+s—Š&Ö VYÁ#÷ÆJ'JåÈ… \s`œçºØö¯·îVµéLæ^µ\ö›Ü9­•Ka­"(Ú‹lyÌ2ùéïpàÈ#E/}²#d¸úYôüµ/rÀF¯xISýØ™.Xçâü ¢³qVùEà’€^íÚGÝ9ß ©¢d ’+è¯ -“³¥žzµÍW¾dÌðVÏÃ~´T*Ÿ$?óÍR°®Ö`˜_„®SŽ«ÇI¸4l=בAÂóI#IjþæÀ=à€ÚêÌe8ꬠÐ>Âç-9ØoV,Éê~¶ï³ú^Àä‡Aÿ*(;ÃCÒwŒÝr“òN±[¤>®â¹í8dÜ·tîrÎG·%컾E2ñ¹{øž›%%oÀº®:rÔ—VÕ°!˜óúòXûãÊ ]U×’tåÇJâTç­g[L|³÷K,Á§§Y²ìË`ÏíñÏ®#¢ÓøÄA5P€H;rñÒ¹úîJ»¹U<犢ä~5ŒÆ‡‡mÑÖ'"ilñª¥w÷sâØ47î ÏÞÜY ' 콢ͷJÉÉ@øÝQ3s¿±¬.žc|¾æ[,Ûj[¾ÅHtxójjµõ­ã_º½N³¶ÑrÀï%´ŠkPjª[ªLèßÍ QIÿÈæãήztóK¾jFKn¶3„HìN~=ç†n½Œƒƒ@“=u?á¼ÔjÓPÇv³#Ñ™¨À[äÍ‚…'aR{¢Î¿=:ª7½–>X €[5Ÿ/÷­#ùv”Ô5ñ"uÊÚ—Ÿœ‘þ­š/¶ÖMO„3­+Q@ùZÛjF:­œ“´êØèJÃDé ëÁXpCË‚Q“â€ð%fÄ(»U<‡õë'_0#ÒHû!žÀž|pßÎP‘NA'yªºa2?š„÷¡\Î3ö'‡ªU&5éî·Úøá¿yÃí k€´ñ„bõe\¹îñø#‹535õbG9_a}•S¬ã„/„m­Â0ó™Æ_‰W Oª–7UçJÄù ç~¶a(m„¾Û%f%:y+7¸Ì_|¼B¤‰ ŸÖü— U\ÇCFªŠnXsó·ïT‹AÕ·'ß¿;á ¦Ã|C8í|tù9—e#Gž@ÉFŸ<¶óØ:”:f&ËÁ:¥ñ.!Nc¿0u>/o7ÌÉ8Âr¥.ßÒENÄb“w^¬-ÿZÍ2R¬í28ëšDgp¸¶a0v€êÐRjUÌÂ~ªX—õœžû7 1 ? HjÁP¬º*]—<9N{*f*i:í%@™ÓrÔ[¾§çpqSŒN’Í}Âgk‹U:-t†‹ˆax»$g$Æ’*cqõµ,±/±Â~SÕ].–ÆϬ”.ºÕ þ4A%³è×”´XV’öÌM8 ‘ÀCuœsþ¦RdÚ¼×{6ùYêx¡*|Ì}š»—Wà+ù§éÖöô]u—±ÃÁ#º¥<ê¶Oï­ áiÁ/êˆ[ª R>>r:ýyßÓà‰“JŠ¸Þ=«ƒk›8tZ£.#$ʇlŸhVHLø«*›Aéí¼vÍ2#²‹:–°á,7B±fËùÛ8†ïÙï¯èSW]åRu§é¹­Í“_o •ñïß5ÿ#&ƽÒ8#3’NËÏblFù“æ;-\õ}Ïôp$ea~8Ô¿®%ÙMêûVDaŒœ–«‚`’=¢P¨m›%³wdé†wÔãùõ{ÜyÛ¡…œ:}t>ʈµ†}j,îK[ìù0:`Æ´·xÌ£¥ÝŠ ‘‡1ˆ™ƒ®`¿‚š³ýb•a•à§ ¤g'JÔØ»”s$Ên¶ï÷5‚ ÝR— éÙI_[\kv? ‰žó»¿_,hð/Û¨,f&ãªÓñ@óIÛß]8©dÛ75!¦ÿ·l‹±ä3rÌ*¾œlŸ?Œeé¬â¾ ˆl!ög{²Æоluv–]Þ!U·ë‘dPÕÞ6RkŽ€Û’¢Ûçe>Ä’Q«ü_˜Øeë4 ½'¿ˆ‹˜’I 0ø ½ÝÎÓ¶[`L‹×02˜x±c¯?=*TDµÈX{À¶N+¬$Awé˜êHobúèa 0ÙíÊø0‹à<‹NÞ²cEQ½Ù6®ƒZ³ÙîGÌiZJ ô¢GȆ—Æþ™Ÿ/çªfžã÷sÃY ª×yJ ¨ä€µ£˜V·ÉlMîk?j]n4@¤bÓ´³aÏú —ÊfÀ€` Õ[MI¬Œ#ÚÐ+"r¨îÙu0Pð€§yw/ÍÅpýNXÜDÈ=yðÄ~¹ôsN6Ð™Ö §*¾ ëèú0¦ ø9Û/W$¢÷zVÞôË9îðÄšlHlópoZf¼;ƒ`[gI§ÙJѼ'@d"y겜€½|–±4;‡ ?Þ>ŽŸüu_·Y8| ƒº…ÖªçñùÇm÷©r[¾ƒ™ëÓ5°¢¸8Q a¹$jŒ™„„=E…¡Ø3G –jÒ3ļòóp”[G‘ÕXá|/é uˆë>`C f0Íí'B‘4=±ÖŠJ/oö“שÅÖ¹qiÑàÉhÙfcxÜ>*¤ ‰ç\eò¬iw²ØÒo:2ôÂu7É‚²FNÈ ÐÒhV¥q—›IüÚß¿S\°$1ÙGnÌ-ê®Òm\m]4…@’ ú0-èCð‡íììá¹ëõ"cLŽ@-Ž~®ÓË>õ³î³A¬æ0ÔÕ_äí“¡ón¿i-û=_DY¬¬è†®ÿðCæJÛ<‡ûnïlŠJ,xíäú¯W.’—è†ö‘cÉ’GEÁä n}2©è &’ʽÑ—9Û#ÃåŽjÈ»‘aaºUö«Na#ôõ¤÷Ò O…u'ÒFÇÿ·IØD±‡ñÊZÀòݘ‚ª"‡%‡:Nʘ« ±¨öbË«oPÒ¸0RºÊ›áD«­÷…‡ÞŸhvv\ÍVÿ¿ýEÆ,-¬ÊÅ*(*x¥²ÐÌ$ï5 <óSÅÖï,¸U¦.×H÷û¬Ž$påֽĔÙHjÉ}nò[ÏÀD·ev÷æfµ>o7#“÷EBúå>`×z§`ô›‚eòQ¯ä…ÁÒÂUVq¨pÕÂs­æ=_¦‹ÏÊ#§Â¶¯Q(]Aj2F(Hæ~ŒcêLqñõ7gšÇÍV‘ i¹½e®1’ÒÍ­OlMª´txwBq@K@ÁQë-ôhhä5u}$aOÌž„˜DåÐâ½yÌoOœ«¥<Þ*RçÜO©ï¤MÅG6úwêa¨eOª²b‡·)Ö—†lb¤©ÀJνEÝ7H Ì$_ë uÑfM‘ÀÍ lGr©~‰}è´ÌØ*I™6¸ÞÝ/(ì¨"4€š²×nâd–Êízn"«£LÓ§vÄó×3Äœæ×IìJ}-P "°{{ˆf]&umáX ã+u½_]«‚æ)TRá¬=±I8,á¡ærÓC‚¼Ë‘Î?l0l$hçÄ<]K–ÄHýØ‚=·oû•Q*"+;æHY”Ñ—…ðE`@°:5z'™/—Í0ÌxbIÏ^GVm;p,TMx×­u×Á#•øÉØFi‹Î鎉Øq@ë£_ïïL’¼x®-‚t¿]ûi£AÄMiívª„/i,)NoœO…YjÚÒÎ÷Þ“%õnrY¥ƒäh0ˆ—gøt¯ÙMtR·ûr7 %Áì•7­&Pù{Ð;X'Ô…ã—uã#¾¸Çxú0º»ª‡–‡”çüzýJöÖ¶Ø ÕÑZ„ŸêCÓÌ¥‹VˆÈø퇥ymœïmyé«¡Œ›£#­5jÁR HDµ9Îœlº*¥3t‡tcž}f\êÌ_Ëo¿ã¾¿c^D‚‰AŠ4®;8œøR®ô~(Âű*éÌ6¤áµøÔÏw’9í¿;°ç†« ÊäùsiÝØ_¼ÆxÆØ"ÅùÑÃÿº2,#R%®ºb{7\ÉmѸ§ÈwY°(4ÖٚĖ–Ïæ¡Åac¥””Ó#¿>fóµy½/7+z¯*{·“ÞÕýÍâ¨ù²9ÈßÍÚ¦©lÂQ·œÉe_‚Ìœf°PcWÇÃ%Œ–¾Þ¢ƒû{\N;0©´б Á8û ã‡:•,µÁý»ÏÆÏ…3@¹ÁS°ÕF¢6É„gû ²'!Ú{áø³¬õ}mæÜzù´–p5Àr+äwæÐH¹ÏBŸyßKî, ²Õ~.i{ƒb¡ G£"z®ÚYò$‘(©ã@d¥MF̾¼jùêü29¨Æ¾â·m­Éàu"{%<Á»!WÁë–—<àU|êBý2ɳa×0®«X¥•ºàÂ&ç…Ï qÿ!ÚÔdØ%(Ì(xgÙý7ò‹Tû·ÊmpÓ̬^-\gÖ Â£h;Ù¯Û:æ<ö|zƒSÁ•Ñ=»é[Ć–¾£M Hfyp‘ª %Ôœ‚©úÊZÇB]†ˆ”àv†|Ì¿m £²˜UñŒ[7t ,Ãí…Ö™ñJz®Ü5èda4÷øV¦o®„ó{·Ïæ _1½" ð–õŸÏn>Ì{³iaÀwCu/øí9Ùbóãtû“fÐv‘­šÆð°YÑ#‰„Qo˜.r¥Pª›C“ÀÎr‘•åÌÏØ_­¬@»CJy(ÖÇ®ÈÏq,Ÿ†—þ¶È‘;è çÚâ¹Ñâ{¤—‘ñíÉ °¸÷ýúÏìÒßáD3y5ÅÔ·ó¿ÀYÌ› ºvªf‚ ×pæú,Â0 –Î$ÄbB5¤ÛrsIIùoŽïŽòt$46ŸžY#KÝ Dg‘ =Њð¾³êë-¥iÕŸWH‰c7©”™z0NáÖæÈÐDàññe“ìÕ+IpÛ°A~\Ý´ë.$ȦŠrþ­2qA›Å¾XéQ$rtÃ…5°‡z"ÕñË(#†n?Î+±k‹¹ˆ•ËªìûªLGëÊ,‚]ë]†»ç7»;‡h{?¸Û·zõÜÔéÒ2–¨ÑsÀ{]¨2´ÇO9;¥«ûþHˆ6N\STaþ)ŸSã×3¡^q¨®øP…ìéí!H¶È²¬øÔÑew¶!-1¼d€‚ÍÙáZL£Oã½s ô¥½re'ë©A¡uõÒ­…á8…!?c–õI~Eƒ85)á ´&ˆBŽ\q„æxípUñÏÿÐsˆ—äâÜÒ ½¸ëÎÿþ¾Gñt_•M* ô»¹t®hûiaxNâÇh1„ûðÿ!ù(Û·<|ûÉÙuu"ÅHå Ô,ý4µ>snK¡áúÊžI³!¬jÞ4R½Džúƒ0ÕtE |­»M²÷z¿7>_I;Þ1å~åf/—i‹ ?Tei‡éÕê z­>Oâ\SË)ϧ&zy¬“w¬@È9ÞALV[”¤%c| ]1£,OJ@B—wŠÈ­· ² ¬ÇÀFRü É1l=¥=Oägç† 7ýÑ¢YÀ ÐÌæ1ÒÛH:øðF¨ ^¤[@K¾^,—8&&ñô§²œ…jÈ.]:制]„K˜ð”ú*…êÝ[àfª“-ªрš{kKŸ•ˆÓçñùîsTtâƒò#;æƒkûõË_íþž‰=·âäÊ£êi@^OÁXJÂæwènLøu+DYcCðc€À§Ã[¦º­& bÅÌ)1̯Q,lBßja/Žü‘ú ÞÈÞ¹CpŠ1ðt³¯?Ì›Hû\Dïòl[RTÒ›ù‡< Nµ0{'úïLn^´ z-. h=TÒKýä ð p UxÙ÷¸µ€%¹kQN„l„%ÃS˜“`:ª ù1C "Pkí"¿ÁØ„hl+g{x…‘§®Å¥ª¹ï\\Æe'LÑÍÛ¶–‹…YîAgô^û”ùêÑH‰€áK•„‰é g} ¥‰ÿâ-upø¡òäíÖ!¿Wñ‚µ™j¬úÅþ`ü‘aÔ^[À®F‘ÛbŇºSÆôÂèŽWŽIh½” A»¬¾¦™mt04—CiKª/©îL]™Z^cäY­ã›?¬e)­ …’/(¹°QVÇm‘Fo‡ZÈ¿ 3 dk» œW*4²°XáNÉId[Ø'{‡m*nU*‡„ÃÄó vì–G$©{2ÆÐâI½†¼(&øLwrÀ†É/,äÊ öXê«œèù'Õûý˜;á=ÍC2½ðIŒã»ƒÌSý–±ãÐ)Îó‰—ÐCÙ}»š¡Ç{¬ÝYŸµVŽX%Š?yƒ€‡g¤öQ”•±kžk‰¡QÁ|zÔcƒ´ÏV0zÖZ°Ržf,±•ZL0?8¦ ¬ 'UžQ) REUòâÁH'¸¯e:ôo±› ó~ö#--ó„ïc¯>úPºN|“uê´vúÃÒh_2˜ö“>0 ~xÑðèch§ãÆA¦ïÝ÷´6æ™ÝLÖ'Ë ÖXÉòï³]¢%ñ9>±+-Sw¹UÏ»ZÓW2žÍh¥F³›ˆ#B Ž:¨7I8¤E‘ø±3C%pw¾õ´QA_|‡#Rþ`ÐÇt,ê-ÄŽlÔ¾ÈݦŽÿ8ý†zè¤Ø™—ò~\N-™4ÚfÙÏOXá#¥&ô°Üä›o tc/‘³ mHâ¾m—Tɾ¸…­‰ÃS §’§PûJæðˆæ~ï&-ï|vm¢iF–X^/žÓUCT ¥frÖým¦©4{9ЯGs…Í’:½Ô™ÿ±ýÞÖ*4'1…¨…5ÂDK&Î>1žr<«aO÷g²(X¥‚VìðŒ·õÙ«ƒÓ8}èæÖ¼U`„4Zþøõµ@G`Én&Mã!ÛV´—¸F¾Ký‰žIGFœj4ÆÒ÷“ø2÷\Qˆ7ðÕ˜¿ÞeÞãi]þÃáxÕl=J×Bc@>¥ï#‹+ÞÀoµemÎ#/­ ¸ƒC;ˆô /šß‡q¼­#¥*IÃÒóißËÃ3tqÇéH†/IH¸.aˆ66›GäÖâh/á×›kýëøRcóœZò!¹v¼ý›æ*¨Q"~@²8Û}ä×ïXgÓ²ÂéBæÝR`ŒS% #üN(Új»8ê²÷Ý|M¬…ï†ÔŒÍl63±E£{ÞÞdD÷zÍÜ)–ˆazy|ÇDD„{¬„>^†:cãK¶Ã^”H~Iùð@÷¾*Êèì¦(Ýc–E›:+HBPú`¢ô° ¥¨µ1ÝŽSšÙ(1Ú!óô2£:¿j Þn Œ Ù?~œ(àZïxõÈÞt‚¤X[{£Ë8†ÑëóàØtëxùÑ°ÕhÜ@P£‚þ*ã1ðSžçSb»™Ç,3w9"`h>Dä +Ap“0)°&t%‘¯â×Ì©"NñUõHÄ‘'¾(¥Gð€˜Ëý—$áXópzöB,.@E¸õBݤŸ‹ÁB؉6íÎw),•ÔóA°j¯‘ ‡¬3Á»-ÞYó}ð" Úbôë-eÍLV¸ºGi¾üwo úÖAJÆS©ø”»öç…o✄Ÿ·‡B ÿ©7ÇŒä.'êN³´Œ2:3;ƒÓ»ëêôyIùKT-x4u„¹?AÝ$¬!Â?/kb3©†A§Åïíb ?8Ð#èfèÕ?ºÒ¸±-O¸?€ ŠÊVûö‹Á(ãÊO»âZÑMj-£–?ëžï)ŽÙN¾¯iëVѹƒˆU~fÏÊK®B‰rÛ'–l ­ ɶ Á9},°>¿ìøùÌœN§§…ÒÇ"/<d¿:™#ÝRSCž‡‰7EwÛ‚ÿì“O Ô-|·‰v ¤#›ÑYfÔ³ym‰sÞYz@Ž&ò%à| P<{Ñ9Ö *‹€ST¨xëÕÊõäuÜY¬Ô•YÚ¸r)L!¸¤ú̸·RVÑ ¿'TyùÂà1*”ÒÜ›8;æ ¾Ie¢A6qÏ¿ç'¯H<ÎæN‡%e~”mìà°ÏA§V$5_1NïZCÐutß\q]=%(ÝÏJx’’.<÷'$š Îh_§¨&|77ŒÜ^§pÃT¶¿ù¯vv…£ð?ÒúNw;++9ú¥8ðýôV ‘AùÉ«Ë ¦Ã;:Jh๜h‘+=SxßöA6á¯"EçÜܵ ÆkàÝKö·l6E4Øp“%þ8=Ñk~‡l>7ÝGø¿ÍŒ­ 4±9ý¬Š`;6f½Çäœ>ù² á¯zNnÛmì[O£~Xø•À¡l‰¼Ùr¤w“Ë;†2uåሠ/ ‘…h+¢)—+-èZ_ýqdjsÊq‹[7ä*ä !3ÒÚ{¡MÈ°×j†E40~É–Ø÷TÁ#[ôIF•‰j-s¶VÑ^ÜŒ’Ìÿ:Ëiu -M4DÏû19Q, ¤ÔÉÿ† ýÒ&àõ/•]tœ|+J boà}íì»q[¤{DÑåâ,€ï¸ê¥¸2+Û‡b“—nè)¼0Áx ÉOþ~ö¢iëòÁä«Eÿ6½d,÷oÉÓ6^Õªdþ!”}ç9M”€S'ѹ…ÍÅÌqÞ¹xrbœžw•“‰uÒýêV $ ò?dÞ¢¨ëì›ÁŽàÇS^©J?ZšìVoád·Þ«tbfœåK0“#¾„FRvdWZ+¨B#CÃNüE·ðÙ˜eâÇFÁß0€ÆÁàDŠtÕ£W%lzØÅw™ÝÎK•ÎmîˆÉ7P ©þå o©·%ï*¸Wkøl¢ÏÊøk;LjZA5Íèv༗Z,g6.H××Pε#YªÞçÄ“’u²z€ÌIW] î˸3£J%K­p­&FÕÀ¦ÄÝ„¿³ƒÂòíö0Ï|w{Y  pÖ³)3ËÈÎd Û‰Øü´m¾í üÿì}¤¾Ó·(ª^r/T´¦ê;¤ C¢Wâ 2ðVÄwKÇßÄyŠ‡V|s3™¬ä¾Ieáäî?ÇyÈ:³©8Z^}Щ%ݨµÉ¥˜‰Žºr=ÏŠáù²D-Í ì?®§l¦SÒ4RzÖC½Mœ£ÈÙ–5¼gbQ’YMuôrR1‰ þrÆ>Å7Dú¶É֙Кá’/p¶ ©Q`Lœ€ïØh¦ï(Ûÿ~Ö@Ó¸œ¸y7œþÎ6ºˆM—Sùd´Ü åB*aÛìýõôÑ(ô$ºâà :,üxt¼/÷³AŒ¡F3¹¼‰[GœJŽò4¼iž"4Sø•öç/ÍR÷[•º( ŽÈ…+½LwÞuðb4óM±æá 1KK}AC™ìŠ’æåï&d–ƒWŒðë؉~t×JE¹“344ÎMÓ®Ó6­m¢å"à’*–‚Ÿ&‘k¢Í¶dæ›~uQäÒØeð“IpêŸlÒçog²{YÁAÄNõ¢Èð[Äc‘½YÕY¦K7¼Iʸ«¿]òžj„³o#l¦a3€ªÛ…2«ù¨áþŽÍ9‹‚äÞPòk)9g©8¯Ýüû¨Íæü‚_Ù‹ £ªÛ(–”vX£KÚÐÑS©ÕGýH¡ƒº¤Ï7RbäÀ"I%óA5X:ÒaR¼š¿ÙR븒dÄ÷ p·E5,W®Ø ‘e<)ÚZ“ì–Õ[|N 3CWk è6­Ã Äá²{×!ÎÔV|PšT3u¦šêïÝõ…[Ä,[åŒMU<]X^”x^[˜§Œ”Ý_Sõ„ÅOEãÓ²jGŽØ %4™è¿ÌëóÀáAî¢ v ª #±ŽA„eÅBÖ+,ÁÝÈ64V¬q™…¤•lDo*Ýá3µ¸ùµMW7äZì˜nŸža~îùæXc.`m;³QÍ‘éq̸¾P(‹½ü9§Ä|w’¹uÚnü‹cê·qœw‘ôô¸ð’bH­I¤¸Ühdã [*>Y¥Ã´Je8±:÷6¤ù€]AY¥HRü®·ªãf-ˆy”-®ÅoЗÆþ8€Ê­š$k±¡üƒ[½lE“üâx… -z1bFß,p ¡Úï¾Ë熃Œk§Ù³cɨD!» dÙÕ2ÚMÒ ü/'*·ù8S†ŠªÜõCïU)qQÔn³~IP"ReQ-ÕT^¨/ýkS/_\:µÔÌ3…@ >’Ä£¶2ßï,9 ”qÁžQ½a%îÓÀ÷y\‹Ñû<1©9”;THÚžvˆQdy©•žäL…œ9nv@­¬9"A?H˜Î9l¹=[F†¶çvððök›ÊP/[ ní_Þ‚4âG»Cˆä}—·K zɳ6š¿:úQic;pý½¶ëœåòŠ¼SÅ|Œz ²Ú9Qh”McccÆþÇ'9Ý7óI5¸xÅ’Â4\§ü[”¦Éû`èßßÕG“QB~:ŠŽÃ$ê=ƤNÎ ²ØgÖmIY-¥tÜ÷GO属^¥_üþÄÊ©!Ö`ÛHÉ•V\TŒX²1J-¿ü“ÑKùI©Ç²UŠïHkó¤–(÷ïÕ3‰h|:aNøÌ’„#`ª¬bghgz\p:Ô„f1>’m^8mK^¥šlÒ%üÿF®ÂÏ\be¸†¼ëD®ãxG÷ ¿˜±š±Qºé1¬0}dJXÚj5¥};»O»ÞÂigBTt.«ë–u‡KÃù¶ÞzvˆÁNèĤôt’BW„{pÊoÃDé+^]©nÒM^eåqá2"ý‘49@Š¤V)&©_䄪Ô(-^›Ý –^ó )7×Y6q. àÍjwý‚^ËŸ‰ z¡/öcNŒ8°‘x÷ôÅ$c[.Zä‡]nI|e Û/§. LÈ«Bã܇Ÿ]ã•úÙºüu‹’zǺÓÁa@L Ý;˜[2\ ñØñ… 댇ŒØ´F aym·Cv$ Ò&P ¶‰E¨è°Êʘ4i2† íG8k[0‡ÎJŽ?¹‚õŒ«û&:J"µÀð„iÕµebàˆå¸Æ’\Ÿ 5VÝ÷)îgr_»ÉØ´$¢MÈÃ@tÿûë³W4AaÍ 6ª¬1D¬{Ž–m¨:§ÑÉ +n9òÝË©s÷5É&©Ñ¥W‘ëž„}rÏe¹lr—l|÷èHÉ…¢ìD66 FœSý³9#@”yž¼ÈDð¥jëaµâ%ªu¯1AÞ£½‹Ô¸’Û‘@MB 3ÆýñâÕ?Wä’X–â%M"÷Yñz²oƒ´[ªM‰/CaUÅÌâ|nAG¥ùVo ‚ÕåØK¦œôäftðCï.èzý³Ñz૤a3„¶7œÒuzaBP)²D¿¡Í¯FòœDší“…Cµ·Ï-èÌõ¨N1â{¶«¢Þ-9 }Å\úµ4ߦyG'Ÿ ,“1i`}¥• ÏwBp«cX_¦³½e´lEvž¦½Óï¯D÷Y¿"lDã½ÛLþsç  1â»S"ÈÞæl!BôÌÅŸ×ÇÅÄÒƒá.ü Ž0ŸƒG¹_(^âÔC gg£ÓÞ™¡h@^þ¸ •‡<|óçO‹Î7–•j¾†< jŸ´†&öžu 4¸ Ò,×Ô!/óPÙnŸ8#›8$•5PŠô¶Ÿ7ÎàYÿTÎ$añ3‘вäAk¸>ò ûj1%”U¿ÛÑó4ªŠà‡aÞu§pµàÞªþfÖïÂX]'Jf;=.!@N Ebu„¬¦‘×|ÃVo…\ˆüfG’\VÞSì¡hxaÓ(Ô6†ˆ’®’ØL}ˆZÞ-†V6èvÛ`ù`µãåmËÈÛíán9)“uÖùÚ· mt{¦¿±_ýË m_ÿˆ:†È÷ÛO'~ºøAµ_L Úi*Ÿýj<ÊjlÇ4{ßlêÒOÉvßßD·ŽƒQ8©ðq[ús¶`*ãáÄO¯èƒü[Å!˜?¤)ß4KÆxéxšÒ‡€ðAZÓÈ`à".|ÁL•ò g !fWéWò!¤^ðý^C¡e ¿£)ŸØ®³¤NAr%s½N 2?R¡ ~ÙpUQ¬Lem!—_šMä‘ÓÈ­m°½¹—õÀ5F ÖïÞ]Gü¤«d²id…»Ü–\m^-hdg œ±ßâñô °xćP¿èÞÐnÇÔ ‰ß}Bƒná!F¸›ï?ZKÒÆhÕ¿.Ô@…ÏiË]!àÀ¦¼ö%EZÃÄx!øCa¾À|ñb‘€Ý”h¹X™Rž®H”•­÷r‚­‰Y Ñ=R§$ Z›kò"Ù¼X÷0hæ¹-›?«õÊYΠ–!øÎ_+‡ó ÑFR”'‰ž@ÈL¥km”ÞÄ}ýnûc Z *½ÀF`625÷èqæ[ú‰1ˆÖDÞq£-'`þkìÝ&Š×%ÞNZãwvA‡.n‰QXæÍ9nPí}nSÑpòÓŠBp“Òœ ŽñHœ¹Û÷lL¢¥×æWˆäAX ›-ŒL^ I{MQÈN¤_±æb­¥Qùò»±A =œkÙ æ2¬ú)¯Õc?ŽëãÀl¸&RôÒ%:BëÉQ˜¼RŽÔý¸$ã€L(üà…ÿó=Íç*wŸ­ÿ÷3ÿCü4ÚÚ‚Æl¾Dû¡º¾W3¯Ø@ò:©GU:i;ÐYñã-ôû$O.®Ê:œ(G•Ò.þR¹NciÊu¥#j¤ãëµùÔö@Âq×Ø¿]ÔHª Jäçbiic|8ùL®(K*ć§ßìæM[^*E`ëH¨«lßëQšç€l¡[‘˜Ï nÔ&ž(§Øü"kkÂ3’9Da¨ûôiå@h1¼~Æ"¼_¬”g ö†æΩ äæ~LY÷ó[~î£D`! oªtÃ":BƒAÂopü~c,èï¸Aãp€‘UXÑÿÚÕ%¦­Jä¿Ïy×·nt%/áMäÆ)}ÉUGåΪR®TYk`fžPÿ#mßÆtøÔHôøØSˆZ ¡?«‚dû¿§èhŒ]…Zö¹ßþ6;ᯪ9ä_%8:y²ávDG/8>Ž%QŽ„#9 “_#oæ†?.~^Q¼Nª–`N%"Ib ÙN:–@C5ŒM(¦È §E.©e4{S|[UL|úÎÂ,?®œÂ¥¼AªÆ i2¸gþ&Þ¼»…’äcò‹'`Dô¥¢y6¸h°lM™¯ Fk¿MÛ‡ ¦’Â/Òy+³IX‹rF—M• ³x¿ÜL¯B§Ï§².8‘L À¸> [ Ó˜ü(ÍÖhy©ÈÓï¢nkÎÁÑ¡/a©Ø;˜;«v¤‘R†_ÛO4ª#»¼Úvð ^ñØà‚¯ÝBÍX“Rí FûÓEí¸t™±¤Å;V÷QåãL Uå²[aè7ŠÑ1.ôõÕ*=a<Å>³wí<<°0w;ÿõ†SBˆ·šUð(œ”½ê4bsS"ò–cIDÍnV7Û™én¬rFøÓ äp±^OP0þ ×ømßùPg&8W>8ﯮ2-&&ÓŽ®æ®l+æňOm‘+lº›z%½šâ¬ÛQ–¡ˆBÕoŽT`]”6Áe|ÓŸ•¥8mKTȆßRÒL"O¥oTF1¶ Oí’ù½ßZ„é#÷ôÄ>¥ÿûÒ’ã•W-ÛÀÈMpb#UqBA‰áH*íý[[Y·rVÍ÷È(÷£$Ôˆch ›ÚÓ80 "t‹7 ê»+Ðó!ykl2X'’âëY;Ô¯ÿ­Û”Æ„Ä*ñÕ¹}íã)+=¥úúˆOܧÍr¨ª ôvϸ­lÖ-Q€'ý±~°gð‹tüiÑ^÷†’]¤;mߥ –2Q°göG§Msv¹B¢ìm›+p‡„DMá)ϦysÌqåˆÚÀK“‹§`QÑ™E€ùU)!qù@aBw¹Í0ýùæ¨t¿%-$YfxB[°/9mkÿð“ui(»‘œs宜œðvÌHÞùžk'£m4N F'êKªDëÜ›¯¬ýóP•ù”µ¬(åêD{ªawL!ž  1h,<ÔB6áºéo-ù®-âf½AÑÜ"XÇ•Ùº\sï/tê…VëOšGF (ä½h¨ñ²q !ª;ÒÓù 7˜=þ4¿7.9º„ÉušE|Ä$æ¯z;ý”ÓP—ÀŸZ´dÀNf†Ù}¬º÷búÖsI±®ÍŸMâ;%¼5 w+ï$[f19¿ž¯£Bñ4ž µ7ËGa›Bº<ÓÀÄhÕVXêVÃ>ñòõe¦à4…„K‰M=dùõw¤3—ï©-öpT-Uöâä”àœ§ýÝ^­x#“@3«ý]†•îWõ&*“]§àVënh ½Øž‘\HéìóÁü› ‹x‚ç ê3íÿ>¬SªÊ/ÅWæÖ† lúª>ËþÞ&DÊÚ`& ÅïS<™ß+£'¦ï/Šè)® R`úÊ‚= }`öí«§LoÅV¹¹0Ê |¸W-èýy“ÊýejöÓòGÿ…´ÂÒªwHŽŸÉ?ÿwm¬ŠÊè6ðÍš¢!?ÉÒ»ü¥1¢:,õ5²VHép ˜j÷âË þY4Pf“UöXÒÓÈø°§õ9Ü©Dèßhæ‰{m5èðѲ¬Ã­­^‰.Äyúž­ÚEjZÔ¿v Žî8î†XÜõô9ð¥fo+|´›'ÖKpôÎ$¿)ÚË|’³Ä=HÏJM;èõIs§ž¿ü« ØL¯’¿Ä~9+V´BA¡Š.^dÙ“<ÃQöRrì¯T ­ ÞMYd¸À#+OCù£ÒyD©]ªµ&T8ùsêŽR¦ |»Êµ?åÍÿ•‡Äc,ð]Ýð —Ÿxôñ”lmy•ÅÂJ–ïIL©½Ûfu?¸Ÿ­å¾uüe­X)P|w)…šh²µ½iäÆ šÛ¸RŽßÌ2fN´œ‰ˆ±_Äyà¾<o—ž2¥¨Œá]hòÁŒ ÇØ1cmå\TýæÉÇ0 O"𿿺æbé!ñ{ø! Š•nÝp« „“g/`»¬Gú™2àsËAÿ%uZ tšR÷ÙŠXÕIáá,$–9 e*jìÓ ‰¥¨â¯yYÐ*¼h™cKT<ÓMoši¬¡ZP ;nqK–]0c+·µ/Ïÿ`Ô7¼òi•ì;¶”e¬Ú>]ø˜î+#|åN6v3¢=}Í{ï÷·Ó[eȵ1¥ tïáÇÃàBäžño ¾¦¦$â:°5s²8ŸxŠÊNæãjïðÁÊ·ÂÖ…¼5Ï —g¡ßØsUœÿŒÇæg¼¾¤;ˆ«†……çW[˜ùhKžºíElT&Ë·¾mƒÊÁ-ß ËÀù‡"¡† Cf±,Z—A߃Š¬ŸÓ€;¼¸»饘ÛvQì8·ÆŠ©àyT¸Á}R*U ô*Óz¸x±¿u=ŸE‡¹Š+ÖúÎT"K•EY‘N磓rÍ`Þ»]ÿ²qY<°wÿó+Šk®iQËô ¸G°/=‚¾R ¥S±å$N{F]côkí©’é^³ŽT³ËK˜⫾ éXø]äWÔþKÑ£´ù7?¤þê^†4ÝçÕC­‰Üîq®ÿÛE@ Ì)ìþÙß\¬syغì›u’²ØÊÅ S{çqÈ«Qð ?Šþ9¹T|rBÍKêÏù‚ÐB£˜S…7ÛÊRŠo¨Õ•â˜œƒ€;” ¨xw7å3šÚ…ÑŸAa Eª4”f ¼We¸ëÆÜz¶¿¶Xh_nšDc5nX)JÇÊ«Snš3ýǨlº=¿¹Þ[ÛVsšR¨ê}‡3*AÀ«š– 骉6Óí)Âj:Fɘôpuñ̳øù±‹_ BàLûŸÅ¿‘Âïïéš:´_Ž€ÞNê ‘ô6[:£n{e»4¬ë¸þl@ò7EstD½^~¿=B ÂàÝ”ÐÜ ˜B†òŒ²˜ÂV‰à®EW1G©kwë Ÿ2‡ ðw#Ô²ÿ1‰|îéR1*¤Ð.ÃFQ<1…± é dX–½¬h¥ZÝ=á¢Ó¹^‚þ”Á¡-5Md¨’XåË3X—ߦn\ Eóæ„NOl¶pΣéo²À…wäÔÓËøP{fBÿÔÂêwŠ:ÿ B8ÐæFô¹KªñÔ”ÕÞñŽ^3ÇÙ=¥u×K×q`TÕ,p…é{ú¢bgZ|uTŸ ·@;;)á?‹s!Ï!Þh‘Dm—¦"µæ»>{vx_ Úϱ{8'¶ÇÝ“­õP?²*v¾}Þ¶bB6Ö~?+ W«#p&»ù¶ÛŒ¸KetЭ]8™sŠ•ÕcïVc*Ío]=êÌIÀxªb‚“I~†_®‚LGgKìGPOífþ”:éSý5ö*AN`¥¶½ßüØ&3£fo+±û£c¹*y;Ä –å’†É{Õ‚mÓ½Ac‹fq2ø­Ãàå^3Î60÷‡®å2âµ±î†Áue=P¡4‡ŠÐ=Q>Ò‚H¹MÙZ{£`V¡—”ËíU±Í`We‚¨÷÷+ö¬k¨l©ûÓÜq§À“ÍžÔ'rÄOT»ÄÞ¬Ù¤à ï[«ÞüWϺ9ÐGèY©íôÓX+L”®î71³ mCáÇ”,¾&*ªþæ…¸¾p/~†|—hú”Ùv!£¼ GíOÚ`ØŽ8s¸Ü“ I7 N‰µ§Yw" XÃÒ'%d/dº¯Çß«ÆgY¿ó¦Pzty®æ«¤ÑïÐé|ãÁaá­”q$zA&ám·fŒoöCѪ)™“ýç(ž£ÕõŸ õû·p¯c>,hO¯ãPÒ¯ô0¥EnJ7êèŧ<;8™Ä ¼ÌÿÃÅ9v‰àLºmÛ¶mÛ¶mÛv÷mÛ¶mÛ¶m›;söÓ¼?"I=IU¢Ù“=üU—–¨²­Ìè{I¨=‘m´}øëëSgn"h›ä0ÇëÃlO¼Ê pØG!MGä9j¦â(¡'<ïù ¾3B!/÷² cj¿1,ùצí’g-,~ü»æ¯ Cóæ‹’Í–hÆô‘6Á$`°P«Ýír튕¡‰¶R]#ã7ÅJºáëë¦÷‰ˆeäè˺®Ù—J•Ý˜yú]86“v|Àòç¾Wò|sþLå†ÿ´µi¾DÂõOz:cËSœÔ‚úd #a¤Ì¯¯T.šMïJ„[ M‚\2Elž(zèZ:ç-4Bڻػìr'½·í ðú]vŠ¾kÙó:w ¢^ã?@æ‘duÑÖjî(«T˜Õa@JžQ¼Féf_£¸Ñ}NŸÄ’õ|ù´/Ào꥔ßp±™@ÊxAyVg‡£(u¥Vù ñÔ„ówsš)œ˜ñ²òj›x‰R¨çØß᳇H+‹=ÀPÈKªCT1£Lk¶.Žlu9 S%"H§ EbÄpÐ_'à”aÆ(» ã슞…ã8_ú–/+ÿµA Bpª‹£áf"W¯XüÉЉFèèÜÍ ³N}Žœ6—#õ´œÑkŸ è ;5p•' æ.™¹9­qrþïÐy_ >§Îi>/J÷ªcF\yÂÉgÞA“SòÄ{«H0#£œ>²T„Ú“—5¾(°A¸G¢t5 »eDX}”«b6\ÜSR ·SÄâ.zæ×&´69_Ó”ÅñŸ5 l7²šÓ€gîÇŸÀ±õBb¦Ñ<ˆ#TTŽ$§–ýb„DÅ'õx0…Ñ?™'›s‡3‚‰Ú¡¤ØX†ºü‚{»Ái£QºØÿo±Þ‰á· ç%5sñ×WòìºU¤£!VkK‹BzK¼®r¨eô—÷íZ/§,ê½ÿ ç7*¨_·(NGHr퇖ä:øðq)Ü$â¤ÇÞîêkµÍ矪WšÓ=g—à%ý²kXÑ)a—uxå)i&¤¯Ãnu<ÑB£õôF‰Ž”‡@Ñ{:)+„iô5 ð’ªÆRãî»1Aže3²ªíÊ4·ZÍ*HAÌãüfx”Õšzýy¹Æï´¨ÐÎ3É v `Á… Jгì1'’ˆ&„¶Çª¤ÀMon 6ÛeS±n©-î÷ ëàÚ?I@µà}ÐÄk•pe-%:'m}SY/ÕÚºa.- ®¯<€D$l‘áZz½Æ o™eÓ}ÞüŠWazSß…„ãK¼Íü§Pœ{ö–\£ã”u]h.ÅTQÉ©«œ1¬Õÿ;4Ú¬> sÚ{¶i{ ª Ù¹¸ÒŽ5Ù¾ž ˆâÒª¯ëJfx°´Ï¾ýiôaC¨sQ¿¥à>Ĩ—©!½ý=ÒG² Ù †Eè&Ð_uRzûjêb®”¹!{í¦¹¿ºLe«¡s#Qêù£ WôÊѺ•r“ ‚-)W¬ž 3ë'Å–4ã˜jØÁPÃßѶ:wï~]ªm°PyÌ…žM+he vm¡ƒâHjk éÉ%;ªØŒùkñM#p•ãGb²EñùtåÓwÖ»zíKª@Í4ˆ\¤ðε›±Ââäg­\BçÒ ÈA¤!V\))»ðùËb÷\XAV%,úêrDdI_@Ø׸%]š«æpÃkÒÍO—<¸ðäŠ×xN1—Ì`èöÛŸ~^ß#fŸÍZz?÷°Íæ´Hé{² Ž%rçA;ïØÑ•ú ¨Ò™¯} fá€-2|@Η2B€ÈIÎî õò'¬;³n ¸E\þFXo·y¦Dû O)lÔ8Ó½ÈÝAĬۖÑ1œ>¤§ÏÅ#ã®pr„:;{VÙ*µ5†Ì Å DÚüÛ>î°ƒG·øœØª\¼c(=Ñ©öú#;i¡@Šµ™pbMüËõs¿Ö@µ7Ê—¹ªõþ úÜÙ?ÝI“e|:6ÜÜß”ºá «¥ûb€¾óœí(dGªñ_˜Rùn×äò‹±AuŒD¼p-Û"(Å™‡Çn¡a™@/èÛ¥Ù¬Ç@[6ÛXŽ ;ï=È=ùA$l ú9¤¥c$$¢ÂÊ:Úá yˆ´ÖÏ€0âO7U¬þu©cmÀ,Å6žcÍ÷Wcé‡y' îÑŠéáQë×¼ðR¿­ìÿÚ¾FÙ(œi›Ò诱Î!}D‰@ ò}R¦|õõѨ"mâ™g¨Iýßœu’@åGž SSÑdNAŸ®õ…¦ÛÝ1Ü={q¿ŒµÉ "rÊãS‡Äe-ÑN°€©mÒ*ª<-΢F9ÙØ76ømÇ ´qr’ÿd¯~zóËP15Råæ²›¬Vªex«;vÿÔìʃ½µ„Xþ)æqxŠè¡tZN[ÅWÿ0©íç½±8¬©½5Úr¸%ÕþÕ‹º~°é⪇…Í߈½µÞ.{åÜɇs«Â¥ôºúæw]µLù½1¾´³AÿÅúJT¥ñoòˆäãA=vàâ,¶Áè=3ÿ»ÚƒZˆgЦÄ2hÚ‡œ…–– Ê%€/-,àŲÌ"+Ö‡y…LÎu{GÓ¹±Ô¥›±o³ÏÙ¶¦Êbj^ÐÛ7!ÎÙž—ð2øþïQ£6`g˜cÁjxˆð/fAnN£9߽ŤžxòYoØøçU¡‘mÒ”Ny38¨„lc6.ðáÒËv,£l¸‚?þ PÙ†ÒúùŪaaÛQ‹Ëj Å„”aûßö¢µÙÁ6WƬdMk8ïø–å/‹DE°2×› :M&ÂþRS„!Å¢ñSt69 ׸ü…)¼ýve‰ä~’þ5!é®M˜£«\ê  KNDvœi~ô›‘½xMð^·ª£”iTvAš$ˆ\þ#'n£&þDÉÙ\è±ê7ÄKhôæùOìém)Ü_¡CœŠbÔÕ^»ɬñ=÷/Ù·|²åÀ-ðãÞÅñÍž®àÞI+œbƒôò(PZéóË;)¥p¾¼t<·ó?S_Ÿ¹õÿ²¢Ìe(¢7/É,‹¯ÄÑS^4€ø þäV …XëëäŸp§øÒx ðqtÅB|iõ³{þ¿h –ù„ØtzèÄcï1IÍįìö‹ö'L^1|Üû/<9:¯H¶ âËçV€Ü¦¹ ƒ. 5P*d¯ ó®øªè¥§‰‹½ôÀ¶½ ¤±@ÙÄ©­Y³²ºU•ðú¸ägWR-@?ÏÛnho•TÜMÚsµâlø—úŸµAûßó y' ÓJ)¦ÉÊÞØ"}u,K÷P}úš6 iýP]3&¹…YÓŒ+(Úí#r Œ(³@@‡’‰P5²ßü>‘¹P“Oå¬+±Ð_A9ן$V6Ù ÀÂÞržB*VÔ':U<7ƒß–]ºmIU‰<¹ih´<¤‘ ¢ãmê©Qk˜ˆDÐ7p„½ìtêœÿÍFñéqàż§æüc¿·â.Y]kL/QÐäÚu\òÅ- šcðÕŒ[\*õgÍ…4>I`àUZ€{â*¡í;`$Y%ÝéàFÛêŠÄåê‡%ìˆàÁ3‚Å›SK§žCÙÞEmÙªo¾²çì9Z‚Ç[6$f-Jãg…ï[¡B÷>—§ÑcK•c.óaòê¿`yÍÇwXÌ]Ã×i_]µ#.¼Šä›9*^Ò5æʧ0,Ö0æh÷¸HÆfÔ5¬.2o1dN´>O4bùƒ€$,½™‘€6rÙ‚¾ÈÌý:Å@[ñ)ï¬;WâCœ}2cÊa;˜y‚œÏ™ì=ÆÃÛÐxÙLüòT™¨«úƵ§dÂÁºZŸY+óN/E­$KcK={UQŠbe>Ÿ+Ð)´úãR« Šÿ³»üÜô\ŸÔ 38ƒÓQ{H4u"¥ÜDyZË $y¨èµÁR·p’2]BH¢ÛŒ4ÒÙá‡ÇׇÊ-/!ˆíí_žütyê„?{ki¯f¬é\ôTÑäzŸˆ¥ÑþtË…¸Q §5 E–ñú•e²³Ë™£uõßBœXM²¦«ýOœŽ#ÕmÈ:V‹ÛøàŸ¡–¯F±¥ YÜ’°óÄ%Ì]Ûò‡pÇI=NŸæıljç-}¥Al\/IYÑ»µêÔ=ý+·%ô‚ž8;Ê—JŸŸ$BL&H£¦œŸœU"Tý¬‘I‰,"Ä¡ÏÇÚ+C²js@ò&½û`¢9Ëžô‚Ð~Z­Ì9ÔÖ·ÆV—h%•`Ѭ?»žÎ`¸fAGS7ï„È˜Õ n}%¢œSÊÅpzd×[Š˜`I/Û›Ñ8Önvﱈì2oX1fEtþöuŒÞ¢('¹Oþ)Þð¬¥Ö>¨Ë|0˜l…nUû‹ñžbwä…òÁá³£¨à±»…|¯¿)ÅHAü þ¥kºØßÖŸªÐl\Þ¶Ù>ØÊ»Àyô>ŠôøAOU_­îÿâr›KšËÐbI¢N¥ÎÒ¦ƒÕ´ ®×mÏDMRµ•M=óŠüä*Ã8e2ÎT ýÖ5Û½·ð‚ŸŽ=ü»"t]ÙF¦_ІžäÅoºÍÆ°h] YŽ°.>|òTcØ‘65¨¤i¸ü½kò—?:«¬×ˆŽ¦Füò­àT ¯27³ÎtÞ_÷ðP@/þôº¸‹Ã§„Ò¼q±Á_F‘ªù/æ…²[Ä1 ©Òå2ôÿº¾3­;úÍÓHÏã5Gü¥êd¹bƒ˜†e9º4l/›Æz÷ðØB–ÜÈZg7yª•W‡LjÉ5ºÆ}åVp¤Ù.ͨÍÄPS˜ïËÜDݲ0–LbŽLg ƒ'› 'µ!ímæeγýË{i(š6ËýÅ<«»mq­’º)«ŸMHø`GjÝŽ”)9¨ÐU„>¼AüÞùÐöµî™  Ò %xQ)rÉ €.4 ³¿-Ìi7ÀCF¦æ,RhÍ“®ÅìÔ‡{üObQÏqÙ#ÐÁW™‡ã™Ñœg¦Ó¶š4˜©ŽëÏé.ó¾™ï_‡ñžêɺ’³®ÿx‡Îë-=•w·ˆÇAïÓèSUl%xÉvÛö'dêrÁ^‹-HWKW=>Aï+Ò=ÂòìN$JÀ×5cÉtòŠgËiËJR#wŸõl“õU¤ÄU½™¦å–²³í»×òJ Ÿ—£~AÛb#’¨¥ŽÁgáãNyHyÀ˼üŒèOLùïÄ¿zey‘C"«õêoÁØ[ºg¾¬~B#Viù`&¹t'f¾·ü|€ÖØ=)qÔ²At"B°–Fz/¥ëµ©CðìîœN’Ù’àñÞæµxŸ¹Ya¤£¨M\ÈÅu‘Ê "Ž<°ñ¯á˼äs¤À½•  *¿74U½ÕÇ9Œ2‰‚xZ¿Êú^ ’J“†Õ§Ò{ìº(WÕàh‚Xim²sñÝ ýJ«#„w¢FŸå>]{z¯fs»)tC’Ÿ'æÞ2b¶ðxÍa~ÿÝÈ[Y¶ñ”ZŠÁv¸Ø0eEŒM¶Ù”£]^˜[¯pÛ“³àjŠÀŸør ë(Cíg !ïeµÝlÌ$¡|·›+ç?,NÞpvÇ¿ ¬Ñ» Ê×Ú qŽbYZð§̈'ECçç£Qªa¾F%²ßå% Šxð Êj.M3ÐÉ…ñŸòÈç€Õ—ßÎ!É" 5# EžL»{Ò«ƒ |–ÀrÕ†UâuÃüKô}ÿ}^}Ó<*0¼Sù6Á¯e_X@s G‹R®*NÈ¿ ®Y†¾"½A>‹–Içí”ϺDS ÙžPd¬4MÚ0X–(Ãé¯Io%iÁ˜ºÃ¨•è¸ƒÿ¨ÝÞ;Ú«…2þëãÈDÝý„ŠÈõŠ ·­©a—ƒ8·Ûªhˆjùü.¯Ê´wkG”]= úôÊþ>C WÊÃ0Æ{¢œ©Fzù¯λ¨F©¢¹3SòVŠZy|W¡D¢‰„,iíÙÐ0NP\u“¸×FQÑÎ #Ö¢?ôƒñ'ls;µ2[>©k¥P\ 7£fg”vÚß̳Y‚!·0óÃÕG±ó–¹zl…”Ùd­:ª#˜Š°ÕmAø@÷–s[8µ( æ~Ð0¥r®N"èÄÚêùïV¸»AžH0,Y¾Y ìYh´‘,UV—X#ý‚ž]TS`éÎ’B£ˆ\ä³ A-`„SåîúABK?b)šýê?ʃ:L²ãô|jrÒL:ë…%(z^¿\žDƒ>À³HÑkÀŽ!5ÕBÛÔgÏýBü„PbÑSTÓw²zd,ˆ`®ú§5´Å«AÓÁ´"Ž&"ÍIrr’XÃ_ø7—ÏsËÀÃpÊK•»(þ÷¯«¶öVE#¾¢æ,T­!RñјïôH;ãwŸˆ1E |W X"œ-ƒ)®&=*hAç´ìÓuäÓd3@+4‡V‘ý‰)–'ãÍöº¦¼æõùr#´€Ÿ­ÈžLWxï÷>’šm/!44ø~àüfaí—Ï{ôÙ©E$\›Í)WÑŠ³nMèŒa¨Ô9}ˆªº¼Ú¥'©´!w[f:H»öØÙQ¬uä@ð>PVíÍ©jòq ÓJd ïß»ÁøñG—ëG»õk„QÙËLæ ÞÖ›•ïÂ7ñ»¢²Š\L¾žNèà·Æï۲ɲóTjãï6Ž« ²v­“£˜î–í²~DWwX1U·ú».LæªÒ©/£¸‡Iêµ¼»÷µX ‡ku7)£?AÇeÐs’Ì$ê;@Ð ë-'‚c…Ó@—õþÚðá¾íäìÙUïäž΄gΨñ[YiŸþgÏúÀoiÕ¨ÚQžòivÖäŽ#íÙ·hCM¾ tç¢S&‡B\1DØX’Yü’|Fã®ÔÞp¸;§´…öP¼¤('÷°mƒ47˜ÿ2ƒi¥JŒG…²òÚ||Äæ튡B.‹«bm£;›ÈíLk”NÀ,ãÄñ “IÕV&…Ð?'|øú&Ô×4esGÆyJeÝJ6Î&Åñ;f0§ AÑúù“>BÙmdÎÏ«ëf°ä¢²a¹i=7ðk×kÞ ÿøË“I¿ãî6ÞËYÛU¸XyÕ¨í]æí¯PÑ€áÃ'‹`Ûuáef e4„´?’ß»r=¢ ®úÆt©ú)Ô£RìjM‡Ä`*â! ‘îžÆ(ÿ±]7L>Ù1t%B24Jî Ýdctqm,½;~º¯ÔØ_–Ô7ç,¯±ùHÓJš£ýÝ’aAS£±n‹¨æSA°äY.àIÙ(‚ž”EŸ¯28ëôÂÏy{`»pj*€ÈÑÜræÂãûFk O'PöVZsyX|tùŒ‚òɬ°n3ÅÖ‚ VÉÐÕƒ Ln·fðñR ÔAñ⬆5˜9ë• õ̼™Ù©mBBeëù^fVvv„ŤMÜyj\4dwæP8è­Ý¢T?sJâ 6( !õO.å ÈþkÚ(*y½.pŒÄ‹ÑE/2]*äÂP zNžÁàÇuG°cð¶ Bì„ÒŠ°º8‹~.(`½¾QUÓÉB[ü½Øô¯|³ÛsˆwÞí_Žýµ ¤Îà…ð– @&«†n0¯!ì:Ü‘ŽÈÖÕÈHúFxÒ"ýðDÑFt È^Ä·€—ûµvÙÃwŒ.„iiÈR2®Ûo¼¿rw¡ýühy÷2¼e҉›ºH\M²ŽáÚò×bG"g]-ƒFˆ¢SE9XžÔét’„+iI[¹5 •á0_ØmÑø~ÃAÀ†NõôÉÅgÕZ»hÀîQª~óÙ´¬7÷éc‹ŠCXOÃÌ&{«4Õçg¥xMœRñJF)‘ôQ$Æ…—^¦ŸEs“0j!ÌÒJØ7¥à¶¹v¾¢*M‚G0ØÑ;­,¶¡N.kfwžû}Åpƒ‹>^f…öjÌ¢Q¨8,Ð%ëSgZ­Fèìý8ð"‡>æ<¸~Ô«0ý:$èÚ¯-?’ÎE°Czq¸4Ðàiæ6d7¸`Áy}sü·‡?A7b$ItäùèÒüÓ{Ñ %ÕY~>+]¿å“Á®åK¤{5%â¹uÂþC†™íª¸Ôñ-(à߃›cÜ02Iç½WHLÕi÷Ÿ;£Ž“2\>¼n½ij¸LB–$ªÕt›ÕKPa• ´F23²/ y=3¹¯ Ä­‡m[>k”ByëFw AVlFðò‡¨o\þî–h/Ët”%w [ÜÄÏY0À|—ç>zÍ7·ÿzdz]ïDKˆ{¨]cÛ(¼DRüŠþÏô21è…Â`çŽÀŽ k%  WIP€Ìµ]c ­T' ¤0“ŠC¹…òrwe9ʳ»Aç‘»-™žwŸŸ5¿6Xˆ)Þ@ºOXÝ¥÷” R:ךºmºÙ™U&MëÃq$–¥¶:Y­<'V5cá´[¸m®v1acã íQJw8‚S/ëø’¤Õ¿KHP¯~qpŸq+…‹ mçBf|En}qþÿ‹(æ,«-N¶8•ëT1Ž ~½"Ûd\ÝYTf ]g¶Ur½N‚\ò¹Y>à =è<Ï<éaÍCUÁ8Ú`ž²®!v>(íW)ëÅn v&¥6ÔDVãL~MãqÏx!QGÉí«s¾Šôžb–þ§Øèöue ³7“–`|¯C¬X Þ @ã´R~#Kñг½-vkÛÎÑ£–å×aý+Žºœ`©>“€ö*Õ›#ƒô„ï„oFÚÖš G”ðC2~³qŒbu¾<9½‹OõÆî€7é#-Æý˜Ê%EUacñºwFlÜäãØõ•.õ8μ|y°prw±„T];j½F †‰L“pu®ì}Tð7÷”pòÍ^ïÓOe¨”6öZ¡¸›8'Ž(éúèE2~ÿÖìì³pPÅ``ú<…:øú"6ãÁ0`o¦ê­©ýAî  Ø‚;»7Á¸£0Xî8»¨S-¹‹õÉ€|>ˆ‰¡à¹&Jªä$SIüñ:MËX>ç3™CW_]*E™ß<Þ*Gÿó;Ì(‘ ™À«yˆ‹QO‰{Š®“ …)`T¿~|¬åMc`Õ`3"Mѵ$ ^rÁøuÏÿŒpçm,Xf*絬çR «5¬3o÷=-¯…Ö©x×KõXã”Gïx¬~$)2×júÿS+0÷ê”–a0öõNÂíÓ ä©ò‘HBNw'9m\±êJ Â{ÕÇ?×…Š<Ø, ]À)4A<Ðq [| »—†ÊD`šJõÔõ‰ö™Ü?y-që­ÕuÉd2ß:Ó[€¨ž4A=ÀˆŠÓ`7N óa0‡±:Y:»¤xdv3| OuòJÿB“cNdõó± JÄÎ× ¤’Í¥õi?kKÝí©‚ãô)Ë2i75Âùkø¹H‹Nl9»€Gº 3³O0Ãv¢¢ƒMÇ|þ¢¯âwH˜k%wŒ ØÓük¨ù¾¦‡—öªDO#_b»ìÛw<ëQŒy=-13µÑ·OºÝ .³Žm9¨uo¬­JÇê¬Å'ñ.%q«×—a± 1®7Æ ÐšQĪ[.ä~cïöçl"F<žGGg!G‘VÖcC¤7¾uOeîÛ!¯z{PFU…`Ùš¡!#9¯gt`iÁEíèZ %AŠKfQG¤_G?ϳËk¶YëL—½+Xü®Î®EpO¨@â&àÇçøÛè\Z?™ƒ€“‘ìF4›oIútôÅrPÞ"sÃ^ÖA&qöuŠZW¨4´3cÿðœÐ[¦ëLˆ5„¥cÿ¬ )ÃéÍå4”Þü?x›ó«B;ÓÕ á-``-|*Ä+îEÄû³þ~ \“\+›¢ìWa÷T[H#5bÔø”Ï”Yú‚Lš¼›4=ù›ð1Èê‡Dr|ì€2/„@ZXÒÍ/œ¬K€øø ÐÃ\W`–! —hàÔWñ=¡!U"GÌ8š²¿z£nÛ;anpä‹Z©íB¢ÃÑ¥š½'uÏéCŽÑ2ÔÀÇó ˆõ^·U­XDßwýÛYÒ„¿œCŸÝà/0Y8É·„AÞ#ÀÜŽ¶N0T’"‚ZücÑq1nŽžÆ$ÚÞô¬–•Ã‹k]ôܯý!¼³Àè„éç¯%hg¼W'ÃD ŠsìŠ~Tv4£öd·•Ï¶H·î^”Ýk­–RÄU„â<Ÿü ð’¨á®x‚“QÿMù·™¥ÆüóÎ@q0ýd$Eg@“/[¡0º ©{½Áµ=ª•Hs›2" Áœ²©Ki…WÓFð·Z y‡ÆÃfVÿ«ËD»5¥ýÝAQRviÊs±-ÀËç}ÿ-·É渷ŒÛ©Ñ†Ó˜ÞÛ8s껢@Š£ÒãÐëÙ˜õI:DÏì1j›xê –,®6ê÷­Éf=šŽ÷9“u<.g[ìÂìóÔ=Pˆ^«È¨6ŒPÇ)¼9o·îøiÙd”‘¢Ÿßs|Ejµ•ÕŠ{°!¶)Ä2ð[ðš^›àVÒMº?ß’c¨Ö±ë!2mMeUô{Âo9»BP¬×ù…D2ž‰}w¿F\¥mGM0(\IaËR«š…u©I!ú– N©Ä ²\»øJ¬ .¬€¾"=cVå{¡4 Út¤ ‰0/ü¯­u•ê¹H ÎÏõ¶äåf‡;®ŸMaC'| ²¤œ:Ä‘sÖÍ“åÖr3!æŸS~Aà–ð‡h {hp¦ò/Fìà $ t³âH òC§C_ÿx½ççbŽX,‘Y°î¡8CmB8²ò2ç­¯ã<çAΟ‡ÞÌãÒ‚ô’”Ö<îT×å8ùEÑ°*t…„ÿk¼‘Oá¢Ð9¡Ÿ7. ed‰P«)k—}?WÙ„ž®ñ¯…ž‡ ÑQ½9ZaÚNƒ¼`áÉw9Œ #/çîlF¼qûAãÈn­ø÷¶®þmE‰UÿûËeËÆ*»™¡ñ\䊻ǥ`£³Ë½&È®P=à|Ù¨F“ô¢I+~±{á))a¼$@µž¼zI„Ã×lЦzôÙÁ´²xQ‰5&º…É)î0BˆnAÞUAK»•&Z¦iB¡•Kt0& «=êT5Ýóš½?Q˜¶VE-À@9ÀU†èOoEî%Å•#Ì‘E¢ ÔŠôM€C+7ʶ˜-]<»Ý®R?}‹5¸lî7µ} ð‘¢ÂÔG|ÚÿŒD>Ò,·EjË%ÆR1€Pg*4Æ‚…1ÄGøØ7ú=8õ‡_aõôi¾wu–i± Súzjë3)nȲ¢ë—uýXâ«=!3;ä½Øâœ%ÈB¨ÿ]JÖ£c.Uw¶ºØ÷¥}³²²wÛëQP ~øÊ×!r’”\ê"ëÄ‚Éw¼ÇË,E(]íRL•J5•ßáÇÛUÝ 1„‡§Ée'+¶wŒaÃóÁ"?ã”Ãm ½àÞ¤ÞZ‰ª€ Ug›}ïgÝ$€]³lt1Cì³?8^$P|ôº³ôýî™÷¿Ä)Y×C;Êû¦íeö&³)Ŧ⠱Ʊªá­Dßw¬‰hÖ®pu?þ}Þ\Üzm¼IÏúuI8MrŠ2€RyU/ >é U¹s²Ä‰ä.?~jôzœê¾`\r­9tƒIˆ« Fo­û¯Y[Ñè#öåÈx7BëD˜µ0ô8Ñ ²? Üó­Ùü|Óæ÷‰ÅMêæÁ·Ùòa²wgÂÝM#ÂxþOHXÞD>ày¡å:À'Á™Æm¡—€‹ÜÝý…ò-/Y5"õÈ–™œ]u{³T)ˆœ´%æ,VPùÛ½LAUXŽœÐvúÂmfÂróçd,4/wa‚šÚÍlMÉ:o˜ÚZp¾‹úgdC£áгÝÒ¾˜´Jc$Ì×ÃÔAÝÐ…`öÿ¨Á÷kì¤~2ëÃçJÀ Cþ2c@M2Uä ‰?qØÑsŠÓ~ˆ`Tý`£/ßdõëZˆFù$—[ …räêï©ÇÃ)þäBú>°…=¥aÛFyŽ£¬æ'6aôu—’rE‘­—âvD oÖKJíÂ<3ªð=Àï­#Uwëâ­¶ŸOËéùÚ준=AJ«Q‹ ÎC¤êÚ™ª¸>I9ø9þ0£3¤½¹¤÷g_\#¼#LNxçÝ8W,H9s²ÐfÇ,‘¯[œŒ=ö9ËÃÿ´™ºg¢ç=×-»½šá¤ÑÆÛÛ4]¶-scÍjiÌæŽåŸ­+µÓ÷nD`&£§~±›~€ŠdÖü"P2|ò¾VÐ6]ô}9™>Kãp>¦1ï½¹…%ÛW_¦ïwSøÖ 0t<74`Ûà3Ï?îþˆoYhÅ? ŸŠ¥ ðÁœâ{T¤&šmUÑWgÇ…˜’;¢Q§Ô)œTüA#^dèÝÐL­¸ô+ÀxŠÄ¬8£%#¨±OÎ ÊQm¾o!Q•;Õè8•©2Ä&N¬IKUŒçól0¿3ªýú‚,æbÑŸÚÕ_)fŸÉt#N\öìç„À_|T eo½¬;éYΗO’ñAìÀÒâFž7:L…BeoWS½jAÊíK\!ÚsqϽ+Eà3~ŽÅe#b‹ ¦¸JÕ·3a^ÒOR½ž`!ÉOï÷½½öî‡öjàsQ©ÜqEœçz(ÏÙ©býí¿õægÕû«åº­K)JŠ¸qXeÖ‘>kÒ…ÊÙ×u4µ.3bÅãkÔ'QŒ•¸>{W%óîÝT'¡%ˆ³¾’¼qf™8y XrmL¦¬ÿ@'î&Í»S¼O˜ثm…)¸¥\~b‡ØùþêÛò­ð ;†á/Õ‹ œ@7ÜdM1¸Èf>ƒø.Ÿ²BõTÆû‘8‰xÍ+zRÆacÕ(¬½D'oÓ ººÇÆ=²\\*ËÞ¥K¤Xvðüv¥M!þ¶·!i _ÎþzRûÚhÝU!T)3Ãr7-׫äǬ®º<ñª1öf,愈;;8Â\ì<§)ð?V”¥.–ö ŽéŸÃ¡¨úþÎówö0(±Ÿ 2¤¶8\Ø{zõ>žË¢—ƒ2صƒf>}טش%:;V„u0ØRëòƒ‚evÖO§×îÓÀ0ö;oècn”@伋ð·Jô2ÛgÓçB ?)ùénW×2×dëFTkz{—5Û² €djÁÏèkÙí`ð»TSðžäÔî³îihd&ÕaœOñ€ÅꪺüzCBŽ2ä^º,W™»nR÷hºÕB§¾ÂŽÑ.·‘½ÊÛ0•(‰úòízáw÷€ö?> `šj.ʽf°!IÄA[H¿yRÔ’„kêðhžW¯°pG€bc.dËЭٺåŠpâÜöÝåœUl{9 ;½í€Pñ¶~i å½v5AË]N¦ÍÆZÅýu}²»8Æ ©ÒZI£ ’a:3ml+¦³á¹V‹&7"¨ÊÊn!–Ã8_ÙŒN™9}S2 ™GI…[l =–1M^Æ«èD/-œf|Q‹c÷ÔÌÏKRg{¹•­?åÖU"^ ”L›ˆæÔxÈ·46Ý àbó§ŠYNˆƒ”¥K×årP¸jyÎÉ©ÆGúGº˜C¸„ê_æW/FÞô¿î¿&ÿœ’„¤øKi®`ÁˆŽü‹^וøªÆ¤ ·„€'Ýb+Y«¹_p·Dè¬ÓÔÈ«¼qê3ž¨óÌ¿yâ˜R1Ó°×å‚ÇŸH+ÆÚypÂ,ÜÒ÷°Å‘EAí÷#—|)êA²´+YIž,@äø\P¢¼ÿ…hž„Ù–‚_<äÜä×t8hÎú,~t׬=„Wå&šÓÚo7~RŒÓ¶°-´ÈfƒÕ|6€eÉ4#Pj|ÂSÖKM”æh.@­ê ¦‡W=j°–ý—KÅ Ì@ÀÏy`µÊuYIŸEœcF>‹ùîœ'YߟµäL^@Åç ù0¾Ò<® ‚2—´´0AD.l M÷ë!+´õãu†cefPv¸DÞ6\¡Ê*½ž3¦Ñ+Z?ÑÆG6ì:ÆO£ÝÄŒ)Ëd5t'%—>…7gruÇôÀ!‘B?Á­àÝ~úò+ï¸[­z½ OKô:²sÅà`‰wá6‚›â¢·&nú<¸IH¯ÊòV>^õ9âIÖŒC"Öñg¼á8/˜ôëvvpB »QL#éWqì먉¥FAqYƒÏ]2²_÷ £.Cñ C‰â¯ ¼?>Àu0¢kùTDæê“å VT Sû‹ªÖ-(~BA¨¯é§kͽ‰PMS.Û夲kÜÙâüYŸ:†(x–¼¸S¾=:WÍôU`!`ÿ÷Ÿá`&ôN?oµp‹m™¡ë #mEǘõEŒŠ9Cy®,/ D¶•®P)ÕM»ˆ´# /6fëMOèÎx9[YE«è,;¨?ÄjÑï ÛCÊD'²|%ÁvM嬙öÆIbOý”‚§kM†ˆ§v!é î‹e«žð¾X«­¢®uÈ´HÔ³¦”ÏñN‘iû?¬Gr$=*ÓÊ×b¹‚sò›*qU9Wè ZQŒ‹™^S5¨c;RZIŸdRãL,1œøN!Ó§±¯Â­6½Dºï_Æ¥¿{º¿óÝxŠôÛ’]%IÐxB‘§T4„ÆAªÔwl¶ÉtÝÜs‰›[N »ñlðê ºš6[à,®fä¶d&e`Ä“°Röñ1»ôL`Á¯ôÎ}~»³âH^¿{Èúæk{Œ˜¶†*UG¢˜w!}ƒæÉÂo#ñ mh|9;Ö¢¿m¿ècWýœøá}XŽ¼÷ÁÐme*ŸLe,nA|BíѸ£:s?.U¯ôÎÑŒJŽr§"Bd¨­Á/‡Ìü¾Ï›¬6tÆ›ÔHr@l‡0¦Wû¡Ÿ ïEéž¡ˆê‰!Kí׸×ètéS>挬»p5á’fó¨’‘1:UW'[GÐ*”™z ƒ—ŽÃI µJGd©×ǺÆv Œg=W´Ÿ*E/ü^Gš„×äJÎz™ëˆ“¹ž¥yÍ@žÁ;Cú·¢P{îõ‹K|Öó­’éØÅ€Gµ)»4¦Gv°£ci¿~Cô@´  –Ÿ(ª×yÒ‚73+BaÝ-ÖR…²®J“`[û¦\‹7HÚ‘?w¬cø†w,Ã?-]˽ƒoåÓV¶{êtÇ™– c»Â^™ˆš+µZ8PCTNttÌPì~µ+ÏE€ Ö:Gññ÷=R]p= ©1`3œËô«Z§£ã„Þ7«­™ Ç,\¸(rÇ9ç¤3.cB¡:X[³·ë¢Á¯Ô`Ý„ÿÈ':¹Rë1k3Ë\Ž¶ÙûòlfØ(€‡ù%p\YT0ñΟ0º «HÒ`N®gB¯>öÆ!«P˃cµÝ¹®îO&èã©Çƺ HkÕM±>Þ® `ë¢|LÇÉO‚„x5aˆ-R•DÇ'qb•r{œ"ºÁÕ£mÜ£ÖÞ¿RB‚ÝÁy8Aœaéyl»˜¼î¤'ºèô']x€Ê¿Çs•×Þ§Yâ¿9›fSl¬Ì¾.¦¦ŽÎÉÕÂ(Ù~ûý³<ÄF¥Wý-EXú€|®áÀ?6ûvû+*ŸBCµ€•æœ*5ñ´F”­OL:l* ²Ë3ƒžî3K˜ŒcÇ©PP†þiz•·§Y—Ñ°ÆøkÀ?3>A ÷é’çüp¯·@B×·˜à:¸vª4J¾°Ávÿ–HmÁŠIÎf75 f\|·~\‚Š¦¯ aÍ}èŠ]`ÍQ¨` ~×”6UÂ^³^¦¯ì3ëÃok½’Ñ¥²N´KoKÙ (GeP$ >ÑHmbŒ?é> äiN8t?kÀiÎAx§%hÔ4R®þe]Í@f-hÇ!¬ž/X,<¦¹’ÄiN°Œ™ëŸÀTEöècF€¬ðP7ML«Þ,tÿ:Ì»¯·Êq(1´+€¨«µc½gœ!™Jrðÿ®çª!› ²Ì˜ë7ž{Ù²q'ÎÕ,uµè`âãËÿ¥áicHfö…C|®®Ý ôçaDCÉßEÃIƇ JGÄÚß›­`¤Ó¶yŠn8ÙåC–Õ¤ӈ“\T8e)-zÞZ”\º”æIàvUº«Cvúî:kd3,à =ñU$v ™3ˆy<²Â*? yÕ%O[Â;¼S ‹¬f3 ×CWMo½ÏõÐu3ISÂ]eÏáÛúOá´ÓÝ™(¡n#9o7 PKoÖ<9ÿšK^ó½óYzjXÌÀ“Ü$¬‹y#éЗ¯®jR;,ÙàAš‹k¥ŠTsÔ³H"ù–g\^úÆz2¸rÖmöŽu|Ú¿ÔŽr „Íè<v­ógÖñ8î]»Ã47¶D"é¿­¶ñÉ’S‰ÏqêÅðºžî]œLÕaÈô¡Æ  \Øÿ x+¢Ä× zÂK2 Ö:Â~¬$DT¥œ¨,|l\®à“0Ä7vÌÜ©aã¹`ÜÕ·)©T&ìV;I¶lÆwBáæÅâjèqC?¼æÿ[˜&gŠˆŽæë¹”Û'WŦµuCn_‡_–k EÖÓ\ɈxHí×o{š!Ée…Ü°³âç1O{ãp$ß7`êñ­u¡‘Lª˜÷0;‚ uw6€l„–€­`)Y÷cV›ÔÍ{1™ âââÆ1ðxóqD'á#H¨ëzæl_À E"½¤g3mЄ˜Éôs' æ÷kÍîôc$_„µ0~VN ?Õ£8oÞ¶r^}Š­Ýž¼MÉ›Áo ×s½v†+”Û#Å…ùµ¶xxzOè[|lé~Aó.jŠ“YÜÝÅ ñ ² *akÌq`i1±gnF?O~ýø héç©dŸØ¨ 6—Ñ|òo,¤*éºÄr`ZßÄ2OêÜ ¨ŒQ“Ô=ëv6xâ@UÍ$ ìkûz¬[Ђ4‹lY¶;ĨÚJ5ñg~ `!œ˜è_Ñ[œ€ŒŽ›A“ÉhûÄðkZE;*ßwC¥šñœ7ˆXùäœÁþ`‘ÆŠÛ¶*/äÊ÷þóÕ7‰Øjnx¡õ¿Bqcÿ*´äOñO!Ú†<[VüH>èÃ>–Q&%góV£,$Éù3ZŽÉÅ»2ˆ5Æ×DäÿYîZ¼âK!³È®™¬M 3ä;ã Å®]ùŒ,°úÑ ²‘ztø#õ°OŒî½Ÿ‰&b÷‚¶Ôc"1ô›Ž¦B`S4çº`Üû¨-ß[›==YA.ì|™&WRÈ©æÅ!V ³CµþÎÍB]œŸ¶2 ‘g—“>ŽÔzUáiydOO…J%o©ä5ÂuO÷óÏ{ûþ´²Km?} SÍgìÏ©Õ»§%’=Ì°s٫̇¾zYér:“ø@©w·ûY³/¼¶U‰¬Ü¼Ð›¶EìÀÖè:täûm â‡.ƒ®)æWA¦²ƒªŸVO§Û§ß½w¶9®sµ7h[âzù‘óH¤o'A°új!ž7ç6-º}É©W+2¨ÀÎæó®}I^€^y –Ï‹ `¡O·_ÅÑp¥;å(9á¯B%aïG#{«‚Q.w:ôù8ÀX;M&ɬ´ î®"Ég9‡šf}½±ÔÉ’pÆ ÔÓ\3ØsáE˜JP¤Ÿ©HÉ;™È´Ú´lƒAóO@qÚÂ=²¿é´æÍË%Ð&é:ZÀ~x2B#EŒTnÉcæëM}HÄ4Q úô/Òo=ô.€±Õ¦¨Aa(Ãî~iÔ@QÉ0áÊǵjöK€ÄÀä5/æ{âÚn®ðò—’ç<¹)Øypnãd} „“ɦw÷ȇŠéúô8¤ÕŸ­E3óŽvHvèÈ~Cò‰"x¯(žW»E·!„Kñ§ùá|ØðÛNË4­‘ÉIa[Çx~÷ Œ#¸OˈÛ5Ãq#ÅÒ$f€]9/Ã[v.ùJß¡±dÿ“«lõj¼Ï¨ äZ™x½òÖ@º¿e¬âqcõâhL¢äžŽ)Ñø¨ý}dHU;*Ô’–ù¸™›ØŸ·d–é(-#ÇTßÀÁ§Ú}9•ÛØîo•Ÿ 7m{dlaR+ÑPÕ £îÔÝY|¦hƒè`ÙÎü-{ þ¾/!=°A½±Ké¨8$«¿j:DŠ| ·Q~$‚ fPÕ‘…«kyZ‡q¿˜¼K¡Qó¾ßCîÉL¨k„ ŒXäªëR·mÅQ„_13¶¤<|) ³h­^DŸ§“÷!5ꦖoáU€ ™hØ‚:Öë°ÏçÒ`:¶®÷íøu‰|œwÄI–¬RÒíASF§°VÆÐoŠ³ •7ùTnÁšÜIëMºnš†Hº`¼­}„Â'ÍX•Ú>Ž™XxKr9:¯jƱ‹ÚU|FÒÝ_UÒRoûVºõU#é¹Ú0Ûq…ÌѤu¼7t·@£§{Ƕ¡e­m.Öë—_[÷)ŽìÛ;<jäÉ„ýï?gÛË­—Þ±7Ãô |Ö&ª=Q’ Má¡€fá’(ŒŸ¼µ%¥²Tî°É0-pj&2…ïÒxZªê´¬nÕ&æ?9óÌ+µÁþîY×™ìj—ÕdݨšÂ÷7MÓvÿw ƒ6ÞƒìlÔºwÕ)ˆèÒ­¥S4”3>èÏ#´ŒÞ•±lãŠs¡ŸëÙ„é==5–jX㯿’%‚a–˜a˜ÉÏÓË"ói=D[×؃qDÐC,óÐ'ÊÛ£û)\ŽÄ¡‡®ìÇ.XŽoŒ0ZVæ)~,*{q!o€áÎ3¦Ø@K×'#PŒ³LÊ"u“º• ìS)B"É7 q¶£ÞOyò¬žŽb™v‹7\ï#Þ™–OfŒñÆ ‚@jšÛ‰nÀØGýÏþx—_Á:nåYø¢šÓâ^ø-ðGôv›0&Iú³ç©Ý©¸±e™Vr#ÛC¤šÚÁ7ëIû`W…£8Ó _¿À/¯4Ì/’)wo²úÝÐ}ŸDG9sx7¾ÅÉ/¡UˆæØÝý/±SÛ˜¾²ºt˜Y«Bªñ鲿÷\yÑ«³È‘ÅK–SçAo©nÙ „› ;B­·¢PwxÑÓ¦ö ÃqɘLsú’(<Π³„b†Ê;¸áö!ï]RöþBV6ñâ¬SÝÌŠN⧹ÊqÒR•­<‹ÂjØ?œ‘{‰‚$FY&ö¢Ÿªx_±ðàóë’c!g«nÏ`ÀËDtd*Ò9A8Y%â;E _6ùõL]¢Mñ-#·ÿ Ø)Ë„‚³6sÐøR›íé7LIpô`Á%SIóóâGtºógA"¨||¾}Žu*ç“Î[*w,yøfœ>ëºówXD÷vãUnçK^pÕä>çGd¯‡&[L=®tÏ/:Áã Ɖ›‰©,7áÔºÒÉêýqÕEOU7„¼)33™–•nò·ÐòŸ@Xå'þÑ:=ºÙ¯Ìê{ÇÑiݱöú&ò–§Î<ì„|ªáäP®à……%‹üæé %Òð€ Ó#Di41]5W»1Øu¢ÔÎ# V÷+fzôÜ‹ãa¿j›ˆ™q•gL#Êù¢î;.=ïpÔªNAo_X¤zdiϯ^{ÖìLj‚Ãr¹ùF<µ‹G2O‡9óx˜Ì9,³2¢Õ(gâ·Keš‡˜šâ§ä~’ãR¨E7’VJ]Ãý™øìSH1Zv¸ÜN‰±ðDá—ú4ôÌööô1Ó <šñ&²Ž{o¼ûˆ‰^ÒZÊ—ö_-ú1Ä® Ó?·‚it¿:vA±K#¡Ò´E]7ëR²Ç}~ŸÉ¤§o¥ÜÏÊ<ü¤}º¤2·/†[h~ñà:®¼%»¦ióÅ.9ì·AJ^Þ óäœK~€ƒñüŒ „¨¿!wa"µ”,Dj£5=r–ÄÍ}ñ×j:á/Ò“î`µÌ sd­ žÎÁRE•,®‘-\¬s0ÓïÔKôE=,‡¿uDWµ1ýãÀàÏ‚ ïÄÑ5Æ Ç¥z h—-£ bŽ-ÌXN_ôYö1ºñe‚^'ŸWz ¤L GÅtMèÌôõ¦Ýü¦VÁq¹üByÆú'|UrW`yT’ÀóçРž©[Ë#CbFèiÌšsè#aÎeŒpÿ[ÖNWDæ®ýqÈìüJQõð–LÂZF¯èðlCjN˜Ëõƒ–×ÉOÁíÖ#¢¿TTxÖS&lÁ•ÒÄd¨í0[ŸØqþì›B‚%EXV"À–âT1¸võ¶¡Eb§Ý6Y PŠ«;òθ'û†5£SƒÄˆa²a2¾ªÓªJŸÞ\!cø¼Ø{wßE³•àHŽƒPúß÷f;v72â×m¶r«úOWmoÉÐ9>ðw›mÆZëÿÝ´N`å‰Ge×Ð@3ýçmãÌFYhYsÄB‹è^BâcjíŒf§fù7û/e‡z°ç×EC8gÛjS>^h}/Ý/˜?Hº—Ú¼ñD£4_›É†—{dˆ[„³V>€µ“ÅU{D,5ñy»éÍc5ó1þ"Û^Ë„]>×Àš¢Ù¢è×áì!!zòKóæôWCL]ÎaË)½Ž[ÊÒv…+ é2÷g!ÒŸT0øÍñ¥_^ÁÔæÔâ®Ñ&&ÉçýlôòŒéÑIô-Q2q¥¯ëb}û¤¤WD߆/H‘BÍ‚œ;¤ÔÛ±¨r¦_•±É9”ëpÈçªWUZÕø¨ì赕ËI †/w¸ÉRÖµâëWYã ‚¼ãZnZ9‡Á˜Ë«Š&%TŸ8„Þ¢Œ†8²åÇ\”é*aÙƒ¹é’êù™òdSu<^%­Ä‘hds¦~H¦HÂmm ›€Ò•kq²O@MZüצ„¬Z«Ï–¯®òt)Çï2ê/œj±«x¹|«ùïs¤6nyö‡Ë{¯:W7›/[]Ó÷ƒjHóÏÖËÜD´¿ œt ŒdÛo˜KÕ­ð*|h 2G¤†àå`œÑHï/Àé+ˆ>d‚§Á~ÍØsÞOµ)+%µØЈï=Hëž™®6E%7ø¨dw~5¾Ozª+Ijò¦ºÙ¢dtáðæ5ðÙdÊlüäõå~ÎaíKiëÏ&4­É,É‚[)ÑØl.rÊ·1Cµ9¸¥çãôÔt5÷JZæP&Kùêcj̲ÓgC‰ÙLzå–LzÀ'\ú@Ÿ†'ÛjÐ!ÞðÕëXÖ=uËÝ0œ®f%ºöÀ¤ú Ï)m§ÚÐ>N@¬Â+%»ß¥Ô¿ÿü€„ª¹¦wr?Ù9=uƒ•°­Õa:FŽD¡ouŸ¶“šYˆb6“ ®`aç|P¨q°“ÖAI¢Á4Ê Äá :úÖ¹Á«þœ?Ûb.*>‘º~˜Ý€™LM D³Ââm:=´Õå÷ܬû¥oM¬ì]ßÞ<µ'¯Ã;á Û ÂGóT¦¿Á½Œóʲe={‹ƒ|“ä¥åª*Zà|.‡~ ÊõÿµéºÒ) Ü ,¹ª¶4ªç3Í^|@÷ik‹ø&”™>"QÚø?Há,ÏFÌÊZn¼{jyÜYÛº-˜µ±<- +û î•úóhx1FÓ0jö“©´{Á;€!8&¤†—%¿oøG—®}iÜÎ.f" Ѳ€HÝâ홺í=‡b[ïUœï“…ßÃç5{GäLñÔh'¼í¯t”õ*…û ¿CÙØìIÜo£f ü‚™›ÜüÄï;Ò3l§—jÈ!|³Ìæ 5dÒÀ²_ ‰öÇÙSô]BXºŠÖOcVã_%«+ u ‘”&›Ö½2na^®Ùä÷záÛVâÃý_úÌ4É-C#“]b(IfpÁÏÞ»ä'ðœ6ÔG# wÍ|SÁÕþ¤üף⻦Í,¦œÎÕФæñ‘voøý˜jæŹ—6¸ð=VÂæÓ,‹àÁ_¡e=ŽÁ ¹k™¿;+»m.¦KgÀ‘`uXP¼mªô³ÿ—ãz!-¾×ú½¾Ž1ôÛD ³¶$rïTjöW6 èIöŸÿ0ÁQ[L]©”%qMµ´tå)ïç¢l`)¶éb5ÜM3š/—.~°ÐÜHûäý®sàn”n?n…63EUr¹½ßnm’i’fâh~%‡, Yr—Ñ£u¤‡õUO”=Ñ Ü‚dE•b°ku“*ò’ÿʨÊì»&1FÚ £æ^ý%é8v¿Z¯=ŽL¼Ö´ƒj´·£c¦]ìÜb@R¬ì±º6®îÅ _Ø É\^âl#.*üXÆï›lŠœš£l¶Â:±gúOóe´ër+»™š_þt4¾7&O*ùø𧓠üá€ÿФÜhTÈ›Í#/º_'JúªÒUÙy<÷MO´",Y)¿xýå.Nn¦/« ‚¥¸Àxl†ùi¡¢+ˆñÕâ¢UäkeKa-Uû+ú¦f?ئŸª*S@,½ü-.bmXÿ[ó¨¯w¹Eg¡&[njÂ¥ë!k’IH†Þ¡œºEÉ?&ä@’›ÁážS”4zKÕ™öÊ0¸”Ž$»ˆL_,7ϧµi©0ðÈAî«UÙ€ÇTHÛÑs Ì“WØ«„?÷¯êA­¬ d’¿à½O(°4qg·¢ á;2“ºïÑ Á¥û Üë‚ 4Bÿ¤g:6í½¦§žQ³Q0vìhÝ„M¶i€n‹˜Ý›@ã7‰½ÑaK>m2rždOÿ\\ÜÏnfêh”«wœ^¯C“ÎC(ÉÏ¥_拉ˆ`U†Râ»ÌüÍ7Q3çE`7·—yü;²Z¹äÌ„áÛ‚eY8¶?õÁDï„,&¦îí&Jgª?ygß - Vî¾µç§j g†µ>ã“È1œ(«Ä¿-=÷ˆ†¦pßL¬ä}fÎFFëÞ&À#¶y:Œ lKŽ£ó®À´`€±Á:m€Á¥¡ jùÑxŽHjß0D²$,u!L\žÅuG>Ý|¬lì£6 u¢EýËŒjq~Õ!òÙä³ß Ò4ÇoËÅ:³m`€£óI ßÇ6?U¡ùm˜ia>{­ñÿß÷† XY,{éÐ΀eÓÑ¿.,mÛ‡%s@/fÿC$[‚qûu »;Ï .©½AM =”¶t‹…×üÒ]¯³p±Š|iß}’€HSwÀK](¨aˆÏ6•I(¬ïc÷œÙ¤dýkZô²ð 4ù È[þ»úoM¾Y+ Å 3)•À×kì48¬¬›ï{.vÍ^NÐîîC¾I„{âÏZ3©Ä±Ž ×;€<æûÚ ‘®»wq£8rœp_Öþ½Xøä¿c0 6)l§Ó¤?‡]f78/9ý{”L‚;qY+}>Ú‰à.”"Ym€y°¨Žö‹q-]v¼Ÿ±Ý’ôL‘,ZÔ .ºæK(3\#å¾°¼R%òýµNÌ ~ª9æ›7‚é. ¥ ßÄŒa¹Ÿ|˜· ÁyÜÖVNbXѽ3RèâŒÃùR®ÀFïJ™–'樊¡ÈfÓÖdý‘µW‚¯w&0O@–JºViœ˜‡gˆnJS¨SÐÑË·LÚÛe,P‘’Ÿ*ôCÇÉtäíÙëª:õ%¤;:Cd©^6¸¸™¶¥"¾›rÏ·ðgñE'l±¬;CÉû%¤TˆSŠ9$'í{t‡¨«à°£we]¡¯ÎB(€ÈÙék ub)Éà·ã‚cŠdxP®!9"ÉàÅ"¹¥·ZzFÛÙ¿ñ±ªï¸8m™…:®1ï~Ú˜?î.x̪£pœúÓþÉ4Cv¥Åô ôÈÙ;¡Í—B¡K°eny`J—ìÙþÇ) ²F­ÁP¦X¼áu&ð'ò‹¿÷E1þS¬Ì£ëË^ë«nÙÛÌaÄìI¶5i•yžÜ•u)Eª)(ò¦ð«$M·ù2Jýžáfp[RÔ¥à,‚ÂÌuGœ6eT‚yµ’ÿRxDí籸ÏØ#=„SNm~÷Ì/cór»| ¥ÏÂp –ÿ`½uu¤Æ\\sp3:;û^ ¹É…£Äí‹úçîÆÄ?%tÕ½ M‘’³ýˆÓ–³¼ËD][¢!‡.¶!¿qÎ1KàIŽªÒƒRmõ‘ÕC†ÞZ~eR-ñ<gkþçŸ*/(þYÛ.ß!5’†Ø|Ý;oÜso-O²ã |[¶úè§m¾é,÷4b^Û  J1ÖøÈxÝ…¼÷¥n$Œ2ÅÇEÂÅà"G'½x9ˆçÇ"šª½ñÁ®eú1Ú]Ãú6¾¬Ýfˆ§ ’vÝ^“OiÜÄ\ÚdUîË¡§k¾¼ñbÂØ|’ãD³pÒvé;t ý¤AÞ8’(“S,Ð&¯ÜŽ£+ø™?ó:øÂ͇܃|;æÞWGIHl":Y½†Azá¬'|á ,°ì9(ø’¹Øà–•÷5XvíoŽ6”ÌŽ=†ÍIù.Íq‹º:&’‘Û’$€S°íi`L,‘£ÃŒç˜»~â›Ð i¯Bà'ßzÓ,fë] $Úº¦WÝØÏ_™ÎÖH¾› †þ íXf‡4ŽÙœš¢§û$n¬ÀÀWô²Z‘`µ€Cn‡-vî;y ŸvIçû0vékÉÕõÀyW´+Pè“’3K [ÏÔÒP¤¬Êí(]X”Xÿ'!¤–ÀÖ¤šý-*l¥*ú…"Â9ahœ?ÝjI—ôHíK@]ÚæÛñ|W,À«i u7˜Š–ˆO{|ŒL)9íTb®í¢Ã~Gj5èOðŸÇdMx•Qk ñÏÙè[Ι#øu^)Jß-*vPy,5ã_¨nÀ2Õ­zü TÎx¥˜H!l¡–ÏÈ8ž—àOr¾ˆ³™KÏNÔ×Êû£ÑbQB²5×B‰áú–CPð4B#þŸ6; m‰°ê‰•¿±çJqo_I˜“‚Þ{â©^¨ðêš=ê@¸ KÄð 6ø¯Ðׇ›‚€ø¨XA~kîöj„,'*Ê eš½é§í ´f[/¿úžvŽ :¬³Ì¥¡ú UÙ"™X¬œõOÂLAÎmgAæãîÔo`çë¼Uü`e´€³UëCQ¬‚AµgšV`«ôÎ3¹§Ô ¼Ñx¬L¸Ls¡þ ¥þ6³©E8bWOm Ý"\z³¥áJbçœÉyt ¶,Œí­4ó´Þñ@—ÿȇlëzÃ7ÅÖ…±‰€šp]aŸ-ÖîW¾¯'cÖ¤£Ë¦Ïš•ƒ¤¦?tLF¹©u¤i>èm2\äâ±ï¨Ž© R±2ãÆ£€ç_ýŠ ‡:[Ì÷.œïG¥Ú¼´IЂÂZÈŒœ+Nðb$¸¤l¸èR);¡P°u±FÆìäÅ; ¡T#×Ò„½dChd"-ã>ÈÓ•™*P¡ÖÛ–1ë_e_ÍȘ%f¸Ž²¥ŒâíõúRþJÑN*ªj_p°$Ü‘ÖÞýŸZÐÕý2ETW êÁâdž¥'é8ZÑud‘Îý<1ûžû„Œí¶ô<5ºJ^Kæ©X)¿ŠíÒzÜ:\#z‚æd%/ARy¿-Ã/›0ýEôsJ®i=”Êþ¸ñý-Œ:…]ɨ3U8#U;EБ/ÔÃeL¤>mËJCÜ%Ú¨7c·>M(LQõÇÈO‘`{8j6E8K(U„‡¼Òä‘›>)/”ü’TÇë”sÖŒ’±sbà‚=‹ òØKöó· OéHŽFëLÔȱ²"8,0.â\õhVÓKÕE²Êq4»ið“Æ>GÃKÄD¹ZP(ìÖêM¯`Òêî7“俉òâzá=r¼êJƒe†‘|Ôkémk̯ÍC³˜1]N»_¤CÆ+Âó[rñ<ëk)y¼S)%¸‰·ƒ´øi—üòÿ4XUBL¢6ùšØñòø”õy#kV~Â" ‘ˆwÈpú?—˜‘/àaôÚÃ[oµu™S=®¼_â÷1:@º›RÌß­G+Èõ}mIÝ!Õ^uVÉH0ø]BŠ>`êv0q‚—Ô(a¾m¿l•>öP]u•ªÊ×…â>²€)_j?ÜBì8ØV|p‰î\ŽøؘÁƒú?ÓZ’À ³8Ó„… &Ð—Ö[Ð$!¿„û›3 !Ø"¿fÍ@—±ô»ºW0QEñûï·¨ã‚kš6 ²lÞÞÌ«ÇeI¾&å|^ÚrpPÐŽÝ!/1fú?Òo[MÃdMéÒ}nc@s$'(7ÝýþìÂL¾x [Â>{›ßhnt€kýjPRn/žá;p© _ =è!=Ý5+¸cnñ‰cëÎU¹,­S™±FÈCï(Úh°iV©äË‚ wçz¯ø~Oc b¦Ã˜];èi¿Ã¨Š~ê¨úÝl©¤@ìlµ`á f« ûöH9K·ôW!kµ´W‚Sã¬ü:$+¸9tAò•Ö7ìîµ6{"ƒ£wñõhªrÆ Vÿ§ bVƒfêV¼´'F-Ž‰Ø?x̤ùÜ) mX…ÛÚJÕöQ¼¦$‚ „´«åž¨Ó7Ž\[ä¯T‚±§1^v:ùKÌ©n'ˆü¥{4¬ª¶ç)WÉ™f†Â]KÊ÷Ê!­–ÈÂsµöcd«Ç‰rü>$> Ëqùr4J÷¦CÃÄ[õÓ.X–LûS‰’x°›qLŸ¹DA6t²Ä¯“‚L/ÇI'b8¬VRV¶†“¼%9æ<žݼ?ãaãϼm«†hìÂ]NA4_}^ÎìÖâ>¶ú“Í×{Dw/MTOj¹&K-ÞuýT\»7ÑËPy]õ©qöèÇ'-9»Ö—qÅ$bAzA™óe6tró‡-EÖncÀ¯u>)Æ(È–—vã© Ú›Xy¿rè@À–šÂÄô…ù:áÕÏçvcu+uU²MZëqí™àŸø *Ó¸ DÃÞ ÀÆ}ôê³´ïãÚ§30Š“Gc4äú¶U\Z/²™.p„# )rcœ6<©ÇÌîÀõ©)‚ï´3ˆ‚³lû Öƒ-˜’ÞØDö°Wé[ °> endobj 1186 0 obj << /Length1 725 /Length2 35107 /Length3 0 /Length 35584 /Filter /FlateDecode >> stream xÚlzcfAmÛv÷4¾¶mÛ¶mÛ¶mLÛ¶5m›Ó=mÛÆ›Ý}»/vãÅý“•y*êÔ©¼q+«.)¡˜½‹Š§ƒ)#-#@YLE…‘“À@Ç CJ*ìdjèbio'bèbÊP75(›:™L Œ0¤a{O'Ks …1å?5CK[K'€š½½›¥±€ÇÍÍMÀÍÙ•ÎÉ•îo'eSS€‹…)ÀÌÒÆ ,¯ ))' —Sˆ›Ú™:Ú\l,2–ƦvΦ”3{'€Í¿c{;Ëpr¦û';7S'—¿ÄÌœìm²¢*‚bòr*Qaza€¡ @Fâow;g®¿hSc—ÌŽàòËÆð?–ѬÿD<þeÑÁ02L,]F¦æ–v0ôÿÐMÒÎÌÀö/·‰«Ã¿C 9ÿå ø«!%ÀÄÔì/ÚÕÆFÎÐÖ@!loëàêbêµ71u²üC{w'Ë¿žC m-m<ÿ¿àÿ‹t1ü+† ù_ùþå²t³ô05Q°tù«·‹“«é¿Üê¦ÿZYSKWÛu7ý«ÉßU5±·³ñü¿#þÊ?Ç£W—WV’¥þOü3,jglobigPvù«§¡“Éÿ +Zþ+ƒþÍåß`ümYC'K€6Ý?R„áÏ¿-Ýÿ‡²÷ð¦eb`Ð2³1™YÌ ¾ÿE×ØÕÉÉÔÎ埢ÿË¿ÛÿL&SSSc˜õ{cî«ôÖ° ?Ñ¢¹Jpê¹UêFàÕéÞkmdË! ýƒKœ;œ|w8¼Ù@#Þ‹mÉ»"ÃxS¸£É©ÂÝœ—âJs¿Ú?oŠ¯ zW`jضÆr`ב·.™¿×ƒ¨¦X²sÏj¥Ëx‘Åc Ê;#m%¬->؉Sê’£di}Á™îQ× ÍY„^ñ¿ì®ƒ„çïÀ]“bë[V:„k4\Z kk*Ì3vK`æiïcc‹ü #Ûr«üŒÚ ¼»dÞ…ÈØt\9Ðy̧Y€ˆ*¿æ‘ßÜ à—Å“¿á‚ÌYú®âQŠÿóÈËRÙåy°3ëöì7£®®Cîà2C®»%µ JÄ&{à=j»Ë•û[zÙEÆqCÀ{‚sc|M… ¯Âòh.Qƒ¨*º=*yáÊé´ðÍ›×f›TH)ÂeÃDÔlÃ$/ÐîNFø$ z줽]* ®‚ù è€~pÌ/A4ÒBj‘c„u5ÌÚ×Ý–úäbù¸Oƒéâö¾‡±)zš*,&5ã;ÝÄñš¶‘Y} v«¶ÝÍPì‹NøjW’t<ñ‚f¥š.¾£â¨—ì}ìÊVe›Žà@8¹±Þ ‡ïåm¯k[“jÙJ/(œ°¹ÕNß ûœv¦ž9~O=Íšy <"Mê¶Myp”gõ¿¥«ß¤R6—õD0 SÕzǤ¡,마ygj×B6"Zw—2~EmÏìNÞâÄJ¤ù4èuñjÃzs•p͆€?s“»_Í8ôÇS]}I¥¥@£¹O!¾kàC.­¨cx3< œÑÃôÄ…ý®n7>+«Š.P:6Pµ|O¼lOÇÚì¦H|=>Sï|Dy$“”£Î‰I2‚W’ ånÚëúé`q€äê¡!V}ï±ßG§í)tÏMh(ò1ƒ¥,ŨóÄ IC7˜ÌE ÃÙ7<ÕÃI"ì<~wK׶ôjm ÕŸP »Cž\•ªJ„½ž·zi3–AëñÙŒìC°ÓïÂfÄ‚-Ç'^šYJ]’q1Ïͱ• D4qθröññ6B¼ªX•¢™:ö®©ÉÄUæ:é§WÐ×VÝÖ®^t›wfŠj ‘>'¾Äð47‚ÿ3€à &{6BÎp‘fÚ·< ­û Ö4n¨ñ°„“Ír” ãSuž|¥q¯¾Â¤LƯÈA–a÷ç˜m,"/˜¸ŒÐŽ`âóœÞLF<žRŽæÀs³. ¶ëc½¬sVL^ëD eŽi‰z½tó¾ Ù¶éguC>à#©Ì¡Ê’.LXFÎ-X„[ÝÜÛœøèåñöØr[ß9ßÌŽ~cS¡öÌ]YNAß*d÷8gòZb;CÊ[e´=õIæ+! zþMë¶ÐuBk+ÚsB@àå1‹I Ó+¿Èœ…Jщ㊟ƒ;Ëî7)ê»úg°£xGÅq•Wjd¼I5M±S{!aú$áúáO®N„÷€Ô®vŽÄ|ËLu]ëMï-šlžD‘çÚët^ë)—A$¹”öu&¡ìÁLjûÕ'´úï‹wµ !‹bMýÄñCõ˜ÍärV†u|c¬ÏÙ7*×Í¿>ƒg`ŒxÍUŽ¨‚XKµïÛœXï–,˜·KsóZa/‰É©íT¨T|ÞaKKâ ©­ënlézzôWÖëÖmk¹Ù<™/ÏB™XÒæ³ok(y±=ùaE{g€€ÕZ¡¦NŽŸ·³â¹6[" ´ËZµK°¡Ò /{xŸ³Ÿ·<šül¯¶²ŠÔÂa œ¨#R"ÓLB‰%·{Œÿfx’Ñ÷ Ù¦»àåk%pîé±?\àŽ´!3ï)a´¾$Í›+‹ÎwÊ^Îh[Øê¢ 3ZGêväÂVÂFHF Ðzùkü·s 4b•ÎŸQlB:lÁý >ý„·›oŒ†&xˈÁ$ÂÞÓ{–8xêUå`T®,6½X€ ¤Õ`4ÎH쑲¾ãùÞª…>À¬‚|˜¨Y½§ß [Ž¨õ$]4?P&·Ä M%·¶=ð¤–%îÙ}×îG,›|vÏñ†ÇU7‡kì«ÖË„A̺=64x¿žôŸ2×¼žlûuD~Æ‚ºFz¶ô_ƒ†r5˜zu®ó%ïÃ^îÒ)ê ®â*ØkáÂ<¨Â4¤5@Ú³ÁØãêÕ0N6xÇÔ¿ ê±S­WQ‚:ÕåÁ#zÂGpC"h“ŽÉ™ëu¤¤Ð#5çýÀƒb&ÇüÔ¤J'}@lÝ!ày›çbðÏ|]ì­©hTØäEß÷ð¥3ÅrìèËïx_j3ôΪ4õ·©îÔ™\¿ì¸-ÐÚñ›~›, âا³d´ÈS6zˆÑ3U•ðúø8ìÅLa£%IM#ãmí r¡´Í¹Òª#ÇS]H:݂ߎs01»²M¿Çº‰ë¯1‹Ô´TZ×m„…«h R÷Ó9àJLÑŽTAD=ýf7®ÚËœZjçqF¾~Û$,ÒF# mé²Åúÿ×n"BC'ús°×k6oµòÞ‡û¾}`Ï»’7<çtod¯Lúœ–äK€).<æt0-+™5Éçý¿#ÖRù{íW3<¶‰ïdËQ²†^Q ˆö«3ç–±ôÇ|ǘyŠЂ%9æF¢*ü&J'ax˜ÞÈh(A‡¾Œ‹h[pä{Ê­hg 6Wú.¬q¥l$/—øl,%R÷Æ” :ó-ÿA>ÂäW¢‰W4,êyPÄT‡´0¡én?šë&‘ØOaØö+ 1¯áÃ<TnNkÿ ~Eti¸Ë'hZ½’’‡,ªà´¢ŒF&BŒ>ßë>„t;¾±}k·G¶Ô7tÚsÒÇxâö]OM%Ä[Y8¹±ð.ÀO€tX’ @u8B iì-•\P„w繉L왂«/RªG'2úÏÏŠ8v)^.-áö'N¢(çN Žú‚Î[ìņ»ÛÎ^‚Bò<:•çoŽÇ gY‚ߧ ‰™ÂÈ”iÒS>dádóD:ÏŠ«rßÿáÕY…šo¥n,•JD U±`–§\œß _þ½R!, ™ÚáŒö¾<ܧAøÖ\ ±Ýv(v†m8Š(øÓŠùç{†8”õ>l¦5†äÙOšTÅ™0øÁ†áÈó"‘‚Ÿ4ÂMßÃŽŽÌ¡aµ8·Ü’4j&ŸÄÄðî}èßqê~Ô’?+MÛìÈÏùÏñ჉Ÿ’¦EÑ1§!îv¨áGP#wtÝDRË´ˆXLe>‘S°×çn(ŒhJßM² ¥˜S˜fJíÈ\:•å/j^îS¸‘œéŠÚ;}u;Ä“S..»²à[”‚K A„po#|?Š%)|Û )sºvZú³cWH^xÎé ®‹HÎT4nmg$çX|~¹¢$pîHßÐ&úFrqÄ¡ˆû£ê}û‰òk}˜å\‘o5CªwÍ÷öW¬ž]Æ\Sy7}e׊¹´ YxtŸàýÔ(ˆ$Ë5¨[Olù|ýÂÛÀB_È•DÜW¤ª·R×?þd4š%¡Ÿ@’¤¿û[Ø8:ßUð¬µ;,Œ±¾XO(ø½W¯-çæ7)7ùÎÊp°²÷’X¥m“hÎü­Zaâ+²- kÑr- …yÞ#" ÇU.od-òd÷‹‘GÖ4EpVò¤Åá2 ô)-?za¶Í«é%•Œ<ÑŸÒäŒk¾„ƒÇ(åBºò\Ÿ»Öjf<òUÛïÌ°FÖæÍÁ“’Ú‡÷F¥Iãbw±ºY¹lvÒq—s ,çË©“$ ƒa²dBdØæ¾ÕÉ?v;[WHÆþxäH©LäÑcЋíðoÉ—ŸÇ áz¾Hબ);iÉáîQDatO=9moxf¢(Â*)mäø_„~ã6}ÏÑ9>¢²)ÓTÛHM,Og˜Ø´[»Ý°L‘‘­ñ€¥j®˜%öUIfjáï>A¶z]]Çlˆ( ‘@Öœëoaýá´]Õ¾ÂðäWÀ´BΓ¿dÛÕ€U —£)ö «& ‡j˜Á‹SmÓ˜7<>lÎ^yˆfÑv{äP,÷DÓq¾B¥`A‚Í{¬ÂýJäCÏ{†®÷„ù]äϼsyÍöÃW¸ãÙbp¶þþWÝ°· Ä'«Êþ´€¥”öé}ÓObñ§û&Oô„ŽZ­Ê N|—Ãe9m>ªœûºlŸgdddï¾tõO(þñ~;THC®Ç ~¦Yók¶m¸›¼25Eht¶ïºsNy3“ —¬¡Ò²3Éá9[Ki¶É´ˆŸ<"×;Y¤U˜ÎÐôš +®ò?ÌàÉqMü°ðó¢{'ßB‹s´sš›wºÍD´$ΣÓõ•«Ú¯YÔˆ·»#O•w÷ž³qnT‹Õæ,&ÏpÛ‘©d¾ÕWù/ ¨sj¡gG´ê˜@ÉÿÌÊ>ô®Pí+†h±CÌéµ?Ñ –¸0ñ™ðÓŒsEÙÛe0ëœËf4?šw`!÷æ¼*NFbÐUBC={аh^Ü&‹þ‹ê±y“ÓDUŠ@á§ÀJ:úÃÅÞ+³~JØ.ë—D´Ï‰§:o­«daL«Ô‰1â1ä' hSc;ΖOç^²?‚v“Áß‹(*þi¼ ’ ý?ª“hC«RvõyPR›a"¦^vE ûø-ù¾/ÁðbÅHÖkS‡2àÛ];¦$±%ýÿ`”¡1³œñúšá\p°MmÖ…‚¡2EYϬ‘å žcälÂûÞñFÓ IkýüúÕ)o±_£ëÜAbÓQð¸ƒ_Þ¥·¸Ìò9?fÓã}6ÞVgx×(úÉõê#ñ— ‘ë8©  lÊÔ:®ïžr_0n‡u]Ü~?1vc¿w‰/t(·Ÿ«ˆË$6²y ©…±¹»_x ×M™Èê×àm$Õ«AÄÑ©­É[?à Së]tuñ&”]æVãdÛa=Èp¨µÈ«î¬„Φ3q fÙ„À¼ÕЗ DÁ½Ô“¹LæyÝÝш>½2²ÊÁtnT䎥¾ƒÐÌ¿±ÅîÖ±©ñb9ùž°Õ­(ë–‡µÖÌgá¦I:eH0â!] i*†¸¯Ö?Á}4a èÝcüì¦Wø|3S>qÙõ°«˜gÇl¦awZ„ûz¼1•ÍMˮ՛VüŠ! tÞ!©>lÅ DŸ…_ÖïîQ(iÑë˜1”’RÿD9½õS?F,5$¥/ ÛQß<<:óé?sBã—ÏQZÌ*èþÜô{Ñ9—ƒ´ø9‚]k Ö¤V‚<ÍöÑC9phÙ×Âœù±'j¡ö2Ìžð‘´°˜a´…óõù]ÍçmLøÆòC5huºÛ•C>õÖë!®É!Ÿd%KÞp 9kø7λK¥Y„Ö†æÐ7a†‚ñÍv­þ&<ÂÉš¥Î–95ÈNß>9ãÏ4Ö¦qË Ô‘pC#·Y*SI±Vo¬ø…L^ÃßU¬ü¾º†/æ«\@Wq¦ÐTÄ­JT&ûª:~fã.tƱïÑîE¾º÷Ðrso͘h2¶q` ²6­ÎLÖß®«ò›Ï¿ˆª |p¹öl ‰7wq²]fb9ÆŽªd§“x›Jsô© ˆ{©†õ?6âN¨ÈÒx!W%Êæá’„?.ˆ aÆ–lž4äzž¯ÖŸ°–ÖöÖ„Òê€'ËñV]ŒÂñÔõÕª¶U2¢ŸB¸/—¤¶—M™œ‡c!7}s¹Zµ7ÙÙ ïôžÁT©Â™›ó~Õ§íÉçÅæ®>… nØÊŒ Ê$ÄÑÉRÏŽË;%w ɸöX"u½Ð—s+­v•¸;éRMt Ñ„ê!IÙþmJŠ7b:©SÛûÍhcMaÕF‡èf¹ÚgÓ‡Ñf£•:ý&¥Pñ¤ÁÚhÉ 'µ$oYN!âŸN‚N cê£r0T¶ÙCÀ!œŸ8ç:É@Î@BòÊÛ³¦–~ÌÇ-[·ºÅà¹~O|Å,ÏbZ¦aUñVÊÇlÁ'3šÛg£Î ¦0ªß¼´†%ûþ¿g3«¼±ø ø¢¢q}šS)ÖÑ, ".´½VÖŸ—l fº È/5³Q8CuZjÊ6Ç÷ þûÈÙ•KEúr¯±#D½BÏ‘0 á1em˜‚Š«¢]»*Só‹–…õK©ý8þazAý¦£²ÚÝGXcÃ…/ËõÕèØsåƒO¿„j`Ã,9ˆÀWGøt c5T³P÷6³ìÑËýê>¢#õðâäf6cN=m"-«€[ Ù3Ž$~Ü[n´Ö$¶m7О£Šµ?dü¾?¤âÚ<8‰•¾ÝBᦄiÑW ’‘é@†Ê"+?4‡ùýX>‚=ÍÓCžØÃBhj[)¥d,ˆÖQ­§‚±úÄ·Šðs£¢ð}Ïqe8ˆn‘<øcº¬èëp$Ùmѧdm–³3Íl4a6‹q¾‘­w“ù®qïb1Ü^®`Ôqckt NZœéXèÏ4[-±yƒ®ÿ$Š5óŽVT~'…PbþÑëy×R{¹ oáæÝ0ÌÐ$ÍÃ~-´WE1I@I1"ŸÙË1Z<0FÝÆ›–,“SŸ~[¼Ï õ±•=Ò«hlXOãÃÇK]¶¦‚DÅ…dÆëaD-‡:®mÓ泩Ü…%Lnò’Ñ:Flÿ×n‹ø Kî—†Z†XŒ(£n/oN³aÜr¹¤¡ 1Ç<Û{ÚÀ(5J£¸k€4‰Ã‰à'—(úŠ;,NÓü±WµÈ,Ó/ >àÒäàkÆx*P4Ë}h=ÎtËn ¾×¦C#=*ú¯ý©´¼èO!FF¿=8©4ånz¦¢]“ Õ Þú-Úµ5ýèÎîÒw­y%˜= (܇â–©Äy¶ÖVå[3ܕ·KGmôžúô¡JØyZk¿ä'öF]ÍšK äÆî??·§K¦èGùýz.džà îˆCËCÍw˜L­º3ƒs}'šó"iÿØ–¶,Ž‹ÛgÃuäw˜êMCË4y~ÞsÔÊê¡U”úóƒBÔZ¾þ±òZ=ö̇$Ð2y³”ÊeyK”Œäˆ´Ž‰a³=´Z÷IÝ«v6œè±6e ¿q?ÎŒ‡ñî¥RyÆ/Ùã]`â,O<=seÄ”n—õ€WJÉšz?H¹mgñ$Œ›ÐÐLò§Š»ËÆŒü«"¶I–eO‘m#Ò!Û »dÏl1—ôK.°%Ú»——y'v € Ò'¶”ßÚålXÎݤúTí}]ŒF” Gã-V—šçý‹ß†O}÷îjm0pöPt@S4{¾±ëè¤åG·2jšÕ ùM§X X„FÕZ"V÷T¥ ™oëèX¾¢_˜ÖmwKÞ£XîÄL*Ì–®ÅßȲò³Ã²†:êîŽ22¨blëÜhÞ u÷ÿ÷¥õ— mYB0C(—P¡"¥I°Ö2?Ô@íù'Î]¯ÞYW®^(Nh¹c&ëÁŽªfñbÏ!R÷áµ}q%—¨s¯b ůã»Oë./ƒ‡+ŸÕyZþ*~ÉÃÞù× · ùü ˆbÖ“2,*¢4=ú’G]p5å })!¾\ã«CV4 Ê\<ùaœë‹F®öä0åÉÁof¨U—`"HT<ÿ´9¾B6–DßžýGFèìÖÛFÕ}U¡ŒK@zÿ•címµL¿'â~Ó‚s“­–ãà5;Y3÷ lîößjü°Ìq‘LŠÇ¿Ÿ`>…x1 AÄì¨9õ†vE¿I˜Û@ü]fpF8YžšóÕÀñmvze¬6¸,×xÖvFæ‡ÿ Æùײà`Xâg±×x°Õù›óˆG½ùŽ¥Õ¶¥Vÿézí:Ô§"Ø(6ŸLò"Îêü¨YPVs– Ç>:Mì>ÚØ¡PÃ,] T¿Àõ(RR:¶Ñ˜#L8‚Z/U0ŸTz’nØêqrû•XT¢½E9œŽ,½KrÀ)ýüNøÜ#1#YƒÉ¦)Wä9î>Ö—o¸7ÑHr$oŸû E¥dŠx^þ!¦Â:×·kZht ì‘o6ükBBr…ˇªsöˆtÚÍç!ìÁÊ´èþ–ø}'8O ,ãƨfˆ÷”§*ÞK:Ò%Îqv(Ð-œˆ«ڛȢ7ÆL¶ßc\"û~q@ü"*ç~a*™jõ‹ ¡BM­qö)âjÄ?¢ÿi=D aâ8ô®,ñrÂ]&CP©fE“ݬ£¹XNò…b xÜAèßÆC%{ý›q0ßš¬\‘h˽¯5x©Ñ£¢l~‚üÍ#Ã|êCúSw¿y®8fZÊnþðŠþ>o˜‹N¹8$H·¡è‰š}È1QÀasâìVmã™ï7’ÈÍô”Ê1Ùüiü œbë GʺšÞ)Œì:¨vaêgé°` "Pò€IE(¢d¦pú£Üçð ô‡ø€²×@c6y—,rª„}™õ.ðvdò"ù8 89`ÏË—]8CG‹Íº™ÚNLJ4[À¶'æYÅKfN»Öuÿœ8œÝ>ÚÁƒàÞD-aÜ©‹0©þ³VÊ*>¨È®Êºà¶‹ 3 <šhŠ¼P:Ó>ÓÚx¬ßÞ]½j®íö…ÅL¶dnŠ(XXœC¢·Ÿ¬ž[C $Ûáá&³p~õÙµ D¾fxÿ1o<„XØ©@­§5Ê…êÕŒµ$bIi™ÉŒ.;Çf¢¾r6ÈÔÑS3W¯n:¡çÛéDÚßxE‡H ˆ¤Î[m_ŠÜ9*í m!se`QŒÉÂCãÎÇóE‡H^‚Òðù½£JJ'WFƒÃ„¹àÚCÞ*HöJ>yBkqG@X]¬\'Ü—Þ‡^¾©Õ~ãIvÃA4ßëߺfqØšNTÀç2¨E\^—ìyGdtã’‘qM­¸¾ùƒÔ*O´¾‰qc,”‚´ ñ«[Ì·Ÿ;²òa¥#4<X\4ðf©%ÅÕ{âOní ØÏlá»Ik®ÚJ Š(¬í|áÅI⃗¯ sÓ¹F &F µ„tÇ-YéS´šÜlZÌË6È?´7ž´#þÁZ nná~)Ö=¾àɪåe¸ÈRjä­ÅUnÐy‡%2î±Å`ô«K¦ þ‘Í 1! òè²ä¢´ÓGÌx[Ert1ÎätþZeï¦Ö'Îíi4\¹Âã KtÏÛ˲n•˜9-«TóÈù¬žòG¯œHä¬3e¡#…És¦ŒýôîwCµôèPò:ê ,iy§YÛÖoªdOk5© ’z×J—x¶Õì\ÝÙ®<ñ`on[Â8œÂY·þg­P4qæ(ƒ{¡Jû]W8Ï%aŽ’mÒ»©DBhh– ÌäâxRMágv•Ý(h-P…oþúÎÓ1ûk``Ï–â3Z µÂˆ¦<ÖÙÄ2‰U(øíào]'kDPmÑØë–ÃÔKÙ‡;C;¦±Yg˜õh$%qñìýÁÔÉ”Îâ))QhµsÃ)"EŸÛ›ò &m2ÌQš*‡3cÚOç8tÕ%ÊÃD;mn*óP­SIô» \ $nï!Ù ¤{í%îöhßÍÑãÖ¡{»j§=;†h;eéRë’ÀÇó‰òÄ•*3xö{Rö …DŠmkçr`m"š>’Ua±,Ú:â›g2)3ž:ˆái“"2³?Ä>ºƒÂÆ·nx6Ìïh –œÅF“o÷b~¿ƒâZ°ßp ö‡¸ ¥<Èdûµ®º'ò‹Ç<@Ž¿ØÃp^ «»¹=Åghr^{qûthR2àU8'Ô¶4 ²ÜôŒ…$`K<ƒ‰üá¹\,™`È©‡)ÒÁ˜Â¢ß›¡ôJ n¥ßÿ%'ñäÔÙ¸& {‰Þ“(‰†óæ™ßcf„SÏ㈋#5Å—ÙþU„>VÇ×­¯–Hà —™Ñ$[&“¡0ÝÀd{;T˜k î3åº ½Œmvy®$÷r¾¬nôÉÿXö—ÛCþ½™N¼3hF¶Kì¾;•ºÿUýƒ€ Hi·ŠäâaƦŽU'cô6ƒÕ³¦"âWn~çål\ׂ'êGm<½s$Š8 Eܶ|IS«+ùÜ×Â-èÏM–(Yojðûîé¡k_À”ßïnŒÇ<¦u.ðÍíI„ 'þ׋F¶.66xBÚÈÕ¼9QêjC WÕáG…Ž¶_ñÛçú!(b@xš%<`Á[/q¾šø{æ?ŠŒ`SF¼Î$øé"û××\æ¼›"´•W B)Û‘^¥ho aîôGèñ:aþ/Ììv¤”àY‡øïëº-7JáVÌÎ¸× ¡ATPšÉã˜Z¸UÄŽn›…žøȹÀØrLF1Є@¡«$#“8&è8od™Ì‘ñýÿ ¶ »œYÓLS®0û‚½©j§Tü÷/•¿È˜,çšÏóÇnÉ+j¦OgÚ@Weï¦g5‰®#ÀûOö§ÊVzâmžWk=¯÷B‡wT%˜0•š4»{²ãa UA©çjNŽ ÁÓá âŒG˜N c‡÷Ô±“òОDÄ*ê¼´ø–"ž²¯‚È[°å*¥‘×í » i: ’]¿Ö0 tÏÉw–ý®°c}sûhØ”[?±m”& ïmǧpéšêøÁ*ò¹ÐU¯ÊŒŸö•Ö¯[žb  sA/Yp÷dÍóQEÔþÞÛ겆­ÆÙ7AE¦YÝß¼‰³6úLÜŽªëð…Èÿlö° Â¥oL )*Åî|¼“)Lׂ"ìùv˜B&Ò³ØbÔAo¡{®'ñ>/ïÏ~Q¹÷‹€åPëßaí²S‘ÙlNp9‰Ù/Ô~ýBUk×ièœyƒejÏÊÏa«€ñà–°~`=&—ýƒ×èÈBŸ'í©›¨å¸3ŒÈ×LÑOw±I­1Q¦(2î®íN&6¤ yŒt4–>°‘Õ;L‚8Ò$Jÿ)ÖDº3p›×¬Ñ^‹N.©exÃZOêmþ U°ä—qøɲœÆGi-D7ʱÁë¬!MºòùG V»LâØ*Ýœ¢goÉãá”mÏ9ND:#)\í#û¸"oï[‘éDÒd&f‡qÄX[_Só¢íÏÌ“tuGœJ¢®é ù¹‚r1=,ŸÜ‹ÒSÖëõþÞóºAŸf+/Õù=8Œx~\¬ÁœX‚7H‹¦+Ç Ïç!âcxH r%ŸG?£/3¶/’1öVC|–Sž$"~K{,’ #bádõ>·jOP+Ù ¢Ú›I‡™Jí®8or·Ü–Dê~Âç 0’Á€/<4jž¼6_õ° qîµ$ÿ`i¤ÙDáû‚2IpÑH¢{Qk¼J6•Àü~îãÎlþÊX­Ö ô@R‚™“OÊ:Ú×~. _#§±Z¿†*ÑÖ:NÚ´-òzÏÖ²jd¬[¤ã+,l”ý×­®ÿY¨ðà²|mÙ6ŠïK=–gpÃÛ ‡¶y¨xjö¦•võ}l´$e8¥v#m…Âoc&ùÀÆÜdâï³È¥QØÌÃè]¾!Î"ø΋> Ó÷VžqÖ2æ°:l»óŸZ3ö•·ý1ûdÓO†<S*(nÏ"wÁÄÅô*è¶Ð¥ÑGUËi>=J‡ÈxIJú±ÀJ–°Ý;›K_\rnÏÏÆ!×1ó„»Ïxøêâs9vŒOâ/u¢îèÔÓ”o8+£˜â¨{t¢R븟þô‹‹ÜfJ)¡ãõ}½ BÏ úî'‡QOF¾G¯:˜¸ø5»H`½ñ#.2Ô½ï+Œ² ŠãS)eNamR¨ú.Q:ŠŸ\ÔÇ&ˆ;s«éÛF9ŠAnPÜåÄ‘Ü-RÎÎ?·Ó+XêÒÙ/în~ÏMþ(„«ùÜ'®™C¡µ”ý²E»!,ý¥LóÊ|È‘O0ÀxöZ¨©n«IŒØL1sLC2€È»–21ÃLè,3Z£ÝbÔipmreJH™ŠôfÆGà*6Hð%æ9¶«WÚ/Ÿ†Š ŸÕvŠH d(ó}èéœ~Ôß„Ÿ¿ñ5w!šÙÅ>1ü¥cé©á2å^Ì_ÑŒm¸bŸ<øJ¶HXÔ¨2n÷¤Š* ìHS±/BfH+׺¤+G•m¼#ð3U<ül§ÂôžàÍ€úôäk‘y2Lz?q‹rèÝ`ýpQèÈ^XÅ­ï=A×¹Êj 3ÈÄÅôòÔwgÖ/X'Ô.¦t3K6dî3AçfÍ©ŠÆsÇ*¹ã2\Y;uÞ¥¬ñ»5ž›À«6 Z*I 'R1=_Os¨þ~¡µª¬[™½­t&J”ðl=ƒòy€¤x²#¿4q4ªP¢^ Ŧªb÷]\kŪd §LáVæÕwP·'ª+Ë57>/êȬ„&f9uRòÈìð‰|%oâî'Ð éKÐX3ÐUráòçxMÍéÎ#çJ`}"ãí÷1¥Ì@^Ö}Å—»H)¡Ÿ¯.D°Sr£érÈ€‘Ê wš»u´‘çmœ ÙTÄx¤ñoM;‡D7ÕzÙÒ¥ó·àkȱ®Ùõ÷ðv(™>_>ɲÜ©¾{®˜] jŽw¢OVafùaŠŠ U¡3*¯r€©tÇôŒ¾²÷é”îIä/NGZ1ÏäD¤Lˆk“¯ÔáÒ9j¸ƒÂÜb‰Û²BÍO&ô+ˆñ(›Î 9G²A‘R¡“ oŒH} ]ÄN^Œá´¸Õ…*óù?BéâUùYc8ݺpaøKál‘8Äæ>7†®!ó.ûù¡øùLªïœ7Á8}h· bXA-R+<¨°cEå²ÆÔüag»õ”ºmdÀA´\ê îÖrŸÜÝ-¼‰Æ[±«è "DzWA ÍïòL)à—O=µù¾õ$ÄøoHv³-ìFY"H©ûS0ÿ„³1Y°<‹"Š¦âÉ,²eˆ‹úÁ>ôü’:¯"Å9ÉDå8U‚QM÷¬lœÌðýh!™¾£ë `”$e߉D-Æ °³ŠsŒ×›XÇJ®j";M"Te6Á…A3zôs’ BýÒ—ae×ä,–ã{&'>G8Ç{?™)ŽUÄ!Ç$»”,¼ÔîC?~±‚2Õèg Ž€˜• ÔA +É "ÉÆXzÏżý©lÌ™/dz Ko¿ð¡×]5W¬Úˆ@¬Ì²Ú£4"§"©Ù̧‘UÍ¥'ótœˆH}žš,$ß\˜¿ËM޸Ǿ1; 6(Ö]´%îs8Môø®¾Õ?Ü­æ{r1Ð\AJßnöp}殂!Ö“óúÖÞà %÷ö$]¡-”…ÔÙ$‘â—Ma÷™U ½dJc5úS#ü×=¡¢§Óg#ד ÒQ½f*3e7ªcñ|… ˜%½ƒìIo8*-EÅ%¹ˆæ - •=² ÅB‡À:ó„:¦dòÔtÜZi¡1ƒFÇy;‘ªm‹ +~¶´:³í%Mß\„†È²ñTÈÁ#ؘu’ú÷+©¢d0IE‹;e‚?ø@W¼(Á‘–øæÖÛ0%…ý…—ŠHôBR9põ@oIxøô g1¡ÑìAAä1H D±ìÈü¼¼4Æ—î€!8 Âß¡­Nâ"+,;m°2\!é/¦Lå{ð™Adeœû-øÙœÊí¨ûÝÑlâÙ¹w"lØGC¡ßcÆ`î3@ÕíkŸ[£RÕ=Ð|Œ #ZcÀû7­‚Ðø! !äó H‰[úM‡u™âVÇ䮩UɬܪÚæçwOË|_S#GO¥þ–y}jIhJ*>s˜]:¾ŒmAj“ÙO™‹HjH$CWœ¶c`)™ož®½Kî°½³ÀtòÚ'ñ¿Y) Lð8£þ,³41ˆù¶Fûmv ̬ö¼Ò’èÿzP@ ü¹ðœ‡Õ9ˆŸ µ Q;½^[¹ëTUïîg¯¾`Wªy ŽùêZánÿ¬Ö²)W3¦_4ë¥iõËúOª‘®h…³JœýšBk©:c!„ÑÒ‚Z¼RÔ?š›&·Ý‹gª@>K ïîÐïù¿þç°+ € ÑkÛ¶mÛ¶mÛ¶mÛ¶õ®mÛ¶îdfÕŸPÉ©³ª”ØBw~‘ g¼T„pÌ=0¬Ë)p„¡¯Sä†ÀF¾*làGï°D€È£¥äQZÝ?û·G—Êá˜öW2z¬5]Ô³xsSœúŸNL‚gX¶ñÚ]}~SMÁ8ÀutμO³B¢Gì<\.ʲýÝðæé"‹š :yf­»4×ÐÓ"2‰õÙ/#®9ì˜HK¨2µTêVµ‘È|Ò}¬ À£¹&aÁ'àë~¢~$(´æ+àjkŒóv[o Qlš]¼ %˜x¹- ~þùUГýöi€œxöá}L$í¾ÔOy'sॗ着h˜Ê‰@,¬§ãR†¦-ÖR”Ê uKT5ùëÊnþçÕ™ì†#rë0{Ïá)žRµh6hAXªÆÒ™3ꎪ qÆù'£àžƒ˜Ãí #è¾P_ÏŽv9ëì4šÏu)n†¯J˜L-š$±4q€ÑáRÚuåí‚GRZ/ÄÕ(›ƒù¶· þò78 nì28BÖBæ‘;/n.„ùÝüïiœ\ ÒR3ßv Z±i¹€!Ä©V®šl׶û€ÛÅúg¿º^Ë“>™È?b¹Ñ{¡C&ÅŽ–Ÿ;דN¦Õ‰UYª4˜)ýeSÓÈVÞKÚº5šÕ9Ü£LÐC²)ñPMÆ- -N´ì:š‘´[l&x~¢ô8'`d"¥DÍPXNÙäpGÃMÍ2ºTïìlëtëjâšÀ  øàñ á‰/l—NkjP+SÒþ;Ö!É$¿I®FqÉ®|\ ýÈ51ðÇjX€kѱ3¯­ÒÓÖù¥#Ï,^Ä`ÔnÄŠØðM²Ç1ìß#Ð}–•™+¡áÆKµØ¤=Îi‡à»Á:²ŒK[гˆ\V>9_1•™[ï’,ÇJô§*ð˜«y23Ö,RrdV`êµ&Š~zÞºåØj4øîf“lK›íÿl\JRpW,a÷ißfÃt­<˜¥¢¤gTÓ3$_5Ï©÷ÿ 2ŠÄXR𠜬vŠÔEØ žÏB({¥n‡øÍQFö J~ýnH×Îâå‚0â&eƒn.ß›äØ ªq£ÒÙÇOÉXv@çØbQ:h c¤2`®¥—Øt@”¥÷TJðITQ¶øû»3ˆç›=õÍ6ËܸbÅ7k¼¬¨2›âzâë†DjóïÔFÁǃRõ&“L-Y°ûôSz49¾(VS¡[nÉb.N÷(7Ÿ‹Ýü\æµ2ÂNu©=k)Xg~)‚ÈØ€álL?.ò¦Öj¦xî¡z5ï=ƒÍ«ã4+wtˆp>e ÓÞ®+l¼ó.!l;nºv±··l.´õDl;kã*uÚ˜ýbÊ’ò0±ü! < êwD¦"9€ûô¯pŽ(6fN¬ÉµÌZlêô®@wzy 9%`ÞLÃc~Pn-’¬¯ý¹o4ƒØ7Õ>¸2wô½Èü¦e?'ÅŽcù0ÿ¸ð¬"œ„.a½³I·­¤õÁÜÜû°ú®½÷œÁŽÍkfœñ­ûM Žƒ£¯,µ¸÷{€èGG·›2óOO­ÉeÚNãoz ×·@£öei( ”¾+~+‚ÂÀÁ tÉÞÕGò€­‘ÿÁŠ^¯”ƒ¿º#Á÷Ä«ÈNôo89¦x€56T‰0 Ú¯ò".cÇ¿XH› •Ž~À%îIÍTr?ÛÖë¹D¿¼…A±Dké à‹›aÆkÝ Nm0› ®@rÄ!=¿ \Œäθ¹÷.šƒ ç _¤›À›e’¨^.žç„, P,èŸ(ëO#!moOkT`}š¹Íæ%/óD²ôþûoðÝÌäP 9§¸U"wQEo/æ,À;}¨P’VåÖ¢â4ô)ÈÖtÜgý¡7ÒÕÔhŽ©Åéñçl&F¼6\qQf†½'¦š‰¡9œÍTK©Å’ªKˆ–¼#½“8;±©ÕŠ`%ÝY îZ@žMrÌ®~ª D²rqk÷©z+ûó?õ÷¢ž“EæaâÓÛ)“p¦£Îñd*÷óê³û5åBÙú¶éü þmÇ-10®'løú†Pª;œ£AÈ /—@õΈaÐ+é?%Þ~bZÊzÐ4äðB*SÄùò¿ÌJÚDtº×\L…©ØÓ8£ëpCtÚ˜1ƒ¿ËÆDÀ‹:b2±ç‚,åc$Oî¾7f—;äþƒ\58ªpŽë_¦2“îM5)Mýù²yk’96+‘»T#¤Ñ’9eÎC¯ÎO"~vsù)w^š§ßé×Oì“æ_ìçoÀJ–MŠT^28´ƒ©uù,°jpÞ™(¦¢’b ËrÓŸž#zBÞ ~ï|hûZH÷‹ÜOrçžÐø}<ÈéKåD«o2Ó4[Ç¥}è)ÔžŠ{(sÐ_­>¯²9Û~²I®L;n`%´Aßå(QY¨íoNÿ"‰Ç[ÿÜÔ«˜×c_àa ‡6êσ(¥ü¯òä‘’}¤³€Dew[žv3‘»Ÿ¯ï·WâÂ+·»æ t- :…yJGŒˆÈšÞ`èg¢pR<žÈV‡òÞ]§N$wAbq¹ðâ€ô[áZü]ôH’l’ôµä)¼ÞgÉæ5ÞØã»áÕhH7w»TúšY§brBYÏ(îÚ„Lz.Šöî_'” Qdy ,Ú<Ó«€åb˜àGÐMã¹Jl¥˜ô©ç£O¹Ðçän;ůp %ÐÃg™ç†çϹZ*ÏИI5 9-½{ùñÀ}£»Z­?Pn(U(sf˜‡ §>ÃÑáŽãÉŸÓðn!¹Ø’BÀþ1÷B&óee~ŸF³ånÏŸ™¦¡ícœÞmܶ ~UDŒúb€ÑdÆa{-7oÉCí%ÿî Ad‚²ÂXðJïâ!÷jU°È¨’“vžFÒžð™ \yúUPÓËõïÎîKi¶ §ëÀí¤Ñ”6D.~ÚÊß ±-’Šª¯­ÿ/½i€im1ÍŒ ¦ßÏýoÂ=µ£šSÉÐ’¼hU«*Á[¾WÍ2°áwRþe©w^z++ÚtÙ¾úÇÞoÖ‚Á¼Ê¸Â}A#2ik^ut$˜Ø¬½+$ v}}2kUü¤Ôp¹*¥÷¤õRË4ÛûxŸÑ†#ÃcË‚ƒÜn¶ø;5vëê,¸å+rpú>ÛáU¡®]f¡ý[RÉ\²Ð¡@…75YÝ¿£—,N(ãðVñ1èðOA´ËeKÃ&R_sŠ†(±ö)Âôûì_ ÀŒëFnqÕÚÖTIõ894¤¹ üÌàT[ÄäŠ<,éÅ6-ÿXBw³YœìJLé )*( øÒÿötæ›Ý§0 ”Š70 ýöó‹ìXÃ}H!Òh'¶ùäù°­ëÄÜÔ„¢‹£hol— °xmªácX#da+>sév¿EîúJ dJWüÏI†­“³³¬P€Yd±N’{Ù”­›eî 3b„Ù;œ-DC¸ø ú¹˜ü·çþ‘$¬ÒÕý¡ZkP Ò.>yŽÒÊùjrÛ²k¶Å¢æ]—œ·¶vøáô™Í8e–cñ:ö®x“çØåTÄ$‡±¥|>UVÔWÓîŸÍ,=…ø31_‚ñ‰ýÒU†Ã7¹G ¨Æ÷]9‘Õ~õà~ƒ8Õ `§»á:àÆ<ø¯¶ Uerýd¼Ë÷ì䲶eÀ¦á¾[;KTž“Ãyì¾6© ê§Ì‘nô:ŠÍÆF:JM°mÊ+=©µpf׎ô<ëÌTšB(X.>ç×'’É1D;¶/myàì(ïÔow]vH-2-—ÇpŠåâ쩈5̦‹–å¹#äk¢Å–L\ÉÌé,ê|Õ£’*ka£«{Â(§]—á8ê¸ûGÜê}uò߈x…>ˆòŸÇ —ĵœšt¤\Œš¾q~Î=€: >~{þ­`9Ù’‰ñ(É Úo„_8ž{³I"¥²(FÌ^eQ¥XCT‘Ü>†¾|V(DœØÈ}…Æõ­ò<¸m†Ìëþ•Å^Y—õ'xäÐMºݤÙQs÷WHLðÖ°‚ã·:ö}@ô˜Ĭ< C.{O%znã8]/ªI½eÌòÐWÃËÅñÛ1æ7Ê'ªjnCÒ§l{o»0žŽ›°`#/0ð‚ÙM!v‹ÞE(' =‹M†VR‘GÏD; Ù¥c‹ëA©z)ÍË¢ÍÔåÚ–Íl➧ÙËÁÐf@~*¥¦†G&g¿}VÜŠ•³˜ø‘ÀÿÙ‚¢¡kaã]˜ŠèMM¾wÝ >¹©³…Þt™â]gŽ’£òo¡“åµnK¥$Ú®g¾>ÜfM˜ \~i8t¬ Å·~º½{“ÓïúÊ™ïä ׈ŸpsGp:¤øÆêjñYóovß‹Ÿ½.ÛÙ‡ ^1‰ŸpXùˆ´"°P>t~CS,Á£¨å~cçšÕ`"_‹Ó5—GÞð´Œ&Ýš—V 6-eÙa5ºå| ýkCÙž©"uW³ª6™ª‹óÒ¦§@ÁE¥YëµÈ¿¯tP¯°? ² ‹ú¯ ¨ÓdËÂúÐÉÿæzštjX‘=”;8º‚¬©”i˜[‰N*/•cÒË©û•uq„üM²’êÔùS„ ¶9^fúÿôÍ9Ï ¼Ùk?ªÇÔ"~Á[XJï@P™Æ}Æ®£°TEÖâËÊåÅÿÌGWZ3#7`NñSOrm¶ýû1CÁ ÒÀÝpßö'éëŸsH_HGë%MaR&Á9ýÆô+A˜$¼‘‰ócgXc¼ˆeFœ–¿ØºÓcø ­Ðˆ+Åp‡§#ÝR|$¸ž&‘ðw)½uµ…‰Ñ›4YwÞ”¦ŽN<>ºïz@_Ji”`uË«/2Pl\;4AyacŠ¿i›¾"îÎó¿ÏšÓä„'0«B™ùòëþHYÆè‘nÁebn¿öâZ`ˆà•Ï½B½^Ó 7rÓGfÄ×{ÌL4·Í7Eœ‚ Qùeg,h‚Åx©¡:Ý’2B±K½’å -ePø±Z}0/õC«å†€ÃÑ{TªÅÆ< Ö†]î¨[\Æà@͇‡TÒ-J·=êˆ&º# ´Â¡‡ó×}oSÀHEhîiYÍr“@ƒ»zXýãÇr ® c>F§Ö8}Pc·s÷ûÊw—Ž&¼d}“Î,z\nYt¿íÀ$¬¸9@mÚ1/Ë,(éœÄÏ¡†lXE”Á]°/ÑÔ.@Ä –燿¨ì+â¨U~\Þæ¹nìEÛoÇ“uxÈõK”Þ0kñ?duË(Þ`ϲtúB·IƒÏŽL½J%¶ZuŸ6J”Ôq¶û,GæÓ–¤V­Ê«Wôí™ç%* Ä=B01èC÷¼l%¤;;Þwt é¸æ`ƒž_y`Ñû@Œ¬Y¿¹¬&aÏ=È52¼SPdá-Á{Û]Œ­gŠý·ÞŒƒûlæt{S„ò™-_Èpg^^_}'ù;v™‡^@r\®Éë·AÍëá`>ñ"vú›ÒNâÈUѹ2‡õcQ-œC…8+Twás uÄÜíÔ>‰•ƒ]äp /ᘰ‘–TÝÑ–ÐèKObƺúq ͤ÷ÝßšTd?‘WTÅ%‚¦Ü=o…†¨L¦ÁþÔÿØÁ½6KI•ÀF^¾¤ŽŸÆiw]R®Ò~MßÌN²Û›zcµq5ôT'HCŽ Ê•¼õqòLÞ2á; •úý@ïí loˆh=`Q–£Åw¡¸€ÎYPñTÓû;Ù’Á¦'Ùz5}ˆ¿, ›»ÍßÚà‚O:Å@,=†m¡°…€Ëy¿u³¬]ƒ6*29™Û™ì ×NÓ©¹»,ù‚¥ÉGu¬ùô4 Áž)È©aŸ­à5+’Ê«4nv°ÿ¥»ÉÈ=ñ~¦·S‚Ƈ˜ @-LÛëôåÂzogËK œl*_&@×-}%Ê´?‹8¬2\ð6ß“&+D™§3dùéP^V$£²à0àˆY n³ž$ë}üýÒ ˆ£ËöI2IvfȳE¸²qICð‰lŠ±¬›µ ¬Þ&Möòç‰@6ŸÆÛñÏ&B®x½ÞôæI‡p7Ðï?£Î)…ôsά˜,ÇÄMé âäÞú"…v9Ë¿Èdc^®_oTµ‡%|j4kq%‹{ )·×A¤”ŸBº°qž×;£`Z+ŠØæ…+«ìž™ õûXÑ…R&Å.OðrÒæVZÃððq+Bz0/Š6±šQ;Ç+á»ö²u€Ó1³à‹_j½awkÕjx‰[sæ‡rPÀ´=ŠE^©hΡZ;:Ie_o}WÒN›¹Ê¿0(CÁ›s8¨¥µÀÈû{ÎêXÿŒøõ(C‚XúLøWÑø:й†áÕBBz”á³^Á3ÂV}"›†h$%î£í)²stxë*sæößÓ¹ÕµçdïÈòô ¾ëÐÓ³A¦Ç”IÙú Ð . úhªÀܤÎí¸Ý赈f¶¿]-ŠOÉcm–¹Y¯Y…ÌJ}¹#·9hœÌ)ìízi±Í ÿŒ‡¦8Kö¬ô"3”i "=X )Çëº5V·»?]tê…ù¾ésõ¸8-¹' ŽÜ•‘[ÛE:}G²ª@žjZWðé«¥;lóåßïQOÀ+Ž-Gr6JEkEäOzŽ®‘€&ðÿ ]§1aç–kwîÖZ¶©¡¼ Zù£ÔpÄWÿá–qy}•S‹ÌÏlJxýõ#Òz{±\3ȼWƒ³®xYc÷°Š‹ûå"ß}6샖÷ð\þê üÐû³d¸ë¡¥Ò –OÎÑ^fXkƒR¢;'ãuò Æ©³%Ògœµ`ûa¡³¶:Ê3 þb†!y{ ÙÀ²¸ª/ÑŠae‘JëBŠüý2íJüVDj‘}nô¬j3z&Â-”ÈÃŽú)ÖY‰›ßXtà~[ç°§ ¹˜SÛ3ˆK&Þ (öa•“G¿ø9]çlô5Aõ€UÐ }jÅìé5ÿŒ5KŒÀ·?f]ß]ƒ^Èæb¹àa„ê‘ãÈrâÊ ëzÌŸ3€·´tÜ…M½(½^cÇ—šÿ˜ºŒŸ#‹Ù©Š”²É)ß½;ðÀñÁîu¡²ÞÁ¸Ô&Ù?8œ×›Ë ÜÅÊ'á{Á2­^xV§ŽÉqùò!*l,Ɖà=DܚŴ×= @…ÛO²Ë–‹ÚµRÅB| v]ï®7I óhŒˆ1÷¡ÇsÑŸ”š€&ÿô+GÏad=Ân©xˆF ûW ½R*.¡‚Òê›ÐJݧ&—cC&>¸ ß½=§€Ç’Æ÷‰Õ]$’l ´U­5δ Íž}ÃT¶êtcî6pVá,:Ž+ÂïRPô§¤Ü ßÇ?@ûÌk†½bÛu™Ô otÚã‹?ThK#dÍ™ÿS¡‹IåD:HÕIúµ¿$W³P}/‹¸Tiݤ"Aþé»Ü.ª“ÇžíÄ|ŽR7Éh‚N'FpN¿Ü¸þ£«%6ô5a²¦'Y&ôÁµ§Òç¿€t÷‡{‚Y.ÂF&oÏ è+52ÎÏ|ÌÀ;h+9Ÿö(ÄÁø©'ikw›UÕíëkÐ#Pï°~ö%'„1þÆÆÐN‹@‰œ]2¹4LíosÀ/«!m×âmô?öG¬ŒkÀó“Éwt’LÅìF³ láNTŸÅB˜ fí<œ¸·Ò 16I†?‘Š3ïAâqdÖYÂó0ˆŸètxŵ<ýpèÆUæóYOÁJsamIQv^ú~{³&¹D^Ÿìœ:\ûttO§QÉäzR!â9iŠ¹ "‚º…ÑI*DO—ßWΧ4B·0> ÷݃ƒ¬æBöGüã½ò2è‹=AÆ0yd_k½ãmy¦ìUÀq÷Èöžk7yŠE/€²(ñå5~çã0¢Ý/サÚÀ¾ŠA¦*uüD†ÖOEÑSN™’Z,%[Ö Ø¡ƒ/éCB q¾3¢ÿ&¤G˯­¯z~fª¾\á{JãD]ßX»23Õ®­Ø°=â(›ðsì{ëæMÏòFÌÎ3`é Žú¡ >}ƶ­{²ÆôåÕcKkŒ„¡ÒéÔB)ÄãÃ2ƒžâ-Ä<³ÓÒ M°Ž€;¥xQ-z=ÕÑèbUT«¶6ƒ¾ŠdGfadeõNÐZFîqv˜5þkЃo¹Á“2…;5ý/V¦æHSŠÝ=‹kÆÕâ– s:wN-äuwP¨E\ÒKˆ±b„Kt1F-XóW×O¢Àzñ|¡ªnÈáÕd ;Á²ôÈènBêïùÇ™áÕî1$dÇìý‘¥9%Ú;mL3áª;”ìåõLc¬Z_“B^BÛ'…ÇyÇy-”˜²‘ÝnÃó¶-è5…ÖêkMlB1KÕ&&l:r[óÆÏl=ÓøLàu ½¿íÆvÃjèF¦þ§fÝñ(ÝXÀ*Ò•=uïÐx‹r:H³†èì â–ö 7–(ccsßçñÍ -o@J ÈûÞ[ P‰uÃË#,éÜ”½%5+2l?†b0të°éïÊ%œ™û'Öˆ•>éÒt^ƒT)Aüjõ˪QŸÚkà½z‰¸6ý»3 烞TÒ»ãædcŒ«n€É ‰êo¯‘¡Ý´‹Ü0™šåìo]r…S ÿ”"1%ۜ﯒£¸*pHDÀ†w6yBr2Qµx}˜àd+¦‚µ@»Ïǹ¶$<ÜQNå²&ß@¥ö#aCo: Ú=ÆÆè|óï\M_>¹(¢²SДw`ù„±a-W&ÏÚŽ°îÀïj½×Há2¾æœ*¼!帗‚B¦¦ƒ\)‘îqm©ã5á!suG«ß5…ªàHö0ÿcN~×Qüº‹ôÙêý×(é8zĸsÿ$» ð&zÐv„‡aÉíÒÁøðô›=vˆ£%X,Óc<•™x¢>†€·’ :b¿,FP];;V.c§ F-ü‚h°¸ôk<¬žR±ÈÛxì;ÎÿÏñ3•–—jô°ý·bkïZyî(xÇCO‹“öBÏB4EÌð…šÁt åîpY½è’Tƒjç[š;‰Ô.ÍMò÷%@ºÿÅKŸ«‘ø#*.h‹×µ#VÇÏæzÓ(ËÜÞ’ R¥v0ÖÛÆév¿‚„Æ’œs!Éb¤$x'H+"Õfm^ú­ú¯¼íwÉk‰Ž…<.hÃf'ÌIÖ+ØÞýããG‘¬Ç¼´Ü¬öhH[ëÝ7(Á\6™c·’nšq#n²Û­·CÜT©Å=˜ÏzŒ¿IïJ”‰< Š=ïŠþ†|yT™±õÔ H ¼8.<Žd§çŒ"*„çp`3Õu¯—f2™ü´ ççÙ±ü›ôa 2S ¢œ½³~Õ®`’œól²oí`Èà;bVDäþ•ûF®ûÛµžU‘;´‡ÞÉ•þ#»Õ©”5h) ]ú5$_ ¾6$œ8¾$‰œ [ ²?`­ÓÂBpzU…ÇÅ|£YÖëÁˆÛ½<DB@dñÒkX΃A ''~b·žŒr6i!Ó`»¿×(£²çVW ÏúUoj‘Ã\ŸÕ­~dõ¼(ÜLò÷ˆíæÎœ ^Ì!åúó_ïEVÅAE¹>ž.j)UøàÞ¹,ñâ±|V N7tŠ}HÔ4VÁ$• ‡®hžÞS0#i¤™1¹4µj*VdW‘×}ÜKÒ' ·yuj@¡vÚÓÓ./ð›Éü´‹@ò7G³jåL¨Åsž#%‰–F(ý¦D±$ûbÌa5¾ÿ?™žOd´¡¬ìªC&z˨×*›Zƒ(*Îå"´_u θ29/ú0Ú«;ÿâÂQ+¦»)Z#DÖá¼$š˜Ý*y‹óæšØ[ØŸ|ë\c*«Sozqï¦òØ=M]–%þ®CñPD¡?ÅðÊ<¹F@ãbu¿¸_§àøÁ©}ÝVVÀz$4J^©Ù ~;a+l*Ú¾•¦¤Ηav¨7¹@}´XkF¶ÜL®2pì\`jOºJô®X„ÙOè(Úýã‰êI¾R_‰Ë²VY³ŠëƒaêNÂwˆ–n”c–J•ULl¸%>íûé•P¡¤gçñ%]Ĩ0µ}òn;4O÷öäºçôÕ K—ªþ;u‡Û»ª~éCìùd&ö·™–“­é.sÕe«êTfìO0äOáý®Ç`|³¡•mÔ+»„êz:Nþ`®Qó b$úž¶Á¸Éä^ Úù” {;ôàVE4Ç8ðýqoL.¨E¶M7aE˜u_x/Oãœf#±BµvB!ùœ°œô{‚JC•¨a†) —säá`‡žOG,-fÖ€n¢3"@sªAM¨ˆa¡S†Æ¼ë¿Š<µºóuŒfU6$/Õ ßÿÂ$×°jã­+êÍ\%ð¨ÒÜáÆ ëðš|§ìÒ†$Z.•UÀCêj€ðÙiju½Yd°¯©wNæuã•äq‘ù) ÔèQMK'¯wÞÔJ•rmvj]_'[ŸeÏÁŠñ/ÛW1'ò9°5úö2ØÞEM¬‰tCùšFUÒq¾ê{˜/]GÑh¸v'R’ì–(’ÿƶ•ËÁŸÂø“š±~PI+;2·Â«[ÓlqdžnfiÉ®“nÇ*G\ 4p5qôt¡) …ÙZ›YÄFIǼø¿žå|?Œ¥lD7ühŸl­|Ðî-B­}ÿØqšÖœŽó³÷a#_(ºq ™²¬<tëüΗ1RB»ö=D•-HòTo~ÕÆc×V^$K]¿æ¡ñó§WÁ…øCgìœu/¦ãj‚ÎÆ&^•f+§sëÒ®ÚËè ñL™c‚i›UIüÈ«[OÑ¢êN!—¶nµQ,ƒ8ýíx8(H¶¦¯lQNîIª91‹Ï†êµˆüÀï0T±zžÖæ˜Ët±MhŠdt«P?òýøkÛê½Ö'ä $tøÌ ­é·;039ml_Q$ ÖfÈ<ºÐ›æ_Fºv¸V·ëeéù<;‚!_fN]µ9bå½`PA³ÛC å=!–žL4†ÔïôiµØWÈŽÏæ•ý>ô7uݸ&s^ª³WG­ãÙõvðÕ8t ~.®n:“ŒÙiS’Ž@¤ÛwvU[YÝäGQ€o¨Ñ>72F¤åÒ­Z3Û: €Ù©=42Ü+3-Â8ûªj‘bØÏ[UÅ<ø´dË8YðÉVB¼ä€Ž™¨¢ƒ_;÷œUU`Ôkž›„88ߺ¶Â)áŸñÑ7ëèùã”syh‘J•Cå=F!دÚÊxË¿#K¡Ïî % ÂÏs lò1ÄŒU¥€QóvºßF4•Nkg?¡¸€M½ièj®Ú¡ð¶ãwüÛzF­ø†¦ù ôÂàãTþÄ:œÃ…H)YM­«´Üg˜qÖ¿í>PVÀé©>œe衤Ž¢]öe[:XMR±cÔAaNqwÿ ¼È€‡Ê/wtõ¼Á¤öa¯çq¤zK¿Ë¶Ô;TÎwž9–%¸´mF­à™ˆ#cÏœƒØ‡õ@=®>Ž>ú›Ã>v òò\¼A°ºS?9ëõ‰^{œÙ($cÔJ¦kå~¹{…‘+LÒZŸ#ÞY‘áa§&ÞÌá0=¥cNüˆHv'-5-Â…«Rj;׬ÎñÃÒWârÛîÌ*ÝÆ –-@Ò¤ 'j.½ë oÎ{ª#¶DL µ1÷ïþ#•ƒ?¾ØL8¼vÇkX"àfLâÜŸ®ÐÈ‚ÿ6þ^!.3ˆÇ®ië«l½®IÁE"Á(] ¹ èÎfÎ;ó“ñ»ŽÀÏõHñ!2…Yž:}ºÀý;µ4†B‘í —ewßÛiÌqñ§Ä”h%qáñØÌ«0‹ïÁ²OM>ˆ€6טƒyæ/lÖ Lú|Òá¼d«<ëgN}|‘rßG91 ^°Ÿ'ä0ùzŠ×ó'µ¸3Óg%3¢§bMo:C²³šn4%-saíÀ ÙÛ”\/'c ÓŸ­ÓÖ±ß3ðþS‘¾#«C_f½}&|]š{Ö²0²—±¤©£ÅFöHå"^s›œ» éêüÔ 9ÄzHÃQr2óß•ÀeRéª+wmeÊRCcRœð´¦û…÷¥ß7ø½©€É$èÀ³€¬@ Ê5°Ð»¹ØK±°&¾ àÓ™AåYÓÐŒ¼ßë))º§í¢_ ÷œ– R-¢ê…hÿAR«±:Fi{b» &hòd£åÁ¢le\qú4 “Œ5·)F¬ÆXH&ã¯G‡°Ÿ~úÐZ£ §ðë|äŒø²U1à†ØqðJ2`gêh †Â^M,;¶óWF„úòL1C¨$,¬vÆ ‰‡Ó¯_}O}îr ´~ßsaðŠ„éñ…‚:~Hžª‘Lô`Y´VI£áÝÝãÓu JÛõ?NTÄìªÌ!²Fz–¯X0­ž=ôy½§í…6·WØj<ž›™bóÌ5`±«N^z:ûn±5 GEªé–{Tgæ³ðó;µH—lž~ºtr<z££¢¿ /g­=[®” ÝLÉ©ï9‹iG3MÙ‹ô;³D•»©QEОúîPdI{߬?,\e«Kí$²w¬—æE>¿$ÿÐbÜá¨ìÝÃ'Rúc4‹dâ=U-îÓ­8ôy£ç/&µ™O) iAޖ«զ/•v^‚5'm²¥×pY„œC¾=öæÇ›4Ë–'A<„ïù¬Ç,1…+^ªXc &½è{ýÜŸr&Ü)º“ÁŽªíKÁM_ßÎ'–V8—•‹Rš_šò…ºW‰N¯mê¸Èk` õPÁ\Œ‚Ìz×X/ç-¼ØLsÍór*ÿ‡~úŠ>Ò|#Ö½ôÝõpÔ¹s2‰,Ž"íF•Îe¼ög]Ï ùÈæHMÉG`Ç|Qá>½yA䔟ҚÝÒ·õ´'C¨´‹×@!{ÁàËÝ_\;KR7á‰G*uÜHAw4ØÌÀ r*¤=v¿÷0ìS1 Ó“©äSø'”*ÏâŸZÃ;Þ9Ç‘°è[œ²®=ªcCƒ+¡¹ÀüžÒøpžã<Î O·k@žÕŸ–I™DhŸWså“2^›PŽÄ}$Þ,|¼¤÷|žq†_÷Ú÷í¹%§×i‘NIZ9xÇÌMyfeçDò9,lž¿÷G‰ný¡'¬¥_Mö`Lóý0^úÞH"Cšë ¹K«â·ºÏx 3Î攞íy8³ÝSþ¨ÈLœk^Wå h(¿[ÁUf„¢ïÛ¿ŒÛxªß}K#NLu®rÌÆcEX¸àÆ•Æ:íEn$;ìÜŸ, æu É,t>ù8È¥M-<>ïH ,žÕ€§bâÝJ°Õç«»•}oªxlµ¢-$šhYõ”ñ ?Á«ÊQìÔ.d[„º…i’Å¢VÏ`E.ô ]ϯ@ÿ˜‘>z8dz;lä¶^§RtoíNFÒ‡ÃréZ(ó§îÉÝlTí“;½<}udQ pôフm“­9ç¡‚`9L•÷#ú‹EêHF¾Àdnö„z˜ýË_¿´„±a]Ë‹Ê9šåáD°ä¦D…¹Åÿ»Ÿ óaµHY€ t©Gn±,™¼37 @á&=ÕFñÝ{Ô§í—2‚}pßtµÝlº@äÈk-r]¤¯ñ<¹á8ãô¯¦¥‘KåèLP½ÿ ŠU)›¬²iA´ƒdv˜7Y=Ñ:'¢“ð/™£ o·ëê R98æDC ‡´®BUL.R^™·XVc¿ß™7:v³)s±®ùKáâ-›hgvÉ:xê”_-²ïn±/óë5óÜ|å̤æß1X{v‚WÙ¯/ÌOQŠ±jš«ãtîû¹$Tcò§¶ª Gý*‹µ‰Ö×¥¼ͦ?xŒyÑZîMçBþÊ{ÃÐÖ™I-$ž´ÜYi“Và ÏÈ:"[E‚EIÍœ‰ŸÎÙ»ÜÆ\‚–«õA¿! nV0»ê¬û`Óû1A'FˆW”Vå_|J%s».–žEk“F °˜˜ùaíüÑ ,ñôP"¶‹ï’¯j¯Ù¯£ì©•áÔðc‰!¹é–®)kܹ#L>§Vd0àì¹Z×´ÝÏjå…]%H4^‡“¸‡è­,öSb´»2Cç:`ò”šQ{{ëuj¥ùvtIäÎMw‡³Ÿ€œé“ô ÃcdžöîÊqÜ[p“ ^]¸lC-ÞG„[â±nH­¥”þÃ]Î+cßjLzêë…íe¢šê …TëÙ¹ª «$è•ÕÞ±ö ¬ pÒ”1¡¿Ñ6XjQïïÝÚ‚\*Ú"†(ÃV1P°ÕÐÀ¹"Ô0¨":•éÞü^“¬ ŽQ2Âr’hI’ùŽ÷î5ŽÎ4 ĸ‘Z ü¶G¸‚š<¹¯;ft+¸I"8¬ÞS‹´÷”:Y϶P½2E@òÊst-óŸ}™-T±'iN ;ëûŸQÌ^ùظ‡Ð.íÅ{ÖZÿ4ņ IÛc¹û'úÓ?#µRJówš{O.°>v©„ÏrTe°d hЕSÍl´Öëi·\j4ZÝài*¸÷‰µÎû×ÚÂù:ÖúTRAí™üÂaU³ÖÞáCM) ö(Ð5šß~ºv¬,¢ä`$ôCŽíymŠˆ ˆmÕD„`/Kšâ ø¹¯<UÕ œäí’"&a&Õ6Æ)ý£=z ø.‘ÇÊ©ºÒ> D íØœ¶…kiøýe ÖýâL¼¢x},õe!‰¥_ßqRïõs3 ~tŸbÂS¼‘…¼vÇY{Dq$é¯Yš­†+pŽO´çÇòdïýÅCŠ ¦G5ÂdŠ A?NwE”yž2FO±9]øY–tÏ£=u^8P½}D¸‹?؆Íxpn„n‰‰xïEÓ&Xj5®nÛOË”Lµ¬uO÷Èo¿O` k@5Ù…ÿŠþ ÂϪqƒüT>“4„þ \L‡{;›‘öÀß5Οò²s-F2sÛi¼j¥3ÿ ý8AÆoô?·_šš>éƒà @Äì\hÖŠœµe§,©›S¼@´J}‰Jgªô* õï”øïï€YÈI{ßÏ,ÿ9ú+8œœ(H£"cõÕåÎ…8Ö ó܃q^ÅFQäz\ ÆêSÞçí"Àe“ZæVi ¥îÕrþ†ùÇ$­½ Ž&%y9d'òv`;šÂ车H5Žr¬{ä¥æ˜Ú •—½Sg'îg5½u×/èï¦@ڲЦ•(»i´*Sɺdñ8\ý ï éË5‘z*€Øv.FÕÞ¾\n5jDXcéÿóAD9þseË®â0Õ?K«ž1Ðçšä{S?Ø „ÝuSžfòô¢ÿýr(ÀÆ‹PpV¤úrÉ []¿³8¬ÚñOE0if’·‚ˆ… K$Ú´gÊOÏéðm‡ojÔÏ3ñõR7-‹H *¨GYEèI¶À€5’À8ÀkX)¤ðm+gh’¯~-¹ñˆÞÖbÅ)5‹›7±ýfû TÞ.Ux? J 8idØv}·‡Ém欑°$¦è+Xµ"oU2æëáµ»8mŠFI7–¡\0œÁ‡Ÿ êçv#hY“+l€£øÙÁei¢èÈ;v«›oã'²aÖ˜ö‡ëîDCi³Å‡HQNHoÅ!‘Î>PÂI~Ù× UQÀn ~€ Máyƒ—ëørzÁêãçJØÆžÿ‹öú°­hjm?p…ûK³ô:VÿÀ¥p*ÀýD&ƒa4+Œá~’fâ ¤¾¾åVÚ+òrá%Ái¡%Ûâü ÿCÇPn*x)£C‘iÑxþˆ+dD§¾âwè‡m…Ìæn¤žï‡Ð]Zvq‚Ⱦݛ^<ͼõ¾¶XáY›Eè½÷î"¥wëÔСinU”ó£ƒvo–úPMšó ¥:­Ý2=®¬¨0ÿó S~žýãKË]dàVéÈ{55N…Ï+~¦a(f^÷s3a_¼%ÝŽ'’6 0}³±Ä8=g- ÏIöÁI¡'ø~»i•ü7Y°í^…•I Î “ÊUïJ4æR„Cþ>@Ýt66ÓZwJðƒ‹#Œ]ÃÊ#?+{KÊ*Ï) â¾Øs^ræ hø ÛßÔ¤­÷ÏØ£$·ä"þ†F±Ú¨ó¡À¢²äþ–óœ[áúãxµø\‚ã+»îô€ AòcÌ Óž(›rn™äuna·“ø÷¹ ï E‰g€7ÁÏì˜{0´Y­8P£ÎI¼Î4La3Žâm¦2˜‡|k}_¬ð~'Ä5Ï°KínÔÈ®½ZÖ­?.sŠ´<ÖK(ì‚úÀp!ÀN¿~nW“_÷ ó€.€Ø»“ÿ“jVŸ-á;¦3¡Ù×®€xÂWª/gj)ßY_R~ ‚_qGsW ™Ù©´;Ò<Áϧ9*}ˆˆ¶ì™ìßp1»SìÒUXÉ$ÕÇÏ«b3×èŸÛ^L@'^" Hߧö//m_೚˜‘wSâ«ýQC§ Í‰Ø«ø—™ê"q"7>©'šl/A<±a£—®±ÐøÐ~¶W8z{¾$}= ÄfÄ҇¿+ ¿Êj"ª/© .œBø? ûõ³-ËQy™F ´K­H$Ä4¥•G_3L‹÷H÷°ä¦bI½ Ž¶ËdgÉtzûR&¿$u‘å$5Â4é;„_F¯ ¦XTá–Qál¤!võ\VüG86EìBŽÕqé8Ò‘)i¾ŒhZÁÙÿ•º}†ãoÅYóÓIYU*Ç sOx;¶†ò庂|ûçNrƒß”Rø±!<·O ̲L*ˆ¾Û;o*¼™&”MÄ>’îš°îHz§xwl¼qDÒÊ…bü™±î*t.Ì1ÉìŠÌÞ>•ë²¦Ü*–_ ù¹wƒsopf÷B[fãó…èD—²\Sbt%Òëþ Ñ^ªÛ]š°L“lì[V ýÃF[3bn¾Œ.!û<²9ˆð¿Ä ¯dZPóA•[XÉ'kÐ"ã+×’ÒN¬ t„±®‘4ÚèI§7ÒÔ@ÚwˆI (ýĬÓdIt!J.™©@x¡J¶’ >1>©»ILê%ZôæÎG ¼ÉâÏ›Xûìz(Ç6hÂ$ZC¿9¼™ê›úÃÔî…¯MÀö«‡’= ^js< .¹e#;PXI€U‹}Q èʇmJ`›ƒ+ •É{Uçšç?‹¸4E•î(·Q1ØO®F©·n_SXfêÌ_ W-(ΩNr:˜$[wo¸ÌÙR?Ô~¨zæ’÷›*»À¾­½ã9Âa$J4)óÉö~`6à_/nIªõ°5çVz:ɇ¤$|—*¼Š?±×ìYÏ}š¬øñÌÙ9æµâj[Æü³/&{h§pDœò×Âf15f«t :q÷‰ƒÛU§ ´ó"²hÂœ.ö…A¦ã£Ã]G•‰ ÚNÕ‹‘§:–-Ç-åâzÚGžW?oÂÏa˜ …ß~©öÍHÑV&†otƒRrmÍaOvÄX»>—Ú’,Õì0^0œÊÅ–Å:øñ­9G;ë~Ìp™Z%£ÈL‘›Û­‰²™¦ú)?\Æ4‘º¨“¢‚.†åf\+r=»ŸÙWµ eo-ä:A>IÑM»ÊZ-„2XÅïÔ•¢®OÚçpkNG¤*݉zNdäQÖ‹7ÕÇë¶}À&^´f©Þ w/߯PO=u¢sð%ä'OÀc¨‹ê…4åÄQöó¢_ˆó˜ͳŸ}Œ¤ˆ¨PYÕãÀ"ÁŸø,mY}=HšQêOµâ5(#hp1)ÎóŒÎâÆý)dY7`€?RàÅç5?kÓK"iŽ–/COš£¹\÷ò­ ±÷9ž’Ü ëΆšW×Ò,‰Í1³0úÜ, Ì”_æ7yP=júKß;͇1{§fúÕª=œÂz¶Äâ kç8ú„j+“Ћ}8Q -èèFÒ;—ûì2ƒŒúV$/¥½ÁA­NS Kjl·™®ø‡ÇoÀ«|Ì­j$Š5Ï´ÛmúHûÛg©ÒÄŒ°Ìé/ê9Ñ)–ic]läÛ Žâ 3Ç ‹’=Ûš9‹¹Í²êÆζB+PDlÜ‚?™hçU€Û‚è œµv(•m‡¥k»„{­åiŒ| WHôn,(X¬XÿòìˆÚÙêðQJwO–ánùÁôw¥r“ì|EW› %I:<\âw³áFã*8Iß)µÙ8—"pR08a`3¯'qK]\ãßÆ“ÎE1¹< ðƒ6»°‚݇„*('50ÖÌL”ûÃÆc=q,SéÚɶ*ÔxLô¬¥ÇíyO³ùBZjÅO?)¬0tª+­`Œî¸Õ§ê8Í6Ãý*_ºïŸÉÅ O]d„¹ÊÎ=dG$ŒöÏi¯#q 6åi¬ èÌ!ðÖÑ•ˆìº Cx7ƒ &€X ’x?Æù€Æ¹¤­OÏx:ì`Y<šf\¦‰$©ÇG å¢p‰øM‰ò9W´QC Péd:D$§õÆ¢„N—Êk‹¥=KI—q|ÃZ>–z#,·w½†™7$+7_^¤Òź/îÿKï#s9ƒ(yÁ žlEe¤óžÜÙ¨§Üj!5^¹+D½ó¥u$E¡£lfÁHNå*²¡%ôÜL§1I4H¨<Ò#á€é&kÓU(_ûÖé¢A7óbòÂM^û.èÊ‹;Y Y@À îðyPa§;Ër ;Ú(ôÄ¥JVΆ çØ<rItð:àö ¤>˜ð }bÐ’X쎀͟`z¼£‰kÌX³¦âVrÙ‹¯¾0æÖöëo´ÝŽ±+fŠM«º·ù³¨k¹¯É‰8-÷ $bï‘¢È4†¡Kù Á¼êT£ß ˜4eý……Ë3J kûß ‡œKLBP)/ª®hóóoža¡ýŸÎR•o_fj½ÎÆ͆ÆÓ‹1%Z`]nóþˆ´(‚u±EÏÒ™^+Á%tƵމ’ 0ˆÖ#—ä¨F€`ò±õ³•4k½w¨ ˆÄ8þ,!e\ EXùd`ÿˆ_ªË¹á:UÉ‘V=+|êæ$Pá{ß—ª%¤d—ô’Xº$$<žpƒ-°ä ­É endstream endobj 1187 0 obj << /Type /FontDescriptor /FontName /WIGSRM+SFTT1095 /Flags 4 /FontBBox [-204 -361 1351 830] /Ascent 0 /CapHeight 0 /Descent 0 /ItalicAngle 0 /StemV 50 /XHeight 430 /CharSet (/A/B/C/D/E/G/I/L/M/N/O/P/Q/R/S/T/U/V/Z/a/at/b/bar/braceleft/braceright/bracketleft/bracketright/c/colon/comma/d/e/equal/f/g/greater/h/i/k/l/less/m/n/o/one/p/parenleft/parenright/period/quoteright/r/s/slash/t/three/two/twosuperior/u/v/w/x/y/z/zero) /FontFile 1186 0 R >> endobj 1188 0 obj << /Length1 725 /Length2 15995 /Length3 0 /Length 16538 /Filter /FlateDecode >> stream xÚm¹stfßÒ-§“tl?aǶmÛ~bÛ6:¶Õ±m۶͎û{Ϲï¹ã|ãûŸªYµöš5WÚ{MF$fgë¬âad¤e¤cà(‹©¨0²°0è˜aÉÈ„†Îv¶"†Î@.€:Р ´02˜aÉÂvöŽfæÎ cÊ5kC  G€šµ«…±9€ÇÕÕUÀÕÉ…ÎÑ…îŸEÊ@ ÀÙ0µ°„å4%åÄârªq -ÐÑРàbdma ±0Ú:)¦vŽë;c;[‹ÿáäD÷/¶®@G爙:ÚÙdEUÅäåT¢Âô*ÂC[€ŒÄ?Ëm¸þÉ;;ÿOu4çÿXÖ†ÿ±Œþcý'êèþo‹–‘`baì 0šYØÂÒÿn’¶¦v¶Ã&.öÿú‡Ó?üÿhH 0šþ“íbm-ghPÛÙØ»8²v&@G[Àÿhïæhñò¿©†6Öÿ¿Éÿ7CÒÙð1mÍþ‘áß…“˜…;ÐDÁÂù½]€ÿ†Õÿ>Y ‰…‹Í¿—ÿÑäŸS5±³µöø¿;þSÊ¿ö£WQ”U£þOü+,jklgbakPvþGOCG“ÿÿ +Zü»ƒþ—Ëÿv€ñÿù²†ÎŽîmºÿi†ÿ¹þ×ÒýYBBvî^´ŒœZf6#3€ƒ™Áç¿è»8:mÿ%ú?µü¯ÿ¯fÝÆ°kËvÆÜÁ–©Í¡e¾¢…³åÔ³ÓªÔËŒ +Sݵ‘-ÍMô¯pîTpòÜ~âÏñî^îHÞ5ÆOLF…ë.㼓”›ùVo½)æ¿Ö…ë]C¨aÛËAü¸uN__ T šdÉÌ9¯–r(áEgŒ6(mïƒDÚŽ[]x°§Ôý…’¡õõ¸O]…4kRpmÄÿ²·–´ûó/¶¾yAY€ýa˜VÞþƒa;òI%ú sðKœK½69d ,ÍeYI“ûAE¢öI[×&¢’ RÖÕð˜ú¥š«5âÜF™½™>7¬Ùµ&@,­Â>¿0Õð¬Îf5³ö¾~sÇ7¥oL'QªMÌü×üa%= nô`^d+•.Y+ðŒÌ’ƒ-sîí“ëÕhÎ-"·Ê¬Å¿~ÖÆÆLxº²­Ä“DXåÄÕ^93çÐœ;Šk)³Ï+x ’\oºŽÈUT4©ÏK&2ódfj|¯.¢¥…>äîÉF§ÃÙ["fòàÅÝ.VªµËجã{ù‚åbÒtòÈôÒÌàù˪ÒÄN$:äæã_zûc/¶¸e†Ÿ´uoûFDù¬`Ï«EŽV‹¨¹Ò!e3€ºØUy^ÆÚxª²ü5^ò2Œ…±äÝÍYÆÕö›LÁuDì©SÙ[vý^' ³l¢Ž—"žÇÑ<ÌÐi0&N(†¥8óœd"0Qñ‡À•­ÇrÎêzÝFçd|2u6L â—@AÐåŠUŸPŽòàÍÕaÿ¤é …—Nab¸¾×Ÿ>_e_ô¦rÑ2"ÂXÈßQ-ÌÀÓnõÃvåG.6KŒÛB„ü8 #¹î¤ãÔ{«ä ƒ)²E»ïËòÆoï)í›ÜõYÃúâ"9Žº¾¯Ö&†C…[3qVL’*¸,lÅΰLJ4.ô…h¨åOë’:Hîe‹Å;»Ÿá7Òc,â¶:ǺÈÿ0Lbm¹iNºˆöe% q¬õÒ³ùÓWíQ€YˆþæåaT¾ìÿ+ýƒf >ÀÐÄò¶ñTsUì0¡fÌ~ÿz}w‰/åEæ*ƒVÖ‘žDVÔÒ´ävE/»]…,ôÜÔ áç3;íéøË/,«S€ÿH˜é‘åá€ñ€àíLú}˜Ó£ý;(4ŒbµÃêÖ“0JJY^®‚AýOÞxZц€%:´R¶e7ÈõÖ¢Ý_IÔPH¹§ t’ß?ߦ2.žz¼ãsb3ÁE3§FàBê ÂØ·„Š“jÌ#.Ò\¹+±U@D%Tøå>Í ¨¥vl6-"ÉåPyé†Sl·€ÙÃø¦“¾J6{þ&*8’¹`û-/?7³A_ž:»~ ã“óÏÆÏ>TÜJb+ó!‘ªÓ.!*¨„Åó÷ÜÜÚ_·Òðó»8SA-ÛW2nÄÿ)^uë¯×©Â…èäû"åýåšà®ë§@ƒŽèï½ñŠãl¸iwWè,yÐ ' s, ¥R¾¦½ˆ\ú+•ã±µ“ÐùÙÛ'“yb¢ª÷]Ü×ä¶*wƒ±qQD÷íñÙ)§ã‰š‚m/Iˆ¸oÅf{o0 G¶DH¸TΘßCòh½‡þdö× í«Ñ….{¤T8* s­or©•dÊR@¦V›:„Àö¯ãQR["|Ypý`kÓÀïIÜ’,Ò#;@Õ`˳)‰H<Î'©3 Å—èVN[FtQî>9ÿU$ºì“²„œo€”¯œd‘ ]­îßt®dm2*fÏGgõ4óW·âužŸÞœÊ5á¹dÖizÀÕNËA ¶É qds±XÍS…*tžƒ3QfrǨfíiòÝÖ{š´¹A<ÉæN¦óŒ@ Çèq¥3ùÛ$\,èZš0ªbÎ@ùœX`“Y^Çè@õ&usóRí(õzíBOä3¢å«O.$§6´õaĘm!’—u¦ÌBö^T¸(ïO¹ì\š'Û6g[:ª“A¤Ähí¡Íü>hÿ°gÐë¶y~ÅÈÓg=¢Ü,k H îDIH(qJãf$õñ÷f‚ yüuïØ›tÙ¬ÓDæ1Ѳ_}–vkzZµeº¿ R0HÔ%h½L?>IÓfoÌä! î°_\vÞ˜NØ%•m¿NùÏ€þçj¨ÌìÄDíMݲ°P¼6×P-Ô9_~ñ=\qÁ"^l.¶Ê§®³OLÆòX&!Ðõ?N|ÕÑÕÁ «ª/¢Ñr~`:;áDC4±n6‘B| û‘ kç÷X––ߨfyÇ« ×L¹*ƒ«–¥xò Âõ S—hÈÀä…‚Êsõ'ùÉ­Œ´å~KFÛdŒÈh0bÌgVyäqéà×Î*ºlÕÎØÕ‹DCµé¡ETZðø<í…í‘0ªö½¾‚§S„|_ç Êž=í5ÎÉNY®ÁËæY]¹sÞ9¶"j¤ûÏSr“&ÛW=8>GÔk"ÜCœÒú—QsIÃìΦëÓÒyx“µªËþráü¹ *´£ µ2M5õ5мÑ÷…j~rcê,Ÿ7ÐbÒšÜZ t¿#¦q/BŒ‰c —…<,}˦ÛþLÎÍYù$itnÞA(®ÑÐ2+ÑÉå$¯ä)!ÎCcÇu¼ÖÉ Ëo úŒ®TQnþ’Ï!loêò5ïcK¼”ÿ¥ fž¼ #}¡œd?b¬_p’Ô)Hî¶`®Á TæÞö¸…H ñ»êƒ¯žq72•-ût®/5G:b’:L‚ ™ŸhŒTs&ç,÷½áx¼@á†xtûÍî§R^n}*Ûêr…’¶‘)»tËIEA¨QC±ˆ3„0T¶Öò:Ï}pŠ.áR®¡6ào6Ö{åF~>’µ ÌLç"Ú“ŸF½îÜTò ª,õ;ÉÚý€'âJÓ©1"ƒé3üı…¯[ýUf˜ÇPΫ°â¢`+É6f»VGå_·é^Q¾-ƧoQJG¢g ¢³dŸÑª*ƒ©¥Q_kúŠ1Ϭ˜Ýœ¹¥(žÞµi²vÐfí¹Ç÷ñ§AÀ,7‰¹4Öé¤Kn9 ¥vN’ÔaYVZKU¿‚áoêDßÛ_kh¯±Ã]ýÄ·ñ“}°Äæ## ;×[cqÝÜ+¢Ž‹!,lH¼+²)Dƒœú%Í÷¶ç¸™íe´SŽp“ô&»Ÿ¯¹ù1 p¡¢ T‡]e€ÅÈnsðu~º;€ôJþ1&×,¹S§TûM²?É's²…Y'-˜MÅ©‰KL=(t.+³óO?4›zZ(5_6˜šéo=Hò 1+‡ì“ì²æ?ð§ìð(D_Â~v(ýP×O’´÷™xoòˆ;ÝÛºlŸm¤ ¶N~ôjpc·Ize¡Õ»…þõ<˜E™Ì(m#ÌcàÒþXìØÑ8Ù•f·ÐÍÓå>¦IÛ©•0×Õ©#Mnä©5šr4’ëþæ9=W[¡ÖmV²õ§ƒç"³ˆq6_tÇžžx÷uWa³BkŽßì_¬Œ²äº0kýfWì¨N€…ò‚|/:9óÇ´ Ý0vZÞÙ3_8Yÿø™Â#$åd’ü¹îäu†¼=Z[þ¥bçM^C"i¨µˆdë-€Œ1ÓM~†Ë3×Z`Ó0²,÷*Ì7ÍÑæŽûÖœ7w&]Þ܉YpV©rÛö³²‚¹åå4ÔÔ.Þב ‡úX-Y_3èÐLR™ŒŽ„ý/œÅ Öý|˜×Beñ½)àV/b(QDæ+l’dv2oh, %ËF0ˆ‘Þ ÓÐl|XÑ/’{[ý¡äÃ)hÏÍz®Æ&>‰Ø­©ØI½>{xÕ™¸¸øãä%BÜÕ´'×½Z×èC"ëšîC NÈ#èÃÚã—%òÝ Ô–=‹¸ä'¸¥¥‹¾»&«‹”Ÿ¢Q¹,OUS@º¹ÀRfxÚàó¬™M[e&ÌþÖóEξ©!°õºÓ@—&µÄöw•¨íêbª|O¼õ³xák ¦@¹Ô*¡ž'8Îk"·—îQ@i¤'QÔûdK%1,ןÕ*çâLDhóîâvgd h„º¹Úcî+ö6Ü…¦&>£Gsmð?…¨Ê\zø¹Â3Þíâq´CŸ¬N(4Ä)Qã^Òðå­ w”](XÆ™’š6 ªÑHµÚ ŒõÙCb—¢¯z&DqsÞª#kˆÐ­÷ŸGùÁ±½S\b%­¦B"íaëÚ ¹NøèUZ·„âe£ª2'Kõ岪Qˆb=ŽT ±Ô«¯Kã› /3]~1ì:n¶t.„ô£eúî™HHîã:f)ï%}£Äü9,?M“é!— ‰ ¯Ûû-çÈ>¸}ã*6MOîBŠ{(OÿÂï7ÔZ%¨9П1ÐË[¡y¹÷ÍœüÅåI{r,XT=&™5ÍM„ªvJ~¤#îÆ‚¶8@3n¬0:ˆF…4é¡ÍÇâhu„ dŒbB§ƒÅ}ëŸÉSk¤ aa}H1[&‚‹ÊmcN_`*Yu¥tŸ5?h×ac5E à¨ÛšÇØ,L¹ _¨ÙGà´zdù>u5jæÖñˆ´~ª¶ÿÐ. âǸA§ö˜!UÞ䔵 ò.)í³Dq馳QÉÇJX¸~­ÉdzJòŠˆqMݸ|Æ1¤?·+Xtc÷‹œ¿O·T÷HèõÜÝ›ÜL“ˆ ifʤÄÅnhŽûI…ó…t—hÙTLb[îô!B˜TÜAp¾f`!PIW5Pµð÷ç§/dÑlˆVã4Z£±hty¶ì’Á)&‡äÕužýRìªO>´“áÍL”ôV¶«Ú·”¡JMüA,ð@Ö$ÿáJ±kà‘.Ù?N9ݤvþ,›ýª*O…¤ñSîªÃŸ¦;UPuï.%?³É•«E’5Ôèyhò!3brç«5!ÉñØüÆ-g žSë“8Ò)KÛOØ#ÖšzúK#?Q¿á¨ßBsº¾:uu0ó¢Ú-µÎ­ØNqӳ΄`¤ctÉ»‘å ^ÓFüZ4Tt14Pò.HVU=±âéÑêKµ0€u&"2É+©–ûuzYƒ»#9Pr”ôçh‹æC€ü —/q;ea ÚÃn­^jãÛ+àyþ²ŸKÓ Îl…VáûV5GÜø {Ea-’„HU‘Ë ÌîŠQû"‹®Ü ¼OÿL”!fÑß®J‘øb˜ï\78X±@t¾}ºÐ±(:ðOÀubsrÊoF¤ ™@ëÁ>¬‹2×Va8)•Ì䢖Ý M nS÷ SÞ„m.¥²å¹œàöì} rÒÑÔo„,Š5Q˜ã¹£É~:°…eÚ«n zuø°ƒ€uáð2È|ŸÇ’n`’Ž``†F¿lÍú"e†•xÃa€Øs°JfŒÅõ(àœ6)€¥Ø8¦ 4.i‡«,U™Þ0ñ­ÆÛ«€/Y“¦þ~ßѸZf}ÝÎU¤X4‘ÊäúY¢£Êj[sÅau¥ãøt% „a3v96OÝ’ûYIyáûwfÅW.|gÆZ^Ù#ÏÁò2]tŒÌQ`q6+w9Šå·€¼oˆ%çíÕà×$)t<Äz¨ÿl;”b=¡)#ò,½ qG„§ò½.HsÝa³ÔÅɘ‹((ŽuzÔwa¼¨3CVŽÂ æ 0}ˤZ)ÿ½,ø´B] ’L‚4¿½õRw ûY£‘\ôÙfZ×"Ïø†¸ƒ(0•½†–“÷Ï2¡ÇTÂ+x{¯?SáLv~®Á^X+^ƒQ«W«o²’ K|ÁØÃB…@2üéÚNO”Ô;‹ P½Ëºá¯á7·bä$YˆÉ!®ë¾²Þ÷¾pÞtnÍÞ×÷q€:øà²xí÷Ue¯5±øûiùö±t½U7v€Ü³@«Ý!G·.k6̳8œG.B“¸ù~§À ø1º‘@iaÞG?_Ù…ÊY#ðoM+æ¸ ÕRµ£¥ÿ,Çw¶áÙ;‚õ…IHÜÄÁ¨œu– æ1*ˆä!ÚtœËF!ÂÜ®êÃè)Ä ÌnŒì$.ªÁRï_ì˜wLþFhÝ9szðDüÏyØßáòã\'5Tšp%‰Ä‘†ØU¹WÄÎÒ¥¸½è%‚ÅE0µöað…;ð^q´#3¿>ºÂbXV¥–†!MÔ¡ãð_k›`oz3O«‘ɼÚÛe-¥ìïaym@*@f9ólmq¯ÝWŽëÐëÖü@ßòPöâÀœá³õÚÜdùˆ¨jÜgÉÚß´ÞŸþ·éVŽNx’Ÿ©t6áâ_gG gvqk›•Oe³ˆºÅÿ¼ôœµïù‡GOï±MÈ…«ÁφîAp<Ìä ü,ô¿QÊ,ûaÍãŽ{Þª;{Ì ýàN—x•àûŽ AšÄVèº;Ÿ>#Óœ‚B Lyã³/ãe‹8ÚrÞ}3½ÉMU+ºÖÉÙ#Uš÷­;Š;dñ¿¹ßï/¿§*Á§ÈÞžÄI ÿÍ=ÿ¹Ž„‚Jˆ÷R>šÄ“Ê_L.mäÖ¸%uMþ© ûÐÖêLbĽƒ”'¯w É³j§­H¯@A3«—>§ZbJ”µFØsÐýçè¼i!7¬GQÞn¹ÔÊÓ?·öâ1“ÞåÙ?—Ç͵>_Ä[©ðg†š¢>9اd’·Ó©ãæt.)kD€„¹·NŠ»‘ÓB6….³ýdÒñxVE¯âktæ™ÕÞÓ(<ÏÛíà@…5dYñDòO鬲h`Ö[¬óÙ©S¹QÁö!}°ÀVq¥¿l(|󲓽ìn/6–F¬¶PÁ!H1¨ÆÏR•«Ñ‡ÅØdŒ$ÙHc–ýš–ù´G>Ër0ˆ€õ4h[Ïô^ Ä,¿‚¸%lX·ºJÓùµar;"õ½ø–òÓæ ûÉm¦À§ýRD 0 ÞÝÁŸ[ôšCçWA¡¡ßÛiKi’½º§Àd¦D8s2#’ôoàkˆJ²[úZûgï¼ñ'‘\?‘­-\=ºH¥r&Õ\ý/ÙºÉ~KbÞj;j9c™µ7cƒn]rÿð^l¡}ß"tëÃÉçëj"_¢íHÔT'OÈíòM»!Ê*;çìT7½Õ§¦äçÒßAÔ½=Ç›Ž¢ßyß;¶–×0áuäÅ'<2c]«×„šŠ•¢”'n|r‚‹¥Üšäš×ºðIˆ¼¼±§âÚa}ÄHTÇ¡Áß'—?u™rC-¶ä?ì3üeÆÙÐ_@ÏhX¯9QNÏæc ?­Ÿ#ØIgÈép Ô×r VJ¾wlG5¨J®£¸·™þôߘò¢BÞÆi–BMbŒqX‹âe s…jÑ5RûÆt“½ o£xm Á’¸À#c+ÁÀZí+Ìô€wyÏ8¥,ˆÜð’:]~±Ú˜Âm 31,çïÂÎX§?â°|ÆhÅPÝr, éÔ¬KÖ|¤é›µ÷ýj|ß–Ö߯F‹'ÕÛ?nß/ 5Ÿa÷Q­ld…õWPkYÔ&äÌœ):GÚ'P¦þ½¹ÙTõIµIâM³â'ºÕŬßcWÛ±‰ˆZßn ¡]ï*@æ½ÝÔªeÒ¥¼²ã™(bsqçÜÊ›O¯ÒÛ*—±ÚݯŽGBµc+à3h‹~P‰ ‰˜¦1´`§mëÔLmA ‡"J\â"Ÿ.lð;ÑÐdGÄ |´^Wp¨Ú“ŒHõBp÷uTHISù*ý‘ùo€¿büÖÓÈåÔ Ci-Äc€¹ñÉü&ósÊÞO1çl>Ç^¦Ët˜] “+¿ø~rdøÏŽScq”ŽÜç‘Ê]bƒ£HÇ$ÈõZÈóã.æßVû¯ 7Ìz'Ñ>ôò¡1[ú;ØššZz&Ñ%vXXFØ ¸O®^a8Üú^¡Ë=L8ÈW²÷ûm 8®’T/~ƒñzãÄñƒöB'Mw»˜Hç¶6¾­7(~Ë þuŠ\˜bÇqðKÿ¿_’ ßÉï!èõ‰Ú»eTU¿æÂŧâŘEð׉WëŽãWŽ@Ĉ“Ý:åh](“æA!ÆV y7òu'3ûë»b_¦©[ªÄÌv¤E{´èbO<Áäü!YnÈe8„æs‘¼~ÏਚÃìÜÎzË%‰Ÿ§È+e=ÙvÕú"KÃë-d<*ïÇz_ÿ4x‘NÊ&˜m™ Àî`°!¸^A•ºh.tç#2Ó}„â­pr™\ÒTeÉšøŒj ºß<εýžâ~夑X¬’«*O;[˜ª9POkûñøû'2§¿½›·rÿù~_Ü5MÏ+¢ýK=`0$äDf‘úÀ¤V,wG¨º x¿ê¼RÀ®äJ“GæÀA‹2} B¶0ÅùJ=áì Žûs+0þ窑/hÈÉR.: [ßâ kň1NÆ™7¯Ó9†Tº¿ŸÃ¯¯Ì@(HÃαÄJeN¹JVeaáƒ@ß¿Ëç>©":øc6siŽ#ª¤›\d!lÓuåPUÝ~ydc½D•‡îƽJaÃÃLorRÞbŠ8ÌÃd‡áN3éÙØ«?»{€¦‹Þ½5cŒH.¸šçÔкJRèc>½ÁÅáQe DHfÛ”‹ªÈü"g4¥Åà÷ßóc-›mçh³kasÓƒì{ö§Ï&­K¹ÛÞûd±Ÿ¬.V&€Æê;XgäÛµ×ôº¤í‡j/6ÒòïhëqXNŒ e'>÷ `Lƒ%WA 1 '?Å¢žÏç‰mrËFä-Þø4W›dºå óxË…uvÖ ÿß-Ç…ŽsvÁãæËZŒ<¿ðŸ,g´3j±b׉®hb:—2m”-³ÃWœ‚Y̾͘EÉÚÓøYÆÄ~ÃÀ:(°ñ¬¤súj“ù¿•14ýw9q©*ÆÃ(‚“s«uêæ‘”ŽÛ…oüxNðVáäwš"9Na)ÖHpPƇºUëRÏEu˜äÍ8Šu?ŠÛ$èzN”Їèºh™Ò7ßœ$Qé&Ém4­šèx*T¢$´0 öüÓ¨zlEx†›]t¥«²@u?> ¹eD-\’9Ä9b´b¯xÊñݦM°>Ú@F%d™f×ЙÎ/¥‹!s3"‚V5”8•ßÒÇ7x4p—»ÉD4ŠÂ¿¹¨¨œ ³AHâ>g2ãlOôU®Ì–©«6}J3ê b­P¹d}63” ¤PvÜD´²ù/Ñ>“H·——-:ƒ,‚Ì=gÁÄîúÀ+D ¶ç¾8…iÀýMÀ5]py{}âbâ–_®ðÃnkš=ú™¥¿¹Kÿ.á\Í×îç\'îÌ ÿT%èzˆ6ã¤ÿGz~[1h½l5èmo)m1—ûÓ§ZBaøs6–ä>¦¾zZP·bOˆBTfXx·´7*|/ü"éÀ}ÿ$ÿ'FÙ,±*k¯ô{šWК5_vbã´"àbÐ:ª„nñ(ôQ8øxÜc4ŠÍi <òL+" •$ìbn‰¨¶tñ@pÌlĈKÊk¢5©.~¡õ‚ôxö¾—X)™aS¤—xF¶€ÿÓ =ëEî5¹n‡ž~ö@9ŽA”ß–]·ŒÌ$ 69iCiï’0Nh b![ëTIçölé¾^Á^¢ÖAòJD]šã¬möjù_°KÝMƒHš±iúÁ\+#§ŠûœÛ³?“Oïd¦¥1UdŠÒ˜¥óÆÅx”]sÕ?"÷sFÑ«ìº]6¦œê7ñƒ+ÐÔ‹¬VÁ@èJ„ÆÝÖQ\l¡2ñ€Ô¾ZXD hPxz„¡¦ Fëjãû›0Ï„Ô2#¿Ë¥©éT7Iwž™2ʯa}´7,Qè1mNFÖ 9l½&½ `EEQ%gôŒªéÐeÍÂŽŽmdJš®[Ð~”5ª½òm°ˆí·º ç^5.£Rdãç³b;Q¸$½D°v’Rèß½™þ\×°‡ðÖ®‹Càý­š-‚4œ3šýˆ|L¤-5í À¬/Tv˜šŠúŽÁ›`›ßÓéæôîÍù¼Ãnö5i¼mq]U²¿‘ÆlrW ,vûÚ¹IÝÏC:xèÇéÑ.8Ô¥ÓýÉf™÷ŒFa\Øâzq~mc¦N˜)Ú¾¼©• Å„8ê%)`íÀ!áäߟ´!cƒ#`SDXì‰|~ÁƒcÉ©”‹§Önëôa–ÞÑY:UèV;+•ì[Ãm£Íþ\ %Wòx¨Xúr#H½›&Ó´µfîÄl«@ ì´],Q…$kÑéÂç°Sœå4ÙØ9ˆ¶œt-ËÖXŒ_ièvlÐs™Þ4{¥]6ÒËÔ–µ)­ ßÏw¢f#FÛj¼ßêÇÜ ]SKìÚ>8é,ãÃz<ÖKí¡Õ%4è^–îZ˜LÙøÓAß ›{Ë$®$þƒZŽw~_oûTû1ae}k…¬Âýt¨a`!Ú5²¶CŸ)ÏWdGv XÖ£SãÔ-÷µˆ*é0ýVî»ï¼Ú©¦FóG!·È GêÑ1L()E1G•¼%MB;I&Ér¦ž1o«¼à=²ý—EùºE¾Bg_9Ã1–gᘸr(ºs`p/Êekð·"ôçU«Õ7pƒŠ§Aá Ñü†×?§•-ý‘m‚2TŽcp!%pšLèšQEže8ùœwu¥GVmö\¾`…08›c 4n5yMŸ™îMaØ‚dCI@\Ò9#aÀlû|„ôÚ%!Ú¦ãï5B öÓ8¹ÆÊ®8s>á«ÿu ?oÝg¦ý¦2—meÝYAaƯ×u ïöŠ î~]h·¨×Ø7 ´¹n/ôF4ÛÔúáFG/=AÆTŽ7Ö–&—o¡B-“¡Ζñ‡Êûêz‰´€9´½ºp I°!=Ørw_*î»ú¥¤ÛûÝéØÏCM×.L|`zD5Nd¼V™—û&sq0sÐá%å;¨´ƒ’²½;”,4o$tRgV„·FiQCl#Ö®Ï~T¬ç>^ÝES ùÇŠã`*N¦ÌŽI~lNTº‹…^Ç=JÜbŽ©ööÃ|‡˜¾ü¼Tß|aTÉñPÊUz83Ä¥•QýÕz!ë_ט“ÒÙßÍjâIïqÝzd*:í-r:ãzO|¾àÔJ¾‰XÄ LÝb)Ëz+’§yVâ;:G1´«WnJµ,WÆH”¾]C~N Ó¢?.#f  Vn"ýQý"v^¦¸}à³ú­\v/ò›Ãé}$OóOZÇè|¬¼-rŽùçoÁô&lþ)ÿ_â×£^ /Åœ¾ØÖOVÅÝ šy®Kfj=²õp{Ë™=åo{½žµo!dk²¿1‡òSã}*Å ÊÌÔ²AvÊÜVU_-\ü’X“GЀY‰™-ï@•ÕcªŠB~ßü9ÏÎB1Ô›4£ëÞÆÝÌúƒœ Õå¤ÇL¼ì?ãD2±”ÿvÍËšã¿Pë;븜˜ã­D·M˜p«åæN]¿½B:VŽñrþ ?îיÌ s èz¿+úÓò[¤CÉKRSþÐjJ`ó0J¾÷Ô‘ÅC±sOÔ5ÈçjöŠ&‹8 ñöa–0×z~;µÀO‡”`ô;·v£|`Fƒ¤eŒ›Íý×JèÞfÒfÌ݃à÷ŸÔç¢Mƒw,’ôúÝ[Y~:gßŽÄ  ù†!>G/JVo2æ({Åv>(¨ÞÜ5¶:‹x*ì“}.ì^²ã Û¿¬ ŸYcƒ­¾ºŠ+¨“î“S‰?”™tÒÁøÀ§ß¯Ç"%jg@Ì=Ùl)8ó÷c¹d˜Ï¼É²ug‹¢`$»¦~ãjèj©ŸÝI^¢¹Ï¤s9)ù=¾1Öf•Ó2î/wëæ"€pƘ(Gý4‡—óP½XG÷_B‡XÆ p¨†LB<Ô )ëÅB–˜ßÊÇmoà<©âÁËE˜7«£šÜnú%)Šœ”:™ÑU°ü®Ò—´YêXÈQ~ 1Üäê¡ïö?ñ½ºa±m ûóYe~æ÷¡«ê3~ ÔÜÈÏω3sjt–CÇ""[­»Or67<«_ ’ˆPq9“µ (¦g„{˼Ît…ë/¶´Á¦bñ®Mí|F°†ÜÑ|ºjYÜGl ²íÊX… Â=´Ó=qoi+ǦJ¿fv¹mhïnU1§Ñ:9»u-Ý[1Ëý­ì8S7K)B ŠCÕH&Ì;˜º°9P²¹m,ÔˆYH®¦û¶"¤·–”«ûtHî±ÆÀ»£qª¢0"¨íb°£S—Iì] !"(݃ÊgÚþbº%´0;AyÐ n†ÂsP\Pî6n—}(Á$ÿÆt} ã(…Ò.Ø1=Ö¦í÷aw˜•EG„Ð+ºàÁ[8~¶°*©£‰%‰>ÆUßÊ—-z¹<0ñ泞yÓ‰i­A„RÇÀϸ˜æÝšvY0%SXýÔmw{LÔÄ-LÚãþ8ì“a0þm*¬õ3Õ 9çeáY“‚hÄÃoœl^†ÿÞŒèäêÐúç/"*MËXÖ3ýÂó½A¾ÌŸÔ“TKÂHÝ«ž#jžJõØe¡èÿ"GØqôþo]¾,ON!O畦ñ€ É¡‹p-ú¿žD¶øÁExØXmÑ P÷8 éÂF‹’šFÆÚÚ¿Di‰JQ ;ÒQfοx’ EÂÉyÍ“÷ƒ0±¡´)‚r£Ä]Ø¿¬s”mŽìlxÔ‡¦Y€e#›ÔF…’ ᣞ“žì­.`¸`Z ä Ïž|ü2 ÉÆÝò2r†ßÚ™1õG'pF(«7Ú1ƒÒIp−†Ã%¼¡3ìS¾‘xm”w[G×!jXêÝÝ•ˆ ¹"=—烙Å噟vÌ%PV2Ù2\ý÷¯Åµd‚-g£A[/¬ÙAWmÛqý_Ú½Þ-aÃzæf´]¼k>Þ4Ù%p1â=UQ–/“ÍaY°ñ¤Œ£ÀzÝ›¹1„œôMb(¸†®Çé¼7â…^@ny1J`Ô]Næ) 'q:£#ÿûl)½< ,RÅ|nä6÷|µ ìU‚&(O¡aºÅ­2"nÅ—ø­}Þ‘økµË{ÔëeJ5á àFÀƒ˜¦còQ]úßcë—Ì>chÑv(N·¬wãtôÚ:â5h¤¼quA_‡K•+?DþÊj˦`ì‚[,f‹ÂW&\w‘+À›Ô x*¦‡ÿ2úøtºÝ.º¥Ö¦8Rª u.Æ८QüQ|ïa%çB£cHF9Ö–?5eém¹Ê¨¨k¦4zW‚j'ø«âŽÓ\ÌR׎ÝJ•41¼dÞÙ[Tñ1ÎËô œe;¨8¦|Þ\¡Ñ³ñÍ“¸¡ñß#GççS B«~$gô*{þQOZê^iØ.ÇQ™žƒÜ¬ñoUq¥B€MóG¦U ‹É,Ößæ_Õ4§˜u[ÔÞc£;Š%xBf¢Ú)(¢|z·ÖnÌ=jÕVÅ â´êÒ\n4ÝûsÃU"²šrܽŒº)ÕG…ºÉ*}âuÇ~…73;Q$EÀœ›·¼XþIøªlÀ\ÈS«=æd¿S®õ@*ø*(‘?¾Ý4wçfHü@… ÎfÄVƒú·žRUòNxraƒ¬Žë$Þ6µ=VA½A0Ó>‚è¦þýEÝp-Ôæ?œê—ž,,,0¬‰z$œír»ÎX`¹}ͦ$.zX{ÞÕrµ ¬€÷G¶C€VÌÍÿÜ&ÝMÎ6R”¡¡ C犦I|b.RïxÀž»Ÿ5¶GSµ§Ji;³aÎéI?¿ïHÌ·»~Í ×Kð«ñÃ"J‘´¶RS“5læ&§d<¡„Þ3YR4JÓ›c|}læ¾uÖ]Üô¶ž°rdªwRJÔ¢þsëð‘ÜíÜر=¨ôç“ö¥Òÿ~cþo'2Llå†ÍÓËÎeÑWbèEoóâšéÉïÔM¹õxFú(˽ãæ0!/(¿ÜuM ÉåÌŒNäÑ™¬Ó´58õh2.bª@hËêUÅéHm&ÊB’ˆÐ¹mœÉ·Cm.ð‹ñ÷ŠéòÉщÈsƒžŽÞâ ”ê†F¹AÅ í>Áô|ía¨ÞVÚvöÛð€OF9¶´Â—! •¼^[‰7È%mžÕ‚b’3’nYj@}*››9 i¾F¬ù÷ïÔÉÒÒÐ?UU‘UTAµô•<òƒ |t@žzÄŒêì8>±øÌ‚uSDt$涋1‘¨¸”0ù9¬ ®‘ŽD´¤Y_EÝœë@†Æ’ô&¡"Ÿ1@¾4 èkd—Ò2Ó’ †¿ì0µÑ~/ýK,WÚ৆ì#Û®]*å|˘=¯ € hÛ¶¿7¦9vd¿|c•h¦‹.ÿÄ4>ÏÞ):G’綵°| Ä-$ýH¥¾±8Âð¦íKI‰ãÂRrtò ¢M˜0eœÒ]]ËÏ¢ Y+ÿZâ! ¦e<Š©¬qÔÄ^íEl%à`C_ ¿êgk:Ú357ùFÄ:Wšdº©ÏøT$hcàÄÑ­Çù£é°ž%ãùÅÛõcÜ/|¹ÆT‚Áº8•A§*ÅÄÿ+5¸j<3“®Ð€ºræ—^îËë_Ç”¿Ä}¢2ÝAŠ‚2õ¹Éu’§TaS¯»ï“ê‘_è)á\…ÐóÏl ¾ ½¤HèU%ƒH´½H ¦{’ š¨©íß(¬¯ÛÐ0÷½ìòéUÐ-Öò<‹ î™û›]g'gHC;âM¯{d>4¾ïMžÍ&CU"ÈS®ïc´W›Ÿy ‰°5à'h4–¥TÍê ÿŒ®·xE?Šœ "Þ§{(¸+Cù –º€kÐêZÑl#û¿ó©ßŒò¿î£ìG}5,ôwJ$v§f{_,÷oßä·DçBCAP½-\<™ú­Mósl¯Hæg-b›§ÛÜt¼â»î‡ (?¨[ Øêk ¾Á¯úq3d¦¢Õèý¶$Ë‚­2 Å£ „è'¶Óá¥T&œÜÚž1ìï&Ä'2££Àg‰ê(=e­ÂPº“ýÓJ±VómY—œïOM{´[\bŠ;ðÁ¯ч·±H©â9ÄŒF iRÛãù?8çÃ×_nšè™ i·~nü>{••Òç¨ù‚¾.ó3›í¾ñ—²<ÐÆg—–s4kg/7»¾# AÆñém×®Ç û„l$÷§}óÓãb3ÿJ¡57ÓòÚ®­N“úþ¦—v¤o蘢T&¦Ô¿Tms²’¨uÀ·9¯^쥷¸½–MÌ3(Št(×IÙÕ‘>–™Ý)ñOÞ룽E. œ`[HPƒ™ç)ñà>ºÊ9úCòb½„A²² ±ÏI¥Py”R˜°ÑZ-ÿLÏ·m¹KÆ-³úE¨;»gûþKÖÄö¤_¤šÒ•ükj¦e–¿î}ñìƒÒécÏHé{j؆Ö3?¶·ÎfÕ>môð¹zz Ñ;4 ²à5œÎLÛýš·)침¼Š]R ±•“¿Ó iO–¢Ü—¶9OÏZM}ãbø“ ó*ÍòÞk|R³ˆ†Âo«Ë¤Ï0Çx…9úî6¯°Ã3PQ`×|PÐÜX˜òEè¬ÆçXÜÀbˆÒNORêÜI›OžsW#zNÂàVý5Y°[xœÌ×ùúîoª€S%S0,d;x²nçŒ*B—Ã¥:×áÌ1ûWO›l¬4ëúIÔ£k#û žªØíî¯Yö.˜‘2PxªÆU[ﺓµ(e67"åÚÔIt} «˜òÁ{zfÛRÇd¤Œ îkOõâ ¯;ªÎv=×ᬸÍo U‹ò¹ÂcF6†ld ;€a‘nÔ¹:r¾dΤ¢O¤Þ5û=¡zl)V~ý"iÓï¢,±ÖU†Xp??£Õÿï=b†/ Ûéa“ǸÓPÉ¢´0Ø–»mÏZ‰he‹Ìó a~‚\·7… #ô•` uçÖÍHý•=¢©ZFÏÌÁlkQ½ò—O/)µ±ŸÉø<ž>Z™2™ÇI[sÈ#y“ø±¥œ¹' ][I\ˆ(ä…’}ìHa”!ø$o-L€À‚ûàIb =­W g†6•e ­^)òõd‰÷»H`Ê »¤ã¼pÂ`%‡9˸yç®ö¤ó}eª¬žN+IØiÑZj+¦D+Wc8~±è>÷¡,CUnZ9J¶´Í®`ýì“ î“hÿzº%gÿM“ž1)nÓÈ•NÊ.ŽÏƒƒNWv5ˆNžXð}™æatàþƒÁ…T£¾Wxˆù•·¯Îàå[èN´³Ây°3#W®‘‡y(Åôg7YçÚí¤lû‡_+êX§}éý^Á5¡Ý$ôBEÁŒúµ¨+¬¾‹{^4‘“K›^Ÿ˜†ô¿Bâ ®–#‹ÿÜÖœ̶ñ.K=²µS;/4J7 OyU‡Cçe”ê˜Áú@VcWÁCC Fýf²Ë3 v ¬L:OÒŽ`"[¹iÈRƒëò” ¾?–ßN4r:Ëw [‰…KF³bàëÏYz126mùçüåâ'ò{·ññP ç$Ö4Gõh†‰Ùa‘Ñ §mL…‡ëšöZUÄ•Îá•K¨”êñÒ“rfÚÒbUëÛÓª2¸Yí›õFúÉjiZ,Ê{o†Õ04ÎÇ®^Ç'NãlÞvš¸ÈŽªâ§þDíPšyZÚ+ÄQ,äìåÓ) ìÎHî+ËOÄ䟳BW*…Í—m&I;è Ò›å·5¬Ç“TšVØó´±jaºÛ㦛¿ö¦!•3‹=@à+î¤úH}—$9Ø\ªÊ­á±Y©[ýäz#MÈÝWðâ€V„FØtR" ¯ ?r;þ‹zÑóf±4ÒørdŠõ{'‰ÚšyP} :—æ¤ÑÕØFÔrá’VÔsšîàÐhÈŽ &“–H" b¨‹ÒˆÍ®SÊ\—HŒNP‹¬øFw•i"vt—ÖÑñ>ä1MQT|ÊÔÙ¦•iXY”e»Ÿ®®§â ±q¼ûk®Dj@ô~äiÆPj5jU^N§½&H46Y•0ʈ}—'.ÎÕ}’‘AæK†ÒæÃý>)ïû•FSŠfV©÷HáöKÕ{ÈtØÞÓPEpwø’ò!RÒX8<›ÄÁÅtnþ Ò.u¼A†J:]ßw#Äȵí&PÊ%»0M> endobj 1190 0 obj << /Length1 725 /Length2 5524 /Length3 0 /Length 6115 /Filter /FlateDecode >> stream xÚmreTjÔ5!ÝÝ0tà Ý-Ý)Í0 08Ô0tH§´4Ò H Š --ˆ ]‚”)-ñrï}ï}×·Ö·ž?çì³Ï:ûìçp²ª{z M½  AP`¢nj ’‘…Dñ99UP0æé¡FBeP'€ Ô ˆ |N€ª§W æâŠð@xÿæp°Ì†˜{Â=ý`W€œŸŸŸ’Ÿ¯ÂWAè¡É  ]¡g P50´ÔÒ×ðhè›4 P0ôu„à ]êáå8{"ðÄÓà ö—&¡¿xøAÈaÎOw€ÞcSeu}SÀcUaSUØà  «ùÐîô‘y`C!Hä_Û ÿEpð‘ãÑUDÀ?‘>p‚AG¨ Ì_ø/ß´<œ=ÿÀN¾^ÿ–ù<èð‹ýlÐö~%O¦zîPÝþ‹t5m~úÔCƒ×–›<ßꎺÁ_O:éSvà¨xù}#®$jð“ÎÞµ¬&ÂëGœUÉÆ)¸l§Žêˆ©¯ÞåW¡ˆçú÷¤V^ìeîÞ¦•â¥V ?ß‚:¢ßX—w·e~¯Eø7yS»ÛT›?@‡1¼ "Ó›Á3Áñ4rG hÆBÝ:+Ú‰DA–˯†Àø³9<¦œvª'cK•œçeKË ìÈ­L¬ç:ã£!…U% #è:vû§ ”ª÷¤”9þ¤3Ÿ‘:ä¤TŠËfŸxNGr?rÍ6Â.Q$¢dËÖË?|Ž ‰¾_þó¥Åuþf2yåWL"mS¢Ä—V~W­{€… s)Gõ•aÎyég¢u[ÌÒ¾¼­ˆ¦ÎîÚ~½ gÇkçŽöÁ_qt:ŽÕ½]¢c]§qß4KW¾šIZ‘‹ãB¦äiðz8 \Ô7“Û{²~4¿3ñVàbå¸Ú÷6“ o¾…¹òOcV‹-%qsð€™ÊfUcÊH¹FyªŠµ¬øˆ„PR,ÕB>1¼9Ùú=ƒ è'nŒ’úÉÓºl (˜»âqáGöí™'^‡Âë#ÿ4Iªz¯ú¡TìdzŒˆ‰ñe:ÁA¡E7Œ6‘zÊ~ÆñrÒÚZgqò™š%UW·]@¸†UŠžf ª¶¶µ*Љ¦þÏ[Ë•&4%ŠŠÝ“3]Îðj ´‰g)®åqkQ?€>¡rG¼°XMâE…ô9Rå”>ã[R(3“§ùNëŽtF~WàPï®åH¾ž›aÌEk9¿cûµ/ð›êv]§×J »*2CÅ[×ó,À¥¯ aðYàW¶P-ÏþnÕ.ÿj$ªmÇÁ™ZʺªbBqþm‚ñSôŒwÏ„M”µ=úÁÑ,u¦s?c¤äµ¶©šmãαÄå‘TÃi¥@HcàYÃrî;2:ñëÆ Ð6Ò9ùüÃA¼®j%|+:ܲPݘÓEæ¯eH·¹:OòP$·žÞu4ð‡L©6}Nq´hO‹ÚiWÇk¿¦¿wÿ2a»H)ŒX™‹È<¨ú^?ƒpˆSjªa»tž,hýœ¬ƒˆžÊ~Líÿi¬£…2õŒŸá¥¯ØÙoUV õhA#Jˆó‹Vu‡«ö›| êE'Œ·FïÍ"Œ^/p<Ú x`v.Æâÿ":ÛpÉšØËh»ø‰þH+7±$˜?ŸFY¦á4F*î•Ö¤f‚îOŽûÍ.»(UðfuåW©ë´rü¢ú£ÛMüNwæ»)NOÁ °‘Æ’‚Õà™“J­Î¤€ïk3›‰bèN·MNñ±Íu%$vzêp?÷)ÜÊ‹©‡Å-­UÉ”‡›‘©”(Fœþ¾ Ÿm¹Þ»™ÅÏþ‚x÷í] iaægÙo”®€pDW,cùψ˜Î?öÙe ç­L–0užT}q/p}¸n×ÙR©és¢ë­¸­OI¾&] v7äÃ,£½µ°Ü²aÎÁKæš—ëÓ/ûL±¸—U0iDÔ[RNódƒÍSñ˜pr\à\ûLØQùïļ I€9ÍòbÕy:îKŒúË4Ó¸ e Ò¬ÏLœ½P•ßáÑ-âz’X}*§ß@ Jôø’ ô%2×mxâ…0”€gxã­rÍ—ÒW‚Þý†(™¨l+QZü½€3ï²²f74÷+•®Ô¦-¦~‹#=þZ£¹Ã{óukþv$ÒW½‘ˆb­ünÖûbX[äº{¾ÎÂŒ—9|?íœäƒÄìø›Qs‘ÓænWd>ýçJ-ÅÅ.´í`RZ 0{BsUÊ iVO\«šZ “þ‡DopÕ¯)NvagÝüÏ–îÇǺnfA¹r¾²Öû l-¨׆PÂRQ;†7)²=âE Ž‹<©vÈLþ^ ìÑWôìBÔ¦™-A ÝKÚ÷7{ Üùç­è®Htá.iI²÷¡ˆ z;xäýÏ0Æ>àvÏß&BîƒïXž5I2Â:†~%•`Ž˜\7ÍÇ®ÞJØõ륓€´wO $IžûÊ©FÅÒ‹Äg°ÞKëãR$À+Ë”î<9AóÞ¬ž¹óÅÛCP1d+Áø2 -_³®è?øtTO¹/a¶X?ñ†A^Eì+oß›Òg5†ëX.ò?C«õ±ÑáŠgf{ú¢˜†ö"b>-•{Ãß5s4_½i¸.Ÿ>´CgVô½7–UÏ®Šé |ßÞP´ ÂT:佰؈@?muÎ]–&_»(ïr o%O‘.Qæ}õ[ºÆª¶²BÓŒš"§0¤Ú4ÖŸŸ¶jˆÒÒ'Ê`­UåÔæIî¯_(yAY™ø»L»Mqyf%éL¢±e‰ÂûíYux[žÎf—°*ñëç#|¨£°—ÉÃnÝù ‰`¦dÁ$õàN4yóáÛúîmŠ‰jn––#N¶#¢_¹)ã ŽD*¤hF…LqEsµ±Z‰Ž tálx\9µ«TÊGj±Ù·³e^Æ&FÔb(‹dÆLe*»ÈĽóë ш´_+ýQ‡à}êDç=7køƒFJʧñ€1ßYx›sb”àe¢>“ÝØ?&:z7(»—Ì![„©äŸ>1rÌR©A‡Ž[º+x³MЪãêÔ§E£všæ­µ7Ù´Tí.{= ñIPWÄ> ‰Ð,5›ä—jG¾_”NO_p$æŒåiâ¯ßl…_½y½Jý~6‘VPü¬c ·Wßá<‚Æ<%­Øq—ˆ²c À;zqßu÷š Ó ÷D®·¯úÆš?ÈìqˆØz‘!‘IÂ| Ï>ÉŒ{g»A0?Šÿ9 ±«@âèe2À’C5ŠŒ?W¼žyK.° 4vM˜eqwM éæ"ã+õ •t8’e æ™äõ8,Nj#Nÿp4ÜYm°â­ÛÈ„œøñ¢£×¡ÈÛ™~µ9n>^O³Þ¼®?`Ø:3O8®¹o¥¯°&9zÑh©g½œ<08v×åD=”«’‰§*vÒ™OCÞZcpW*þÁ~ ¥ÐÕÑ…ÐN~““̤ô–Cõëäƒ-ŸÙõW5! Yž-7M«4—¯‰œÇ·Kå Fˆ/¤ fø5àE¾k®½çnõÁe‚æz‡WNmQÜ`¼/ÚŸûZ¸—Y°¹Nu†­JfÝ»ÒT–ÃïmdÒ€ß,8?O‘' œGº sšÕë Áðz^”¥/Í(É… 'ŸKe‚ýÝ 0zW]pNœá‘˜0Ûê˜GŸ´ò^Ì= œs> \ï©À[âajÅ¡µÑí\HÑ/Ó×ÕÕ|joXcG&gÿ­ö÷|PAŒ¿šË.{<9@ZZH|û L^"=Ü–tOÛ˜·D PV]ÂóÍ=·ýí9Ùvó^|àWöQ2ójYN톪·žÊµãظ6i¨]íùS®Ärº—¥zpoIϬc ZvÌ•8½ãîä€Å3^Ætúº)tž@¾k;®úëE»#ª ¸+Ë¥ŒT,cÞJëýᗾΰIýK–¨Ä˜lNýÌ îÑw˜ÚWDÌ•55„ëe  [w‚%³`µ†x—·_ïµ^Nr‰®>=5B£9í°?q5]cJ<Á<”¢î4.´šVÄlÆMôÕŸþªõñ÷ÔqKæ¸5UK8 2Z%±]Û“À{O‚;<[ºs;!nÂÓäé¸j=6˜6:ìäAà —ï.£*ä=§éK}Ó[;h‡ÛC/¸j6ðû–’•×CËœ±3˜ºJ,n-¹2oèÔ¿ñaÞ8a»}7î.ôExß‘6]¶û:4I¹üÉ·–t*ž=± ö?.6ñ5²q_¸ú)ãðæ‹T&•¥­I¤ÐÔ¼_®gªb`óeä÷³b}w©µ²Ž$i8%WòÕþ¦à×J¶êìS#¶/«¬±œjþ˜#e07Mgé`>.O<ÿqZ¾ƒyR‰ e;¡‹|ä*—)‹¹ ½YµJrjPŸÁveŸk˜4=8üTòÂު̶üF;|h„½LšyU8V8EíPEQn× (eÞå&‹î&”Äìl‚7táRÝ%ô’ûqEÕÙö*ú'Ößéõºdôè`OÖEÍ\†7—IÏ~[äÑ6¤óôU‹“ß`¹ÿ>|”]#ž´j%t¢²F±Åå.,%:–¤ï~´üêPNyvµïôH¶€ã‡öF¡ÐóÆ¥¹²42 ôÄê÷ntî–ãXIÌõ%j›°¨Æ˜k Ûé9ó’IìSÊ’64ïJŸÆ'Y|4·ü"“õRËWþÒÑãHòÄÂàÚY¹¹º¡|?Q©‹qÉŒ$š¤ï:GÓ8½6mDhW¡çJ-¤MQ¾w,àæ²üBÿ| ?µ†¥Àµ¬²èéÒÜ‹sóNìÜ @;¨jâ¶0‡/©¸ï®¥puÎP(ý)&LZÄ­uª÷häÈŠ|.ã±ÞJú»ÓÛÕ<ŠŽSÔÎÚ-n™‚+©íZêæJ¾•ó½¯ò­Ÿ3GçuÜw Äë"¥”{ΤkµÈ=sq… "Í. ;» H?´ŸÊäI³G–/Ï1áÌéaôf×ãwÛš}ýË nL4R&0I© ô±s\©ŽooO¡œW”3{Â; x2|ÔÚý†ÝË”|g[tïiL¹‘§¿R‰¤û§,KA´&»¹*!Ûϸmš%菱b¦NP|7' OõK–Àe5Ÿó?r KšÞRDt2ˆ…øs?†r¿PO¢ÞÃÉ4ÿSêØAP«Û«”fGåÔò‹Ð[J„¾?rM½Iž’É:³A€ ¼¾ŠÇçÀç²ñïn;Ôo_Zéâ™2RKªµc)ÒÕ çà¡[ë1ûJñ],}Ú®f–—¼*EÜñÇÕ8õ©ó@bst÷ôe…ëÜô˜þ¾ëÉ”XotózX£™âÇh 1÷xÂ.}Ä«ì ×ߌƒ*߯<.¤ª½Îbu?`ø&ÆS4Xˆ08^;°ù® ²‹IY¶àø\Žpú|=ù¹äˆfÿzÕ½—‘#N&뻶aÔEÛ©R¾ ¶g¡¥ÙÎá‚ü——Õ¾„}¤1YW¾×¥f6|VXÞ “nGêŸBè€MâgÝézTt’ƒÚyX­¡ˆÇ{3¬V²öß"A”5ô¸Ý1?+s‹oc2¢Zzt£¾Î-àâRr¾}Πٸ—jB¶; X±a?ny™G&0nŸA²k„hµ=Ó±ºÍèg}Ï.i#Ê^Þ5U«»hUÌ?©®•°’ªï5Lsù§Þ¦n (ŒáÐÑZ®šdPèÏ[»FUµŽŒØÙ´Ö+¼…#J‰dtŽŠ^þ!ÇmyKÓLã©n_æ Õ¢Ýó.^T‰Ôxõ)ÄÖÖíÕõ!öd}sN_(Ÿ¼ñÉÒ«óþ1S>ˆhIžõˆLͺgÁ#ùjkë¢Þ#‹5–Ža[h9Ü2»nýÔjÁúŠ¹ÁüEÐ%=*ÃWªûõÚð¤ìÝ^yxQêYSòëö²ûô¬/2ŸêiaTq‚¦ª§s– õÞ»Y¦¨A²³Î=mNZY{RâK+øÐ^·lÝ°Ä5½=Fè—ìhž;M“=£¡qýl6XƒB»þþ†ºô½8ru›ýÍéa™qeoæÿzÅ| endstream endobj 1191 0 obj << /Type /FontDescriptor /FontName /YCSVIU+SFTT1728 /Flags 4 /FontBBox [-194 -360 1303 830] /Ascent 611 /CapHeight 611 /Descent -222 /ItalicAngle 0 /StemV 50 /XHeight 430 /CharSet (/G/L/M/P) /FontFile 1190 0 R >> endobj 1142 0 obj << /Type /Encoding /Differences [2/fi/fl 38/ampersand/quoteright/parenleft/parenright 44/comma/hyphen/period/slash/zero/one/two/three/four/five/six/seven/eight/nine/colon/semicolon 65/A/B/C/D/E/F/G/H/I/J/K/L/M/N/O/P 82/R/S/T/U/V/W/X 92/backslash 97/a/b/c/d/e/f/g/h/i/j/k/l/m/n/o/p/q/r/s/t/u/v/w/x/y/z 149/bullet 152/tilde 178/twosuperior 233/eacute] >> endobj 1133 0 obj << /Type /Encoding /Differences [34/quotedbl 37/percent/ampersand/quoteright/parenleft/parenright/asterisk/plus/comma/hyphen/period/slash/zero/one/two/three/four/five/six/seven/eight/nine/colon/semicolon/less/equal/greater/question/at/A/B/C/D/E/F/G/H/I 75/K/L/M/N/O/P/Q/R/S/T/U/V/W 89/Y/Z/bracketleft/backslash/bracketright/asciicircum/underscore 97/a/b/c/d/e/f/g/h/i/j/k/l/m/n/o/p/q/r/s/t/u/v/w/x/y/z/braceleft/bar/braceright/asciitilde] >> endobj 1122 0 obj << /Type /Encoding /Differences [178/twosuperior] >> endobj 206 0 obj << /Type /Font /Subtype /Type1 /BaseFont /JTQSEL+CMEX10 /FontDescriptor 1149 0 R /FirstChar 0 /LastChar 113 /Widths 1136 0 R >> endobj 223 0 obj << /Type /Font /Subtype /Type1 /BaseFont /IGNLSA+CMSY10 /FontDescriptor 1151 0 R /FirstChar 1 /LastChar 112 /Widths 1132 0 R >> endobj 551 0 obj << /Type /Font /Subtype /Type1 /BaseFont /HWTUNG+CMSY7 /FontDescriptor 1153 0 R /FirstChar 3 /LastChar 112 /Widths 1131 0 R >> endobj 149 0 obj << /Type /Font /Subtype /Type1 /BaseFont /YWGOBB+SFTT1000 /FontDescriptor 1185 0 R /FirstChar 34 /LastChar 126 /Widths 1144 0 R /Encoding 1133 0 R >> endobj 190 0 obj << /Type /Font /Subtype /Type1 /BaseFont /WIGSRM+SFTT1095 /FontDescriptor 1187 0 R /FirstChar 39 /LastChar 125 /Widths 1141 0 R /Encoding 1133 0 R >> endobj 221 0 obj << /Type /Font /Subtype /Type1 /BaseFont /WDUAMV+SFTT1440 /FontDescriptor 1189 0 R /FirstChar 103 /LastChar 112 /Widths 1134 0 R /Encoding 1133 0 R >> endobj 147 0 obj << /Type /Font /Subtype /Type1 /BaseFont /YCSVIU+SFTT1728 /FontDescriptor 1191 0 R /FirstChar 71 /LastChar 80 /Widths 1146 0 R /Encoding 1133 0 R >> endobj 552 0 obj << /Type /Font /Subtype /Type1 /BaseFont /KJNCRM+EUEX10 /FontDescriptor 1155 0 R /FirstChar 14 /LastChar 125 /Widths 1130 0 R >> endobj 203 0 obj << /Type /Font /Subtype /Type1 /BaseFont /OATESS+EUFM10 /FontDescriptor 1157 0 R /FirstChar 40 /LastChar 93 /Widths 1138 0 R >> endobj 207 0 obj << /Type /Font /Subtype /Type1 /BaseFont /HSNTCT+EUFM7 /FontDescriptor 1159 0 R /FirstChar 40 /LastChar 61 /Widths 1135 0 R >> endobj 574 0 obj << /Type /Font /Subtype /Type1 /BaseFont /BYCRVB+EURB10 /FontDescriptor 1161 0 R /FirstChar 3 /LastChar 32 /Widths 1126 0 R >> endobj 573 0 obj << /Type /Font /Subtype /Type1 /BaseFont /BCJSZT+EURB7 /FontDescriptor 1163 0 R /FirstChar 3 /LastChar 32 /Widths 1127 0 R >> endobj 201 0 obj << /Type /Font /Subtype /Type1 /BaseFont /HKOQTE+EURM10 /FontDescriptor 1165 0 R /FirstChar 0 /LastChar 122 /Widths 1140 0 R >> endobj 202 0 obj << /Type /Font /Subtype /Type1 /BaseFont /SIDADR+EURM7 /FontDescriptor 1167 0 R /FirstChar 0 /LastChar 122 /Widths 1139 0 R >> endobj 204 0 obj << /Type /Font /Subtype /Type1 /BaseFont /EJLECR+EUSM10 /FontDescriptor 1169 0 R /FirstChar 68 /LastChar 106 /Widths 1137 0 R >> endobj 678 0 obj << /Type /Font /Subtype /Type1 /BaseFont /MCIEGV+EUSM7 /FontDescriptor 1171 0 R /FirstChar 68 /LastChar 106 /Widths 1125 0 R >> endobj 566 0 obj << /Type /Font /Subtype /Type1 /BaseFont /YFMQYV+MSAM10 /FontDescriptor 1173 0 R /FirstChar 62 /LastChar 88 /Widths 1129 0 R >> endobj 572 0 obj << /Type /Font /Subtype /Type1 /BaseFont /FITBVO+MSBM10 /FontDescriptor 1175 0 R /FirstChar 69 /LastChar 86 /Widths 1128 0 R >> endobj 148 0 obj << /Type /Font /Subtype /Type1 /BaseFont /OZLTXQ+Sabon-Bold /FontDescriptor 1177 0 R /FirstChar 39 /LastChar 121 /Widths 1145 0 R /Encoding 1142 0 R >> endobj 146 0 obj << /Type /Font /Subtype /Type1 /BaseFont /UIRQAT+Sabon-Roman /FontDescriptor 1179 0 R /FirstChar 2 /LastChar 233 /Widths 1147 0 R /Encoding 1142 0 R >> endobj 150 0 obj << /Type /Font /Subtype /Type1 /BaseFont /BIRPMW+Sabon-Italic /FontDescriptor 1181 0 R /FirstChar 2 /LastChar 122 /Widths 1143 0 R /Encoding 1142 0 R >> endobj 1068 0 obj << /Type /Font /Subtype /Type1 /BaseFont /SYEPTZ+SFTT0800 /FontDescriptor 1183 0 R /FirstChar 178 /LastChar 178 /Widths 1123 0 R /Encoding 1122 0 R >> endobj 1067 0 obj << /Type /Font /Subtype /Type1 /BaseFont /WIGSRM+SFTT1095 /FontDescriptor 1187 0 R /FirstChar 178 /LastChar 178 /Widths 1124 0 R /Encoding 1122 0 R >> endobj 151 0 obj << /Type /Pages /Count 6 /Parent 1192 0 R /Kids [138 0 R 186 0 R 198 0 R 218 0 R 306 0 R 370 0 R] >> endobj 454 0 obj << /Type /Pages /Count 6 /Parent 1192 0 R /Kids [424 0 R 460 0 R 506 0 R 556 0 R 563 0 R 569 0 R] >> endobj 634 0 obj << /Type /Pages /Count 6 /Parent 1192 0 R /Kids [579 0 R 637 0 R 684 0 R 727 0 R 784 0 R 823 0 R] >> endobj 830 0 obj << /Type /Pages /Count 6 /Parent 1192 0 R /Kids [827 0 R 832 0 R 836 0 R 840 0 R 846 0 R 899 0 R] >> endobj 946 0 obj << /Type /Pages /Count 5 /Parent 1192 0 R /Kids [921 0 R 950 0 R 999 0 R 1064 0 R 1075 0 R] >> endobj 1192 0 obj << /Type /Pages /Count 29 /Kids [151 0 R 454 0 R 634 0 R 830 0 R 946 0 R] >> endobj 1193 0 obj << /Type /Outlines /First 7 0 R /Last 123 0 R /Count 6 >> endobj 135 0 obj << /Title 136 0 R /A 133 0 R /Parent 123 0 R /Prev 131 0 R >> endobj 131 0 obj << /Title 132 0 R /A 129 0 R /Parent 123 0 R /Prev 127 0 R /Next 135 0 R >> endobj 127 0 obj << /Title 128 0 R /A 125 0 R /Parent 123 0 R /Next 131 0 R >> endobj 123 0 obj << /Title 124 0 R /A 121 0 R /Parent 1193 0 R /Prev 107 0 R /First 127 0 R /Last 135 0 R /Count -3 >> endobj 119 0 obj << /Title 120 0 R /A 117 0 R /Parent 107 0 R /Prev 115 0 R >> endobj 115 0 obj << /Title 116 0 R /A 113 0 R /Parent 107 0 R /Prev 111 0 R /Next 119 0 R >> endobj 111 0 obj << /Title 112 0 R /A 109 0 R /Parent 107 0 R /Next 115 0 R >> endobj 107 0 obj << /Title 108 0 R /A 105 0 R /Parent 1193 0 R /Prev 35 0 R /Next 123 0 R /First 111 0 R /Last 119 0 R /Count -3 >> endobj 103 0 obj << /Title 104 0 R /A 101 0 R /Parent 99 0 R >> endobj 99 0 obj << /Title 100 0 R /A 97 0 R /Parent 35 0 R /Prev 91 0 R /First 103 0 R /Last 103 0 R /Count -1 >> endobj 95 0 obj << /Title 96 0 R /A 93 0 R /Parent 91 0 R >> endobj 91 0 obj << /Title 92 0 R /A 89 0 R /Parent 35 0 R /Prev 75 0 R /Next 99 0 R /First 95 0 R /Last 95 0 R /Count -1 >> endobj 87 0 obj << /Title 88 0 R /A 85 0 R /Parent 75 0 R /Prev 83 0 R >> endobj 83 0 obj << /Title 84 0 R /A 81 0 R /Parent 75 0 R /Prev 79 0 R /Next 87 0 R >> endobj 79 0 obj << /Title 80 0 R /A 77 0 R /Parent 75 0 R /Next 83 0 R >> endobj 75 0 obj << /Title 76 0 R /A 73 0 R /Parent 35 0 R /Prev 55 0 R /Next 91 0 R /First 79 0 R /Last 87 0 R /Count -3 >> endobj 71 0 obj << /Title 72 0 R /A 69 0 R /Parent 55 0 R /Prev 67 0 R >> endobj 67 0 obj << /Title 68 0 R /A 65 0 R /Parent 55 0 R /Prev 63 0 R /Next 71 0 R >> endobj 63 0 obj << /Title 64 0 R /A 61 0 R /Parent 55 0 R /Prev 59 0 R /Next 67 0 R >> endobj 59 0 obj << /Title 60 0 R /A 57 0 R /Parent 55 0 R /Next 63 0 R >> endobj 55 0 obj << /Title 56 0 R /A 53 0 R /Parent 35 0 R /Prev 51 0 R /Next 75 0 R /First 59 0 R /Last 71 0 R /Count -4 >> endobj 51 0 obj << /Title 52 0 R /A 49 0 R /Parent 35 0 R /Prev 47 0 R /Next 55 0 R >> endobj 47 0 obj << /Title 48 0 R /A 45 0 R /Parent 35 0 R /Prev 43 0 R /Next 51 0 R >> endobj 43 0 obj << /Title 44 0 R /A 41 0 R /Parent 35 0 R /Prev 39 0 R /Next 47 0 R >> endobj 39 0 obj << /Title 40 0 R /A 37 0 R /Parent 35 0 R /Next 43 0 R >> endobj 35 0 obj << /Title 36 0 R /A 33 0 R /Parent 1193 0 R /Prev 15 0 R /Next 107 0 R /First 39 0 R /Last 99 0 R /Count -8 >> endobj 31 0 obj << /Title 32 0 R /A 29 0 R /Parent 15 0 R /Prev 27 0 R >> endobj 27 0 obj << /Title 28 0 R /A 25 0 R /Parent 15 0 R /Prev 23 0 R /Next 31 0 R >> endobj 23 0 obj << /Title 24 0 R /A 21 0 R /Parent 15 0 R /Prev 19 0 R /Next 27 0 R >> endobj 19 0 obj << /Title 20 0 R /A 17 0 R /Parent 15 0 R /Next 23 0 R >> endobj 15 0 obj << /Title 16 0 R /A 13 0 R /Parent 1193 0 R /Prev 11 0 R /Next 35 0 R /First 19 0 R /Last 31 0 R /Count -4 >> endobj 11 0 obj << /Title 12 0 R /A 9 0 R /Parent 1193 0 R /Prev 7 0 R /Next 15 0 R >> endobj 7 0 obj << /Title 8 0 R /A 5 0 R /Parent 1193 0 R /Next 11 0 R >> endobj 1194 0 obj << /Names [(Doc-Start) 145 0 R (equation.1.1) 205 0 R (equation.3.2) 550 0 R (equation.4.3) 575 0 R (lstlisting.-1) 224 0 R (lstlisting.-10) 438 0 R] /Limits [(Doc-Start) (lstlisting.-10)] >> endobj 1195 0 obj << /Names [(lstlisting.-11) 463 0 R (lstlisting.-12) 509 0 R (lstlisting.-13) 582 0 R (lstlisting.-14) 687 0 R (lstlisting.-15) 738 0 R (lstlisting.-16) 755 0 R] /Limits [(lstlisting.-11) (lstlisting.-16)] >> endobj 1196 0 obj << /Names [(lstlisting.-17) 787 0 R (lstlisting.-18) 802 0 R (lstlisting.-19) 849 0 R (lstlisting.-2) 235 0 R (lstlisting.-20) 905 0 R (lstlisting.-21) 953 0 R] /Limits [(lstlisting.-17) (lstlisting.-21)] >> endobj 1197 0 obj << /Names [(lstlisting.-22) 1069 0 R (lstlisting.-3) 316 0 R (lstlisting.-4) 318 0 R (lstlisting.-5) 323 0 R (lstlisting.-6) 331 0 R (lstlisting.-7) 346 0 R] /Limits [(lstlisting.-22) (lstlisting.-7)] >> endobj 1198 0 obj << /Names [(lstlisting.-8) 377 0 R (lstlisting.-9) 403 0 R (lstnumber.-1.1) 225 0 R (lstnumber.-1.2) 226 0 R (lstnumber.-1.3) 227 0 R (lstnumber.-1.4) 228 0 R] /Limits [(lstlisting.-8) (lstnumber.-1.4)] >> endobj 1199 0 obj << /Names [(lstnumber.-1.5) 229 0 R (lstnumber.-1.6) 230 0 R (lstnumber.-1.7) 231 0 R (lstnumber.-1.8) 232 0 R (lstnumber.-1.9) 233 0 R (lstnumber.-10.1) 439 0 R] /Limits [(lstnumber.-1.5) (lstnumber.-10.1)] >> endobj 1200 0 obj << /Names [(lstnumber.-10.10) 448 0 R (lstnumber.-10.11) 449 0 R (lstnumber.-10.12) 450 0 R (lstnumber.-10.13) 451 0 R (lstnumber.-10.14) 452 0 R (lstnumber.-10.15) 453 0 R] /Limits [(lstnumber.-10.10) (lstnumber.-10.15)] >> endobj 1201 0 obj << /Names [(lstnumber.-10.2) 440 0 R (lstnumber.-10.3) 441 0 R (lstnumber.-10.4) 442 0 R (lstnumber.-10.5) 443 0 R (lstnumber.-10.6) 444 0 R (lstnumber.-10.7) 445 0 R] /Limits [(lstnumber.-10.2) (lstnumber.-10.7)] >> endobj 1202 0 obj << /Names [(lstnumber.-10.8) 446 0 R (lstnumber.-10.9) 447 0 R (lstnumber.-11.1) 464 0 R (lstnumber.-11.10) 473 0 R (lstnumber.-11.11) 474 0 R (lstnumber.-11.12) 475 0 R] /Limits [(lstnumber.-10.8) (lstnumber.-11.12)] >> endobj 1203 0 obj << /Names [(lstnumber.-11.13) 476 0 R (lstnumber.-11.14) 477 0 R (lstnumber.-11.15) 478 0 R (lstnumber.-11.16) 479 0 R (lstnumber.-11.17) 480 0 R (lstnumber.-11.18) 481 0 R] /Limits [(lstnumber.-11.13) (lstnumber.-11.18)] >> endobj 1204 0 obj << /Names [(lstnumber.-11.19) 482 0 R (lstnumber.-11.2) 465 0 R (lstnumber.-11.20) 483 0 R (lstnumber.-11.21) 484 0 R (lstnumber.-11.22) 485 0 R (lstnumber.-11.23) 486 0 R] /Limits [(lstnumber.-11.19) (lstnumber.-11.23)] >> endobj 1205 0 obj << /Names [(lstnumber.-11.24) 487 0 R (lstnumber.-11.25) 488 0 R (lstnumber.-11.26) 489 0 R (lstnumber.-11.27) 490 0 R (lstnumber.-11.28) 491 0 R (lstnumber.-11.29) 492 0 R] /Limits [(lstnumber.-11.24) (lstnumber.-11.29)] >> endobj 1206 0 obj << /Names [(lstnumber.-11.3) 466 0 R (lstnumber.-11.30) 493 0 R (lstnumber.-11.31) 494 0 R (lstnumber.-11.32) 495 0 R (lstnumber.-11.33) 496 0 R (lstnumber.-11.34) 497 0 R] /Limits [(lstnumber.-11.3) (lstnumber.-11.34)] >> endobj 1207 0 obj << /Names [(lstnumber.-11.35) 498 0 R (lstnumber.-11.36) 499 0 R (lstnumber.-11.37) 500 0 R (lstnumber.-11.38) 501 0 R (lstnumber.-11.4) 467 0 R (lstnumber.-11.5) 468 0 R] /Limits [(lstnumber.-11.35) (lstnumber.-11.5)] >> endobj 1208 0 obj << /Names [(lstnumber.-11.6) 469 0 R (lstnumber.-11.7) 470 0 R (lstnumber.-11.8) 471 0 R (lstnumber.-11.9) 472 0 R (lstnumber.-12.1) 510 0 R (lstnumber.-12.10) 519 0 R] /Limits [(lstnumber.-11.6) (lstnumber.-12.10)] >> endobj 1209 0 obj << /Names [(lstnumber.-12.11) 520 0 R (lstnumber.-12.12) 521 0 R (lstnumber.-12.13) 522 0 R (lstnumber.-12.14) 523 0 R (lstnumber.-12.15) 524 0 R (lstnumber.-12.16) 525 0 R] /Limits [(lstnumber.-12.11) (lstnumber.-12.16)] >> endobj 1210 0 obj << /Names [(lstnumber.-12.17) 526 0 R (lstnumber.-12.18) 527 0 R (lstnumber.-12.19) 528 0 R (lstnumber.-12.2) 511 0 R (lstnumber.-12.20) 529 0 R (lstnumber.-12.21) 530 0 R] /Limits [(lstnumber.-12.17) (lstnumber.-12.21)] >> endobj 1211 0 obj << /Names [(lstnumber.-12.22) 531 0 R (lstnumber.-12.23) 532 0 R (lstnumber.-12.24) 533 0 R (lstnumber.-12.25) 534 0 R (lstnumber.-12.26) 535 0 R (lstnumber.-12.27) 536 0 R] /Limits [(lstnumber.-12.22) (lstnumber.-12.27)] >> endobj 1212 0 obj << /Names [(lstnumber.-12.28) 537 0 R (lstnumber.-12.29) 538 0 R (lstnumber.-12.3) 512 0 R (lstnumber.-12.30) 539 0 R (lstnumber.-12.31) 540 0 R (lstnumber.-12.32) 541 0 R] /Limits [(lstnumber.-12.28) (lstnumber.-12.32)] >> endobj 1213 0 obj << /Names [(lstnumber.-12.33) 542 0 R (lstnumber.-12.34) 543 0 R (lstnumber.-12.35) 544 0 R (lstnumber.-12.36) 545 0 R (lstnumber.-12.37) 546 0 R (lstnumber.-12.38) 547 0 R] /Limits [(lstnumber.-12.33) (lstnumber.-12.38)] >> endobj 1214 0 obj << /Names [(lstnumber.-12.39) 548 0 R (lstnumber.-12.4) 513 0 R (lstnumber.-12.40) 549 0 R (lstnumber.-12.5) 514 0 R (lstnumber.-12.6) 515 0 R (lstnumber.-12.7) 516 0 R] /Limits [(lstnumber.-12.39) (lstnumber.-12.7)] >> endobj 1215 0 obj << /Names [(lstnumber.-12.8) 517 0 R (lstnumber.-12.9) 518 0 R (lstnumber.-13.1) 583 0 R (lstnumber.-13.10) 592 0 R (lstnumber.-13.11) 593 0 R (lstnumber.-13.12) 594 0 R] /Limits [(lstnumber.-12.8) (lstnumber.-13.12)] >> endobj 1216 0 obj << /Names [(lstnumber.-13.13) 595 0 R (lstnumber.-13.14) 596 0 R (lstnumber.-13.15) 597 0 R (lstnumber.-13.16) 598 0 R (lstnumber.-13.17) 599 0 R (lstnumber.-13.18) 600 0 R] /Limits [(lstnumber.-13.13) (lstnumber.-13.18)] >> endobj 1217 0 obj << /Names [(lstnumber.-13.19) 601 0 R (lstnumber.-13.2) 584 0 R (lstnumber.-13.20) 602 0 R (lstnumber.-13.21) 603 0 R (lstnumber.-13.22) 604 0 R (lstnumber.-13.23) 605 0 R] /Limits [(lstnumber.-13.19) (lstnumber.-13.23)] >> endobj 1218 0 obj << /Names [(lstnumber.-13.24) 606 0 R (lstnumber.-13.25) 607 0 R (lstnumber.-13.26) 608 0 R (lstnumber.-13.27) 609 0 R (lstnumber.-13.28) 610 0 R (lstnumber.-13.29) 611 0 R] /Limits [(lstnumber.-13.24) (lstnumber.-13.29)] >> endobj 1219 0 obj << /Names [(lstnumber.-13.3) 585 0 R (lstnumber.-13.30) 612 0 R (lstnumber.-13.31) 613 0 R (lstnumber.-13.32) 614 0 R (lstnumber.-13.33) 615 0 R (lstnumber.-13.34) 616 0 R] /Limits [(lstnumber.-13.3) (lstnumber.-13.34)] >> endobj 1220 0 obj << /Names [(lstnumber.-13.35) 617 0 R (lstnumber.-13.36) 618 0 R (lstnumber.-13.37) 619 0 R (lstnumber.-13.38) 620 0 R (lstnumber.-13.39) 621 0 R (lstnumber.-13.4) 586 0 R] /Limits [(lstnumber.-13.35) (lstnumber.-13.4)] >> endobj 1221 0 obj << /Names [(lstnumber.-13.40) 622 0 R (lstnumber.-13.41) 623 0 R (lstnumber.-13.42) 624 0 R (lstnumber.-13.43) 625 0 R (lstnumber.-13.44) 626 0 R (lstnumber.-13.45) 627 0 R] /Limits [(lstnumber.-13.40) (lstnumber.-13.45)] >> endobj 1222 0 obj << /Names [(lstnumber.-13.46) 628 0 R (lstnumber.-13.47) 629 0 R (lstnumber.-13.48) 630 0 R (lstnumber.-13.49) 631 0 R (lstnumber.-13.5) 587 0 R (lstnumber.-13.50) 632 0 R] /Limits [(lstnumber.-13.46) (lstnumber.-13.50)] >> endobj 1223 0 obj << /Names [(lstnumber.-13.51) 633 0 R (lstnumber.-13.52) 640 0 R (lstnumber.-13.53) 641 0 R (lstnumber.-13.54) 642 0 R (lstnumber.-13.55) 643 0 R (lstnumber.-13.56) 644 0 R] /Limits [(lstnumber.-13.51) (lstnumber.-13.56)] >> endobj 1224 0 obj << /Names [(lstnumber.-13.57) 645 0 R (lstnumber.-13.58) 646 0 R (lstnumber.-13.59) 647 0 R (lstnumber.-13.6) 588 0 R (lstnumber.-13.60) 648 0 R (lstnumber.-13.61) 649 0 R] /Limits [(lstnumber.-13.57) (lstnumber.-13.61)] >> endobj 1225 0 obj << /Names [(lstnumber.-13.62) 650 0 R (lstnumber.-13.63) 651 0 R (lstnumber.-13.64) 652 0 R (lstnumber.-13.65) 653 0 R (lstnumber.-13.66) 654 0 R (lstnumber.-13.67) 655 0 R] /Limits [(lstnumber.-13.62) (lstnumber.-13.67)] >> endobj 1226 0 obj << /Names [(lstnumber.-13.68) 656 0 R (lstnumber.-13.69) 657 0 R (lstnumber.-13.7) 589 0 R (lstnumber.-13.70) 658 0 R (lstnumber.-13.71) 659 0 R (lstnumber.-13.72) 660 0 R] /Limits [(lstnumber.-13.68) (lstnumber.-13.72)] >> endobj 1227 0 obj << /Names [(lstnumber.-13.73) 661 0 R (lstnumber.-13.74) 662 0 R (lstnumber.-13.75) 663 0 R (lstnumber.-13.76) 664 0 R (lstnumber.-13.77) 665 0 R (lstnumber.-13.78) 666 0 R] /Limits [(lstnumber.-13.73) (lstnumber.-13.78)] >> endobj 1228 0 obj << /Names [(lstnumber.-13.79) 667 0 R (lstnumber.-13.8) 590 0 R (lstnumber.-13.80) 668 0 R (lstnumber.-13.81) 669 0 R (lstnumber.-13.82) 670 0 R (lstnumber.-13.83) 671 0 R] /Limits [(lstnumber.-13.79) (lstnumber.-13.83)] >> endobj 1229 0 obj << /Names [(lstnumber.-13.84) 672 0 R (lstnumber.-13.85) 673 0 R (lstnumber.-13.86) 674 0 R (lstnumber.-13.87) 675 0 R (lstnumber.-13.88) 676 0 R (lstnumber.-13.89) 677 0 R] /Limits [(lstnumber.-13.84) (lstnumber.-13.89)] >> endobj 1230 0 obj << /Names [(lstnumber.-13.9) 591 0 R (lstnumber.-14.1) 688 0 R (lstnumber.-14.10) 697 0 R (lstnumber.-14.11) 698 0 R (lstnumber.-14.12) 699 0 R (lstnumber.-14.13) 700 0 R] /Limits [(lstnumber.-13.9) (lstnumber.-14.13)] >> endobj 1231 0 obj << /Names [(lstnumber.-14.14) 701 0 R (lstnumber.-14.15) 702 0 R (lstnumber.-14.16) 703 0 R (lstnumber.-14.17) 704 0 R (lstnumber.-14.18) 705 0 R (lstnumber.-14.19) 706 0 R] /Limits [(lstnumber.-14.14) (lstnumber.-14.19)] >> endobj 1232 0 obj << /Names [(lstnumber.-14.2) 689 0 R (lstnumber.-14.20) 707 0 R (lstnumber.-14.21) 708 0 R (lstnumber.-14.22) 709 0 R (lstnumber.-14.23) 710 0 R (lstnumber.-14.24) 711 0 R] /Limits [(lstnumber.-14.2) (lstnumber.-14.24)] >> endobj 1233 0 obj << /Names [(lstnumber.-14.25) 712 0 R (lstnumber.-14.26) 713 0 R (lstnumber.-14.27) 714 0 R (lstnumber.-14.28) 730 0 R (lstnumber.-14.29) 731 0 R (lstnumber.-14.3) 690 0 R] /Limits [(lstnumber.-14.25) (lstnumber.-14.3)] >> endobj 1234 0 obj << /Names [(lstnumber.-14.30) 732 0 R (lstnumber.-14.31) 733 0 R (lstnumber.-14.32) 734 0 R (lstnumber.-14.33) 735 0 R (lstnumber.-14.34) 736 0 R (lstnumber.-14.35) 737 0 R] /Limits [(lstnumber.-14.30) (lstnumber.-14.35)] >> endobj 1235 0 obj << /Names [(lstnumber.-14.4) 691 0 R (lstnumber.-14.5) 692 0 R (lstnumber.-14.6) 693 0 R (lstnumber.-14.7) 694 0 R (lstnumber.-14.8) 695 0 R (lstnumber.-14.9) 696 0 R] /Limits [(lstnumber.-14.4) (lstnumber.-14.9)] >> endobj 1236 0 obj << /Names [(lstnumber.-15.1) 739 0 R (lstnumber.-15.10) 748 0 R (lstnumber.-15.11) 749 0 R (lstnumber.-15.12) 750 0 R (lstnumber.-15.13) 751 0 R (lstnumber.-15.14) 752 0 R] /Limits [(lstnumber.-15.1) (lstnumber.-15.14)] >> endobj 1237 0 obj << /Names [(lstnumber.-15.15) 753 0 R (lstnumber.-15.2) 740 0 R (lstnumber.-15.3) 741 0 R (lstnumber.-15.4) 742 0 R (lstnumber.-15.5) 743 0 R (lstnumber.-15.6) 744 0 R] /Limits [(lstnumber.-15.15) (lstnumber.-15.6)] >> endobj 1238 0 obj << /Names [(lstnumber.-15.7) 745 0 R (lstnumber.-15.8) 746 0 R (lstnumber.-15.9) 747 0 R (lstnumber.-16.1) 756 0 R (lstnumber.-16.10) 765 0 R (lstnumber.-16.11) 766 0 R] /Limits [(lstnumber.-15.7) (lstnumber.-16.11)] >> endobj 1239 0 obj << /Names [(lstnumber.-16.12) 767 0 R (lstnumber.-16.13) 768 0 R (lstnumber.-16.14) 769 0 R (lstnumber.-16.15) 770 0 R (lstnumber.-16.16) 771 0 R (lstnumber.-16.17) 772 0 R] /Limits [(lstnumber.-16.12) (lstnumber.-16.17)] >> endobj 1240 0 obj << /Names [(lstnumber.-16.18) 773 0 R (lstnumber.-16.19) 774 0 R (lstnumber.-16.2) 757 0 R (lstnumber.-16.3) 758 0 R (lstnumber.-16.4) 759 0 R (lstnumber.-16.5) 760 0 R] /Limits [(lstnumber.-16.18) (lstnumber.-16.5)] >> endobj 1241 0 obj << /Names [(lstnumber.-16.6) 761 0 R (lstnumber.-16.7) 762 0 R (lstnumber.-16.8) 763 0 R (lstnumber.-16.9) 764 0 R (lstnumber.-17.1) 788 0 R (lstnumber.-17.10) 797 0 R] /Limits [(lstnumber.-16.6) (lstnumber.-17.10)] >> endobj 1242 0 obj << /Names [(lstnumber.-17.11) 798 0 R (lstnumber.-17.12) 799 0 R (lstnumber.-17.13) 800 0 R (lstnumber.-17.14) 801 0 R (lstnumber.-17.2) 789 0 R (lstnumber.-17.3) 790 0 R] /Limits [(lstnumber.-17.11) (lstnumber.-17.3)] >> endobj 1243 0 obj << /Names [(lstnumber.-17.4) 791 0 R (lstnumber.-17.5) 792 0 R (lstnumber.-17.6) 793 0 R (lstnumber.-17.7) 794 0 R (lstnumber.-17.8) 795 0 R (lstnumber.-17.9) 796 0 R] /Limits [(lstnumber.-17.4) (lstnumber.-17.9)] >> endobj 1244 0 obj << /Names [(lstnumber.-18.1) 803 0 R (lstnumber.-18.10) 812 0 R (lstnumber.-18.11) 813 0 R (lstnumber.-18.12) 814 0 R (lstnumber.-18.13) 815 0 R (lstnumber.-18.14) 816 0 R] /Limits [(lstnumber.-18.1) (lstnumber.-18.14)] >> endobj 1245 0 obj << /Names [(lstnumber.-18.15) 817 0 R (lstnumber.-18.16) 818 0 R (lstnumber.-18.17) 819 0 R (lstnumber.-18.18) 820 0 R (lstnumber.-18.19) 821 0 R (lstnumber.-18.2) 804 0 R] /Limits [(lstnumber.-18.15) (lstnumber.-18.2)] >> endobj 1246 0 obj << /Names [(lstnumber.-18.3) 805 0 R (lstnumber.-18.4) 806 0 R (lstnumber.-18.5) 807 0 R (lstnumber.-18.6) 808 0 R (lstnumber.-18.7) 809 0 R (lstnumber.-18.8) 810 0 R] /Limits [(lstnumber.-18.3) (lstnumber.-18.8)] >> endobj 1247 0 obj << /Names [(lstnumber.-18.9) 811 0 R (lstnumber.-19.1) 850 0 R (lstnumber.-19.10) 859 0 R (lstnumber.-19.11) 860 0 R (lstnumber.-19.12) 861 0 R (lstnumber.-19.13) 862 0 R] /Limits [(lstnumber.-18.9) (lstnumber.-19.13)] >> endobj 1248 0 obj << /Names [(lstnumber.-19.14) 863 0 R (lstnumber.-19.15) 864 0 R (lstnumber.-19.16) 865 0 R (lstnumber.-19.17) 866 0 R (lstnumber.-19.18) 867 0 R (lstnumber.-19.19) 868 0 R] /Limits [(lstnumber.-19.14) (lstnumber.-19.19)] >> endobj 1249 0 obj << /Names [(lstnumber.-19.2) 851 0 R (lstnumber.-19.20) 869 0 R (lstnumber.-19.21) 870 0 R (lstnumber.-19.22) 871 0 R (lstnumber.-19.23) 872 0 R (lstnumber.-19.24) 873 0 R] /Limits [(lstnumber.-19.2) (lstnumber.-19.24)] >> endobj 1250 0 obj << /Names [(lstnumber.-19.25) 874 0 R (lstnumber.-19.26) 875 0 R (lstnumber.-19.27) 876 0 R (lstnumber.-19.28) 877 0 R (lstnumber.-19.29) 878 0 R (lstnumber.-19.3) 852 0 R] /Limits [(lstnumber.-19.25) (lstnumber.-19.3)] >> endobj 1251 0 obj << /Names [(lstnumber.-19.30) 879 0 R (lstnumber.-19.31) 880 0 R (lstnumber.-19.32) 881 0 R (lstnumber.-19.33) 882 0 R (lstnumber.-19.34) 883 0 R (lstnumber.-19.35) 884 0 R] /Limits [(lstnumber.-19.30) (lstnumber.-19.35)] >> endobj 1252 0 obj << /Names [(lstnumber.-19.36) 885 0 R (lstnumber.-19.37) 886 0 R (lstnumber.-19.38) 887 0 R (lstnumber.-19.39) 888 0 R (lstnumber.-19.4) 853 0 R (lstnumber.-19.40) 889 0 R] /Limits [(lstnumber.-19.36) (lstnumber.-19.40)] >> endobj 1253 0 obj << /Names [(lstnumber.-19.41) 890 0 R (lstnumber.-19.42) 891 0 R (lstnumber.-19.43) 892 0 R (lstnumber.-19.44) 893 0 R (lstnumber.-19.45) 902 0 R (lstnumber.-19.46) 903 0 R] /Limits [(lstnumber.-19.41) (lstnumber.-19.46)] >> endobj 1254 0 obj << /Names [(lstnumber.-19.47) 904 0 R (lstnumber.-19.5) 854 0 R (lstnumber.-19.6) 855 0 R (lstnumber.-19.7) 856 0 R (lstnumber.-19.8) 857 0 R (lstnumber.-19.9) 858 0 R] /Limits [(lstnumber.-19.47) (lstnumber.-19.9)] >> endobj 1255 0 obj << /Names [(lstnumber.-2.1) 236 0 R (lstnumber.-2.10) 245 0 R (lstnumber.-2.11) 246 0 R (lstnumber.-2.12) 247 0 R (lstnumber.-2.13) 248 0 R (lstnumber.-2.14) 249 0 R] /Limits [(lstnumber.-2.1) (lstnumber.-2.14)] >> endobj 1256 0 obj << /Names [(lstnumber.-2.15) 250 0 R (lstnumber.-2.16) 251 0 R (lstnumber.-2.17) 252 0 R (lstnumber.-2.18) 253 0 R (lstnumber.-2.19) 254 0 R (lstnumber.-2.2) 237 0 R] /Limits [(lstnumber.-2.15) (lstnumber.-2.2)] >> endobj 1257 0 obj << /Names [(lstnumber.-2.20) 255 0 R (lstnumber.-2.21) 256 0 R (lstnumber.-2.22) 257 0 R (lstnumber.-2.23) 258 0 R (lstnumber.-2.24) 259 0 R (lstnumber.-2.25) 260 0 R] /Limits [(lstnumber.-2.20) (lstnumber.-2.25)] >> endobj 1258 0 obj << /Names [(lstnumber.-2.26) 261 0 R (lstnumber.-2.27) 262 0 R (lstnumber.-2.28) 263 0 R (lstnumber.-2.29) 264 0 R (lstnumber.-2.3) 238 0 R (lstnumber.-2.30) 265 0 R] /Limits [(lstnumber.-2.26) (lstnumber.-2.30)] >> endobj 1259 0 obj << /Names [(lstnumber.-2.31) 266 0 R (lstnumber.-2.32) 267 0 R (lstnumber.-2.33) 268 0 R (lstnumber.-2.34) 269 0 R (lstnumber.-2.35) 270 0 R (lstnumber.-2.36) 271 0 R] /Limits [(lstnumber.-2.31) (lstnumber.-2.36)] >> endobj 1260 0 obj << /Names [(lstnumber.-2.37) 272 0 R (lstnumber.-2.38) 273 0 R (lstnumber.-2.39) 274 0 R (lstnumber.-2.4) 239 0 R (lstnumber.-2.40) 309 0 R (lstnumber.-2.41) 310 0 R] /Limits [(lstnumber.-2.37) (lstnumber.-2.41)] >> endobj 1261 0 obj << /Names [(lstnumber.-2.42) 311 0 R (lstnumber.-2.43) 312 0 R (lstnumber.-2.44) 313 0 R (lstnumber.-2.45) 314 0 R (lstnumber.-2.5) 240 0 R (lstnumber.-2.6) 241 0 R] /Limits [(lstnumber.-2.42) (lstnumber.-2.6)] >> endobj 1262 0 obj << /Names [(lstnumber.-2.7) 242 0 R (lstnumber.-2.8) 243 0 R (lstnumber.-2.9) 244 0 R (lstnumber.-20.1) 906 0 R (lstnumber.-20.10) 915 0 R (lstnumber.-20.11) 916 0 R] /Limits [(lstnumber.-2.7) (lstnumber.-20.11)] >> endobj 1263 0 obj << /Names [(lstnumber.-20.12) 917 0 R (lstnumber.-20.13) 918 0 R (lstnumber.-20.14) 919 0 R (lstnumber.-20.15) 924 0 R (lstnumber.-20.16) 925 0 R (lstnumber.-20.17) 926 0 R] /Limits [(lstnumber.-20.12) (lstnumber.-20.17)] >> endobj 1264 0 obj << /Names [(lstnumber.-20.18) 927 0 R (lstnumber.-20.19) 928 0 R (lstnumber.-20.2) 907 0 R (lstnumber.-20.20) 929 0 R (lstnumber.-20.21) 930 0 R (lstnumber.-20.22) 931 0 R] /Limits [(lstnumber.-20.18) (lstnumber.-20.22)] >> endobj 1265 0 obj << /Names [(lstnumber.-20.23) 932 0 R (lstnumber.-20.24) 933 0 R (lstnumber.-20.25) 934 0 R (lstnumber.-20.26) 935 0 R (lstnumber.-20.27) 936 0 R (lstnumber.-20.28) 937 0 R] /Limits [(lstnumber.-20.23) (lstnumber.-20.28)] >> endobj 1266 0 obj << /Names [(lstnumber.-20.29) 938 0 R (lstnumber.-20.3) 908 0 R (lstnumber.-20.30) 939 0 R (lstnumber.-20.31) 940 0 R (lstnumber.-20.32) 941 0 R (lstnumber.-20.33) 942 0 R] /Limits [(lstnumber.-20.29) (lstnumber.-20.33)] >> endobj 1267 0 obj << /Names [(lstnumber.-20.34) 943 0 R (lstnumber.-20.35) 944 0 R (lstnumber.-20.36) 945 0 R (lstnumber.-20.4) 909 0 R (lstnumber.-20.5) 910 0 R (lstnumber.-20.6) 911 0 R] /Limits [(lstnumber.-20.34) (lstnumber.-20.6)] >> endobj 1268 0 obj << /Names [(lstnumber.-20.7) 912 0 R (lstnumber.-20.8) 913 0 R (lstnumber.-20.9) 914 0 R (lstnumber.-21.1) 954 0 R (lstnumber.-21.10) 963 0 R (lstnumber.-21.100) 1058 0 R] /Limits [(lstnumber.-20.7) (lstnumber.-21.100)] >> endobj 1269 0 obj << /Names [(lstnumber.-21.101) 1059 0 R (lstnumber.-21.11) 964 0 R (lstnumber.-21.12) 965 0 R (lstnumber.-21.13) 966 0 R (lstnumber.-21.14) 967 0 R (lstnumber.-21.15) 968 0 R] /Limits [(lstnumber.-21.101) (lstnumber.-21.15)] >> endobj 1270 0 obj << /Names [(lstnumber.-21.16) 969 0 R (lstnumber.-21.17) 970 0 R (lstnumber.-21.18) 971 0 R (lstnumber.-21.19) 972 0 R (lstnumber.-21.2) 955 0 R (lstnumber.-21.20) 973 0 R] /Limits [(lstnumber.-21.16) (lstnumber.-21.20)] >> endobj 1271 0 obj << /Names [(lstnumber.-21.21) 974 0 R (lstnumber.-21.22) 975 0 R (lstnumber.-21.23) 976 0 R (lstnumber.-21.24) 977 0 R (lstnumber.-21.25) 978 0 R (lstnumber.-21.26) 979 0 R] /Limits [(lstnumber.-21.21) (lstnumber.-21.26)] >> endobj 1272 0 obj << /Names [(lstnumber.-21.27) 980 0 R (lstnumber.-21.28) 981 0 R (lstnumber.-21.29) 982 0 R (lstnumber.-21.3) 956 0 R (lstnumber.-21.30) 983 0 R (lstnumber.-21.31) 984 0 R] /Limits [(lstnumber.-21.27) (lstnumber.-21.31)] >> endobj 1273 0 obj << /Names [(lstnumber.-21.32) 985 0 R (lstnumber.-21.33) 986 0 R (lstnumber.-21.34) 987 0 R (lstnumber.-21.35) 988 0 R (lstnumber.-21.36) 989 0 R (lstnumber.-21.37) 990 0 R] /Limits [(lstnumber.-21.32) (lstnumber.-21.37)] >> endobj 1274 0 obj << /Names [(lstnumber.-21.38) 991 0 R (lstnumber.-21.39) 992 0 R (lstnumber.-21.4) 957 0 R (lstnumber.-21.40) 993 0 R (lstnumber.-21.41) 994 0 R (lstnumber.-21.42) 995 0 R] /Limits [(lstnumber.-21.38) (lstnumber.-21.42)] >> endobj 1275 0 obj << /Names [(lstnumber.-21.43) 996 0 R (lstnumber.-21.44) 1002 0 R (lstnumber.-21.45) 1003 0 R (lstnumber.-21.46) 1004 0 R (lstnumber.-21.47) 1005 0 R (lstnumber.-21.48) 1006 0 R] /Limits [(lstnumber.-21.43) (lstnumber.-21.48)] >> endobj 1276 0 obj << /Names [(lstnumber.-21.49) 1007 0 R (lstnumber.-21.5) 958 0 R (lstnumber.-21.50) 1008 0 R (lstnumber.-21.51) 1009 0 R (lstnumber.-21.52) 1010 0 R (lstnumber.-21.53) 1011 0 R] /Limits [(lstnumber.-21.49) (lstnumber.-21.53)] >> endobj 1277 0 obj << /Names [(lstnumber.-21.54) 1012 0 R (lstnumber.-21.55) 1013 0 R (lstnumber.-21.56) 1014 0 R (lstnumber.-21.57) 1015 0 R (lstnumber.-21.58) 1016 0 R (lstnumber.-21.59) 1017 0 R] /Limits [(lstnumber.-21.54) (lstnumber.-21.59)] >> endobj 1278 0 obj << /Names [(lstnumber.-21.6) 959 0 R (lstnumber.-21.60) 1018 0 R (lstnumber.-21.61) 1019 0 R (lstnumber.-21.62) 1020 0 R (lstnumber.-21.63) 1021 0 R (lstnumber.-21.64) 1022 0 R] /Limits [(lstnumber.-21.6) (lstnumber.-21.64)] >> endobj 1279 0 obj << /Names [(lstnumber.-21.65) 1023 0 R (lstnumber.-21.66) 1024 0 R (lstnumber.-21.67) 1025 0 R (lstnumber.-21.68) 1026 0 R (lstnumber.-21.69) 1027 0 R (lstnumber.-21.7) 960 0 R] /Limits [(lstnumber.-21.65) (lstnumber.-21.7)] >> endobj 1280 0 obj << /Names [(lstnumber.-21.70) 1028 0 R (lstnumber.-21.71) 1029 0 R (lstnumber.-21.72) 1030 0 R (lstnumber.-21.73) 1031 0 R (lstnumber.-21.74) 1032 0 R (lstnumber.-21.75) 1033 0 R] /Limits [(lstnumber.-21.70) (lstnumber.-21.75)] >> endobj 1281 0 obj << /Names [(lstnumber.-21.76) 1034 0 R (lstnumber.-21.77) 1035 0 R (lstnumber.-21.78) 1036 0 R (lstnumber.-21.79) 1037 0 R (lstnumber.-21.8) 961 0 R (lstnumber.-21.80) 1038 0 R] /Limits [(lstnumber.-21.76) (lstnumber.-21.80)] >> endobj 1282 0 obj << /Names [(lstnumber.-21.81) 1039 0 R (lstnumber.-21.82) 1040 0 R (lstnumber.-21.83) 1041 0 R (lstnumber.-21.84) 1042 0 R (lstnumber.-21.85) 1043 0 R (lstnumber.-21.86) 1044 0 R] /Limits [(lstnumber.-21.81) (lstnumber.-21.86)] >> endobj 1283 0 obj << /Names [(lstnumber.-21.87) 1045 0 R (lstnumber.-21.88) 1046 0 R (lstnumber.-21.89) 1047 0 R (lstnumber.-21.9) 962 0 R (lstnumber.-21.90) 1048 0 R (lstnumber.-21.91) 1049 0 R] /Limits [(lstnumber.-21.87) (lstnumber.-21.91)] >> endobj 1284 0 obj << /Names [(lstnumber.-21.92) 1050 0 R (lstnumber.-21.93) 1051 0 R (lstnumber.-21.94) 1052 0 R (lstnumber.-21.95) 1053 0 R (lstnumber.-21.96) 1054 0 R (lstnumber.-21.97) 1055 0 R] /Limits [(lstnumber.-21.92) (lstnumber.-21.97)] >> endobj 1285 0 obj << /Names [(lstnumber.-21.98) 1056 0 R (lstnumber.-21.99) 1057 0 R (lstnumber.-22.1) 1070 0 R (lstnumber.-22.10) 1083 0 R (lstnumber.-22.11) 1084 0 R (lstnumber.-22.12) 1085 0 R] /Limits [(lstnumber.-21.98) (lstnumber.-22.12)] >> endobj 1286 0 obj << /Names [(lstnumber.-22.13) 1086 0 R (lstnumber.-22.14) 1087 0 R (lstnumber.-22.15) 1088 0 R (lstnumber.-22.16) 1089 0 R (lstnumber.-22.17) 1090 0 R (lstnumber.-22.18) 1091 0 R] /Limits [(lstnumber.-22.13) (lstnumber.-22.18)] >> endobj 1287 0 obj << /Names [(lstnumber.-22.19) 1092 0 R (lstnumber.-22.2) 1071 0 R (lstnumber.-22.20) 1093 0 R (lstnumber.-22.21) 1094 0 R (lstnumber.-22.22) 1095 0 R (lstnumber.-22.23) 1096 0 R] /Limits [(lstnumber.-22.19) (lstnumber.-22.23)] >> endobj 1288 0 obj << /Names [(lstnumber.-22.24) 1097 0 R (lstnumber.-22.25) 1098 0 R (lstnumber.-22.26) 1099 0 R (lstnumber.-22.27) 1100 0 R (lstnumber.-22.28) 1101 0 R (lstnumber.-22.29) 1102 0 R] /Limits [(lstnumber.-22.24) (lstnumber.-22.29)] >> endobj 1289 0 obj << /Names [(lstnumber.-22.3) 1072 0 R (lstnumber.-22.30) 1103 0 R (lstnumber.-22.31) 1104 0 R (lstnumber.-22.32) 1105 0 R (lstnumber.-22.33) 1106 0 R (lstnumber.-22.34) 1107 0 R] /Limits [(lstnumber.-22.3) (lstnumber.-22.34)] >> endobj 1290 0 obj << /Names [(lstnumber.-22.35) 1108 0 R (lstnumber.-22.36) 1109 0 R (lstnumber.-22.37) 1110 0 R (lstnumber.-22.38) 1111 0 R (lstnumber.-22.39) 1112 0 R (lstnumber.-22.4) 1073 0 R] /Limits [(lstnumber.-22.35) (lstnumber.-22.4)] >> endobj 1291 0 obj << /Names [(lstnumber.-22.40) 1113 0 R (lstnumber.-22.41) 1114 0 R (lstnumber.-22.42) 1115 0 R (lstnumber.-22.43) 1116 0 R (lstnumber.-22.44) 1117 0 R (lstnumber.-22.45) 1118 0 R] /Limits [(lstnumber.-22.40) (lstnumber.-22.45)] >> endobj 1292 0 obj << /Names [(lstnumber.-22.46) 1119 0 R (lstnumber.-22.47) 1120 0 R (lstnumber.-22.48) 1121 0 R (lstnumber.-22.5) 1078 0 R (lstnumber.-22.6) 1079 0 R (lstnumber.-22.7) 1080 0 R] /Limits [(lstnumber.-22.46) (lstnumber.-22.7)] >> endobj 1293 0 obj << /Names [(lstnumber.-22.8) 1081 0 R (lstnumber.-22.9) 1082 0 R (lstnumber.-3.1) 317 0 R (lstnumber.-4.1) 319 0 R (lstnumber.-4.2) 320 0 R (lstnumber.-4.3) 321 0 R] /Limits [(lstnumber.-22.8) (lstnumber.-4.3)] >> endobj 1294 0 obj << /Names [(lstnumber.-5.1) 324 0 R (lstnumber.-5.2) 325 0 R (lstnumber.-5.3) 326 0 R (lstnumber.-5.4) 327 0 R (lstnumber.-5.5) 328 0 R (lstnumber.-5.6) 329 0 R] /Limits [(lstnumber.-5.1) (lstnumber.-5.6)] >> endobj 1295 0 obj << /Names [(lstnumber.-6.1) 332 0 R (lstnumber.-6.10) 341 0 R (lstnumber.-6.11) 342 0 R (lstnumber.-6.12) 343 0 R (lstnumber.-6.13) 344 0 R (lstnumber.-6.2) 333 0 R] /Limits [(lstnumber.-6.1) (lstnumber.-6.2)] >> endobj 1296 0 obj << /Names [(lstnumber.-6.3) 334 0 R (lstnumber.-6.4) 335 0 R (lstnumber.-6.5) 336 0 R (lstnumber.-6.6) 337 0 R (lstnumber.-6.7) 338 0 R (lstnumber.-6.8) 339 0 R] /Limits [(lstnumber.-6.3) (lstnumber.-6.8)] >> endobj 1297 0 obj << /Names [(lstnumber.-6.9) 340 0 R (lstnumber.-7.1) 347 0 R (lstnumber.-7.10) 374 0 R (lstnumber.-7.11) 375 0 R (lstnumber.-7.12) 376 0 R (lstnumber.-7.2) 348 0 R] /Limits [(lstnumber.-6.9) (lstnumber.-7.2)] >> endobj 1298 0 obj << /Names [(lstnumber.-7.3) 349 0 R (lstnumber.-7.4) 350 0 R (lstnumber.-7.5) 351 0 R (lstnumber.-7.6) 352 0 R (lstnumber.-7.7) 353 0 R (lstnumber.-7.8) 354 0 R] /Limits [(lstnumber.-7.3) (lstnumber.-7.8)] >> endobj 1299 0 obj << /Names [(lstnumber.-7.9) 373 0 R (lstnumber.-8.1) 378 0 R (lstnumber.-8.10) 387 0 R (lstnumber.-8.11) 388 0 R (lstnumber.-8.12) 389 0 R (lstnumber.-8.13) 390 0 R] /Limits [(lstnumber.-7.9) (lstnumber.-8.13)] >> endobj 1300 0 obj << /Names [(lstnumber.-8.14) 391 0 R (lstnumber.-8.15) 392 0 R (lstnumber.-8.16) 393 0 R (lstnumber.-8.17) 394 0 R (lstnumber.-8.18) 395 0 R (lstnumber.-8.19) 396 0 R] /Limits [(lstnumber.-8.14) (lstnumber.-8.19)] >> endobj 1301 0 obj << /Names [(lstnumber.-8.2) 379 0 R (lstnumber.-8.20) 397 0 R (lstnumber.-8.21) 398 0 R (lstnumber.-8.22) 399 0 R (lstnumber.-8.23) 400 0 R (lstnumber.-8.24) 401 0 R] /Limits [(lstnumber.-8.2) (lstnumber.-8.24)] >> endobj 1302 0 obj << /Names [(lstnumber.-8.25) 402 0 R (lstnumber.-8.3) 380 0 R (lstnumber.-8.4) 381 0 R (lstnumber.-8.5) 382 0 R (lstnumber.-8.6) 383 0 R (lstnumber.-8.7) 384 0 R] /Limits [(lstnumber.-8.25) (lstnumber.-8.7)] >> endobj 1303 0 obj << /Names [(lstnumber.-8.8) 385 0 R (lstnumber.-8.9) 386 0 R (lstnumber.-9.1) 404 0 R (lstnumber.-9.10) 413 0 R (lstnumber.-9.11) 414 0 R (lstnumber.-9.12) 415 0 R] /Limits [(lstnumber.-8.8) (lstnumber.-9.12)] >> endobj 1304 0 obj << /Names [(lstnumber.-9.13) 416 0 R (lstnumber.-9.14) 417 0 R (lstnumber.-9.15) 418 0 R (lstnumber.-9.16) 427 0 R (lstnumber.-9.17) 428 0 R (lstnumber.-9.18) 429 0 R] /Limits [(lstnumber.-9.13) (lstnumber.-9.18)] >> endobj 1305 0 obj << /Names [(lstnumber.-9.19) 430 0 R (lstnumber.-9.2) 405 0 R (lstnumber.-9.20) 431 0 R (lstnumber.-9.21) 432 0 R (lstnumber.-9.22) 433 0 R (lstnumber.-9.23) 434 0 R] /Limits [(lstnumber.-9.19) (lstnumber.-9.23)] >> endobj 1306 0 obj << /Names [(lstnumber.-9.24) 435 0 R (lstnumber.-9.25) 436 0 R (lstnumber.-9.3) 406 0 R (lstnumber.-9.4) 407 0 R (lstnumber.-9.5) 408 0 R (lstnumber.-9.6) 409 0 R] /Limits [(lstnumber.-9.24) (lstnumber.-9.6)] >> endobj 1307 0 obj << /Names [(lstnumber.-9.7) 410 0 R (lstnumber.-9.8) 411 0 R (lstnumber.-9.9) 412 0 R (noweb.NW2i2LRH-10PmPC-1) 776 0 R (noweb.NW2i2LRH-1KNHNk-1) 359 0 R (noweb.NW2i2LRH-1ep4xx-1) 437 0 R] /Limits [(lstnumber.-9.7) (noweb.NW2i2LRH-1ep4xx-1)] >> endobj 1308 0 obj << /Names [(noweb.NW2i2LRH-1fgq0k-1) 330 0 R (noweb.NW2i2LRH-1oKncb-1) 357 0 R (noweb.NW2i2LRH-1sNM4H-1) 775 0 R (noweb.NW2i2LRH-1uBDkH-1) 275 0 R (noweb.NW2i2LRH-27nDss-1) 345 0 R (noweb.NW2i2LRH-2o3xLS-1) 276 0 R] /Limits [(noweb.NW2i2LRH-1fgq0k-1) (noweb.NW2i2LRH-2o3xLS-1)] >> endobj 1309 0 obj << /Names [(noweb.NW2i2LRH-2yIl95-1) 355 0 R (noweb.NW2i2LRH-2zwz7b-1) 715 0 R (noweb.NW2i2LRH-3Gkbo6-1) 315 0 R (noweb.NW2i2LRH-3dLroe-1) 362 0 R (noweb.NW2i2LRH-4BH61R-1) 360 0 R (noweb.NW2i2LRH-4GHEiM-1) 754 0 R] /Limits [(noweb.NW2i2LRH-2yIl95-1) (noweb.NW2i2LRH-4GHEiM-1)] >> endobj 1310 0 obj << /Names [(noweb.NW2i2LRH-4Mxm99-1) 361 0 R (noweb.NW2i2LRH-4XnUPC-1) 222 0 R (noweb.NW2i2LRH-4ZZkMy-1) 356 0 R (noweb.NW2i2LRH-CpvDB-1) 322 0 R (noweb.NW2i2LRH-FUd2L-1) 358 0 R (noweb.NW2i2LRH-KwRQ2-1) 234 0 R] /Limits [(noweb.NW2i2LRH-4Mxm99-1) (noweb.NW2i2LRH-KwRQ2-1)] >> endobj 1311 0 obj << /Names [(noweb.NW2i2LRH-OTtxn-1) 277 0 R (page.1) 144 0 R (page.10) 558 0 R (page.11) 565 0 R (page.12) 571 0 R (page.13) 581 0 R] /Limits [(noweb.NW2i2LRH-OTtxn-1) (page.13)] >> endobj 1312 0 obj << /Names [(page.14) 639 0 R (page.15) 686 0 R (page.16) 729 0 R (page.17) 786 0 R (page.18) 825 0 R (page.19) 829 0 R] /Limits [(page.14) (page.19)] >> endobj 1313 0 obj << /Names [(page.2) 188 0 R (page.20) 834 0 R (page.21) 838 0 R (page.22) 842 0 R (page.23) 848 0 R (page.24) 901 0 R] /Limits [(page.2) (page.24)] >> endobj 1314 0 obj << /Names [(page.25) 923 0 R (page.26) 952 0 R (page.27) 1001 0 R (page.28) 1066 0 R (page.29) 1077 0 R (page.3) 200 0 R] /Limits [(page.25) (page.3)] >> endobj 1315 0 obj << /Names [(page.4) 220 0 R (page.5) 308 0 R (page.6) 372 0 R (page.7) 426 0 R (page.8) 462 0 R (page.9) 508 0 R] /Limits [(page.4) (page.9)] >> endobj 1316 0 obj << /Names [(section*.1) 189 0 R (section.1) 6 0 R (section.2) 10 0 R (section.3) 14 0 R (section.4) 34 0 R (section.5) 106 0 R] /Limits [(section*.1) (section.5)] >> endobj 1317 0 obj << /Names [(section.6) 122 0 R (subsection.3.1) 18 0 R (subsection.3.2) 22 0 R (subsection.3.3) 26 0 R (subsection.3.4) 30 0 R (subsection.4.1) 38 0 R] /Limits [(section.6) (subsection.4.1)] >> endobj 1318 0 obj << /Names [(subsection.4.2) 42 0 R (subsection.4.3) 46 0 R (subsection.4.4) 50 0 R (subsection.4.5) 54 0 R (subsection.4.6) 74 0 R (subsection.4.7) 90 0 R] /Limits [(subsection.4.2) (subsection.4.7)] >> endobj 1319 0 obj << /Names [(subsection.4.8) 98 0 R (subsection.5.1) 110 0 R (subsection.5.2) 114 0 R (subsection.5.3) 118 0 R (subsection.6.1) 126 0 R (subsection.6.2) 130 0 R] /Limits [(subsection.4.8) (subsection.6.2)] >> endobj 1320 0 obj << /Names [(subsection.6.3) 134 0 R (subsubsection.4.5.1) 58 0 R (subsubsection.4.5.2) 62 0 R (subsubsection.4.5.3) 66 0 R (subsubsection.4.5.4) 70 0 R (subsubsection.4.6.1) 78 0 R] /Limits [(subsection.6.3) (subsubsection.4.6.1)] >> endobj 1321 0 obj << /Names [(subsubsection.4.6.2) 82 0 R (subsubsection.4.6.3) 86 0 R (subsubsection.4.7.1) 94 0 R (subsubsection.4.8.1) 102 0 R] /Limits [(subsubsection.4.6.2) (subsubsection.4.8.1)] >> endobj 1322 0 obj << /Kids [1194 0 R 1195 0 R 1196 0 R 1197 0 R 1198 0 R 1199 0 R] /Limits [(Doc-Start) (lstnumber.-10.1)] >> endobj 1323 0 obj << /Kids [1200 0 R 1201 0 R 1202 0 R 1203 0 R 1204 0 R 1205 0 R] /Limits [(lstnumber.-10.10) (lstnumber.-11.29)] >> endobj 1324 0 obj << /Kids [1206 0 R 1207 0 R 1208 0 R 1209 0 R 1210 0 R 1211 0 R] /Limits [(lstnumber.-11.3) (lstnumber.-12.27)] >> endobj 1325 0 obj << /Kids [1212 0 R 1213 0 R 1214 0 R 1215 0 R 1216 0 R 1217 0 R] /Limits [(lstnumber.-12.28) (lstnumber.-13.23)] >> endobj 1326 0 obj << /Kids [1218 0 R 1219 0 R 1220 0 R 1221 0 R 1222 0 R 1223 0 R] /Limits [(lstnumber.-13.24) (lstnumber.-13.56)] >> endobj 1327 0 obj << /Kids [1224 0 R 1225 0 R 1226 0 R 1227 0 R 1228 0 R 1229 0 R] /Limits [(lstnumber.-13.57) (lstnumber.-13.89)] >> endobj 1328 0 obj << /Kids [1230 0 R 1231 0 R 1232 0 R 1233 0 R 1234 0 R 1235 0 R] /Limits [(lstnumber.-13.9) (lstnumber.-14.9)] >> endobj 1329 0 obj << /Kids [1236 0 R 1237 0 R 1238 0 R 1239 0 R 1240 0 R 1241 0 R] /Limits [(lstnumber.-15.1) (lstnumber.-17.10)] >> endobj 1330 0 obj << /Kids [1242 0 R 1243 0 R 1244 0 R 1245 0 R 1246 0 R 1247 0 R] /Limits [(lstnumber.-17.11) (lstnumber.-19.13)] >> endobj 1331 0 obj << /Kids [1248 0 R 1249 0 R 1250 0 R 1251 0 R 1252 0 R 1253 0 R] /Limits [(lstnumber.-19.14) (lstnumber.-19.46)] >> endobj 1332 0 obj << /Kids [1254 0 R 1255 0 R 1256 0 R 1257 0 R 1258 0 R 1259 0 R] /Limits [(lstnumber.-19.47) (lstnumber.-2.36)] >> endobj 1333 0 obj << /Kids [1260 0 R 1261 0 R 1262 0 R 1263 0 R 1264 0 R 1265 0 R] /Limits [(lstnumber.-2.37) (lstnumber.-20.28)] >> endobj 1334 0 obj << /Kids [1266 0 R 1267 0 R 1268 0 R 1269 0 R 1270 0 R 1271 0 R] /Limits [(lstnumber.-20.29) (lstnumber.-21.26)] >> endobj 1335 0 obj << /Kids [1272 0 R 1273 0 R 1274 0 R 1275 0 R 1276 0 R 1277 0 R] /Limits [(lstnumber.-21.27) (lstnumber.-21.59)] >> endobj 1336 0 obj << /Kids [1278 0 R 1279 0 R 1280 0 R 1281 0 R 1282 0 R 1283 0 R] /Limits [(lstnumber.-21.6) (lstnumber.-21.91)] >> endobj 1337 0 obj << /Kids [1284 0 R 1285 0 R 1286 0 R 1287 0 R 1288 0 R 1289 0 R] /Limits [(lstnumber.-21.92) (lstnumber.-22.34)] >> endobj 1338 0 obj << /Kids [1290 0 R 1291 0 R 1292 0 R 1293 0 R 1294 0 R 1295 0 R] /Limits [(lstnumber.-22.35) (lstnumber.-6.2)] >> endobj 1339 0 obj << /Kids [1296 0 R 1297 0 R 1298 0 R 1299 0 R 1300 0 R 1301 0 R] /Limits [(lstnumber.-6.3) (lstnumber.-8.24)] >> endobj 1340 0 obj << /Kids [1302 0 R 1303 0 R 1304 0 R 1305 0 R 1306 0 R 1307 0 R] /Limits [(lstnumber.-8.25) (noweb.NW2i2LRH-1ep4xx-1)] >> endobj 1341 0 obj << /Kids [1308 0 R 1309 0 R 1310 0 R 1311 0 R 1312 0 R 1313 0 R] /Limits [(noweb.NW2i2LRH-1fgq0k-1) (page.24)] >> endobj 1342 0 obj << /Kids [1314 0 R 1315 0 R 1316 0 R 1317 0 R 1318 0 R 1319 0 R] /Limits [(page.25) (subsection.6.2)] >> endobj 1343 0 obj << /Kids [1320 0 R 1321 0 R] /Limits [(subsection.6.3) (subsubsection.4.8.1)] >> endobj 1344 0 obj << /Kids [1322 0 R 1323 0 R 1324 0 R 1325 0 R 1326 0 R 1327 0 R] /Limits [(Doc-Start) (lstnumber.-13.89)] >> endobj 1345 0 obj << /Kids [1328 0 R 1329 0 R 1330 0 R 1331 0 R 1332 0 R 1333 0 R] /Limits [(lstnumber.-13.9) (lstnumber.-20.28)] >> endobj 1346 0 obj << /Kids [1334 0 R 1335 0 R 1336 0 R 1337 0 R 1338 0 R 1339 0 R] /Limits [(lstnumber.-20.29) (lstnumber.-8.24)] >> endobj 1347 0 obj << /Kids [1340 0 R 1341 0 R 1342 0 R 1343 0 R] /Limits [(lstnumber.-8.25) (subsubsection.4.8.1)] >> endobj 1348 0 obj << /Kids [1344 0 R 1345 0 R 1346 0 R 1347 0 R] /Limits [(Doc-Start) (subsubsection.4.8.1)] >> endobj 1349 0 obj << /Dests 1348 0 R >> endobj 1350 0 obj << /Type /Catalog /Pages 1192 0 R /Outlines 1193 0 R /Names 1349 0 R /PageMode/UseOutlines /OpenAction 137 0 R >> endobj 1351 0 obj << /Author()/Title()/Subject()/Creator(LaTeX with hyperref package)/Producer(pdfTeX-1.40.10)/Keywords() /CreationDate (D:20100723141812+01'00') /ModDate (D:20100723141812+01'00') /Trapped /False /PTEX.Fullbanner (This is pdfTeX, Version 3.1415926-1.40.10-2.2 (TeX Live 2009) kpathsea version 5.0.0) >> endobj xref 0 1352 0000000001 65535 f 0000000002 00000 f 0000000003 00000 f 0000000004 00000 f 0000000000 00000 f 0000000015 00000 n 0000016710 00000 n 0000507311 00000 n 0000000060 00000 n 0000000118 00000 n 0000022620 00000 n 0000507224 00000 n 0000000163 00000 n 0000000197 00000 n 0000056673 00000 n 0000507098 00000 n 0000000243 00000 n 0000000279 00000 n 0000059162 00000 n 0000507024 00000 n 0000000330 00000 n 0000000365 00000 n 0000067377 00000 n 0000506937 00000 n 0000000416 00000 n 0000000458 00000 n 0000071124 00000 n 0000506850 00000 n 0000000509 00000 n 0000000551 00000 n 0000074188 00000 n 0000506776 00000 n 0000000602 00000 n 0000000638 00000 n 0000078013 00000 n 0000506649 00000 n 0000000684 00000 n 0000000723 00000 n 0000078069 00000 n 0000506575 00000 n 0000000774 00000 n 0000000803 00000 n 0000083183 00000 n 0000506488 00000 n 0000000854 00000 n 0000000882 00000 n 0000094223 00000 n 0000506401 00000 n 0000000933 00000 n 0000000984 00000 n 0000099064 00000 n 0000506314 00000 n 0000001035 00000 n 0000001109 00000 n 0000099120 00000 n 0000506190 00000 n 0000001160 00000 n 0000001198 00000 n 0000107490 00000 n 0000506116 00000 n 0000001254 00000 n 0000001288 00000 n 0000107545 00000 n 0000506029 00000 n 0000001344 00000 n 0000001386 00000 n 0000112979 00000 n 0000505942 00000 n 0000001442 00000 n 0000001484 00000 n 0000114007 00000 n 0000505868 00000 n 0000001540 00000 n 0000001576 00000 n 0000118876 00000 n 0000505744 00000 n 0000001627 00000 n 0000001664 00000 n 0000118932 00000 n 0000505670 00000 n 0000001720 00000 n 0000001762 00000 n 0000118988 00000 n 0000505583 00000 n 0000001818 00000 n 0000001860 00000 n 0000128415 00000 n 0000505509 00000 n 0000001916 00000 n 0000001952 00000 n 0000132405 00000 n 0000505385 00000 n 0000002003 00000 n 0000002044 00000 n 0000132461 00000 n 0000505324 00000 n 0000002100 00000 n 0000002142 00000 n 0000132517 00000 n 0000505210 00000 n 0000002193 00000 n 0000002243 00000 n 0000134115 00000 n 0000505146 00000 n 0000002300 00000 n 0000002343 00000 n 0000138255 00000 n 0000505014 00000 n 0000002390 00000 n 0000002425 00000 n 0000138312 00000 n 0000504935 00000 n 0000002477 00000 n 0000002506 00000 n 0000146238 00000 n 0000504842 00000 n 0000002558 00000 n 0000002604 00000 n 0000146295 00000 n 0000504763 00000 n 0000002656 00000 n 0000002711 00000 n 0000155725 00000 n 0000504644 00000 n 0000002758 00000 n 0000002798 00000 n 0000155782 00000 n 0000504565 00000 n 0000002850 00000 n 0000002879 00000 n 0000173429 00000 n 0000504472 00000 n 0000002931 00000 n 0000002983 00000 n 0000173487 00000 n 0000504393 00000 n 0000003035 00000 n 0000003096 00000 n 0000004592 00000 n 0000004740 00000 n 0000004920 00000 n 0000005130 00000 n 0000005429 00000 n 0000003148 00000 n 0000005315 00000 n 0000005372 00000 n 0000502961 00000 n 0000501034 00000 n 0000502792 00000 n 0000500532 00000 n 0000503130 00000 n 0000503638 00000 n 0000006925 00000 n 0000007076 00000 n 0000007226 00000 n 0000007377 00000 n 0000007533 00000 n 0000007689 00000 n 0000007845 00000 n 0000008000 00000 n 0000008150 00000 n 0000008305 00000 n 0000008461 00000 n 0000008617 00000 n 0000008773 00000 n 0000008929 00000 n 0000009091 00000 n 0000009253 00000 n 0000009414 00000 n 0000009576 00000 n 0000009731 00000 n 0000009891 00000 n 0000010052 00000 n 0000010214 00000 n 0000010369 00000 n 0000010531 00000 n 0000010687 00000 n 0000010849 00000 n 0000011000 00000 n 0000011156 00000 n 0000011312 00000 n 0000011468 00000 n 0000011617 00000 n 0000011773 00000 n 0000011929 00000 n 0000012204 00000 n 0000006537 00000 n 0000005553 00000 n 0000012085 00000 n 0000012147 00000 n 0000500699 00000 n 0000015741 00000 n 0000015893 00000 n 0000016045 00000 n 0000016197 00000 n 0000016349 00000 n 0000016501 00000 n 0000016826 00000 n 0000015569 00000 n 0000012302 00000 n 0000016653 00000 n 0000501922 00000 n 0000502067 00000 n 0000501346 00000 n 0000502211 00000 n 0000016765 00000 n 0000500098 00000 n 0000501491 00000 n 0000021073 00000 n 0000021237 00000 n 0000021402 00000 n 0000021567 00000 n 0000021733 00000 n 0000021899 00000 n 0000022064 00000 n 0000022227 00000 n 0000022392 00000 n 0000025830 00000 n 0000020877 00000 n 0000017002 00000 n 0000022558 00000 n 0000500866 00000 n 0000022676 00000 n 0000500243 00000 n 0000022736 00000 n 0000022793 00000 n 0000022853 00000 n 0000022914 00000 n 0000022973 00000 n 0000023034 00000 n 0000023095 00000 n 0000023156 00000 n 0000023217 00000 n 0000023278 00000 n 0000023339 00000 n 0000023400 00000 n 0000023457 00000 n 0000023518 00000 n 0000023579 00000 n 0000023640 00000 n 0000023701 00000 n 0000023762 00000 n 0000023823 00000 n 0000023884 00000 n 0000023945 00000 n 0000024006 00000 n 0000024067 00000 n 0000024128 00000 n 0000024189 00000 n 0000024250 00000 n 0000024311 00000 n 0000024372 00000 n 0000024433 00000 n 0000024494 00000 n 0000024555 00000 n 0000024616 00000 n 0000024677 00000 n 0000024738 00000 n 0000024799 00000 n 0000024860 00000 n 0000024921 00000 n 0000024982 00000 n 0000025042 00000 n 0000025103 00000 n 0000025163 00000 n 0000025224 00000 n 0000025284 00000 n 0000025345 00000 n 0000025406 00000 n 0000025467 00000 n 0000025528 00000 n 0000025589 00000 n 0000025650 00000 n 0000025710 00000 n 0000025770 00000 n 0000035513 00000 n 0000043880 00000 n 0000045519 00000 n 0000030457 00000 n 0000030623 00000 n 0000030787 00000 n 0000030953 00000 n 0000031117 00000 n 0000031282 00000 n 0000031448 00000 n 0000031614 00000 n 0000031779 00000 n 0000031945 00000 n 0000032111 00000 n 0000032277 00000 n 0000032443 00000 n 0000032607 00000 n 0000032773 00000 n 0000032939 00000 n 0000033104 00000 n 0000033270 00000 n 0000033435 00000 n 0000033598 00000 n 0000033762 00000 n 0000033927 00000 n 0000034091 00000 n 0000034256 00000 n 0000034422 00000 n 0000034586 00000 n 0000034751 00000 n 0000037807 00000 n 0000030117 00000 n 0000025980 00000 n 0000034917 00000 n 0000034974 00000 n 0000035035 00000 n 0000035096 00000 n 0000035156 00000 n 0000035217 00000 n 0000035277 00000 n 0000035338 00000 n 0000035397 00000 n 0000035454 00000 n 0000035574 00000 n 0000035631 00000 n 0000035692 00000 n 0000035753 00000 n 0000035813 00000 n 0000035874 00000 n 0000035931 00000 n 0000035992 00000 n 0000036053 00000 n 0000036114 00000 n 0000036175 00000 n 0000036236 00000 n 0000036297 00000 n 0000036358 00000 n 0000036415 00000 n 0000036476 00000 n 0000036537 00000 n 0000036598 00000 n 0000036659 00000 n 0000036719 00000 n 0000036780 00000 n 0000036840 00000 n 0000036901 00000 n 0000036961 00000 n 0000037022 00000 n 0000037083 00000 n 0000037144 00000 n 0000037205 00000 n 0000037266 00000 n 0000037322 00000 n 0000037383 00000 n 0000037444 00000 n 0000037505 00000 n 0000037566 00000 n 0000037627 00000 n 0000037687 00000 n 0000037747 00000 n 0000056729 00000 n 0000064823 00000 n 0000083239 00000 n 0000099176 00000 n 0000138369 00000 n 0000146351 00000 n 0000155838 00000 n 0000173545 00000 n 0000042589 00000 n 0000042751 00000 n 0000042917 00000 n 0000043083 00000 n 0000043246 00000 n 0000043410 00000 n 0000046549 00000 n 0000042417 00000 n 0000037931 00000 n 0000043575 00000 n 0000043637 00000 n 0000043698 00000 n 0000043759 00000 n 0000043819 00000 n 0000043941 00000 n 0000043998 00000 n 0000044059 00000 n 0000044120 00000 n 0000044181 00000 n 0000044242 00000 n 0000044303 00000 n 0000044363 00000 n 0000044424 00000 n 0000044483 00000 n 0000044544 00000 n 0000044604 00000 n 0000044665 00000 n 0000044726 00000 n 0000044787 00000 n 0000044848 00000 n 0000044909 00000 n 0000044970 00000 n 0000045031 00000 n 0000045092 00000 n 0000045153 00000 n 0000045214 00000 n 0000045275 00000 n 0000045336 00000 n 0000045397 00000 n 0000045458 00000 n 0000045580 00000 n 0000045637 00000 n 0000045698 00000 n 0000045759 00000 n 0000045820 00000 n 0000045881 00000 n 0000045942 00000 n 0000046003 00000 n 0000046064 00000 n 0000046125 00000 n 0000046186 00000 n 0000046247 00000 n 0000046308 00000 n 0000046369 00000 n 0000046429 00000 n 0000046489 00000 n 0000049270 00000 n 0000049436 00000 n 0000049600 00000 n 0000049766 00000 n 0000051628 00000 n 0000049114 00000 n 0000046673 00000 n 0000049931 00000 n 0000049988 00000 n 0000050049 00000 n 0000050110 00000 n 0000050170 00000 n 0000050231 00000 n 0000050291 00000 n 0000050352 00000 n 0000050412 00000 n 0000050473 00000 n 0000050534 00000 n 0000050595 00000 n 0000050656 00000 n 0000050713 00000 n 0000050774 00000 n 0000050835 00000 n 0000050896 00000 n 0000050957 00000 n 0000051018 00000 n 0000051079 00000 n 0000051140 00000 n 0000051201 00000 n 0000051262 00000 n 0000051323 00000 n 0000051384 00000 n 0000051445 00000 n 0000051506 00000 n 0000051567 00000 n 0000503756 00000 n 0000055965 00000 n 0000056129 00000 n 0000056294 00000 n 0000056459 00000 n 0000059218 00000 n 0000055809 00000 n 0000051739 00000 n 0000056611 00000 n 0000056790 00000 n 0000056847 00000 n 0000056908 00000 n 0000056969 00000 n 0000057030 00000 n 0000057091 00000 n 0000057152 00000 n 0000057213 00000 n 0000057274 00000 n 0000057335 00000 n 0000057396 00000 n 0000057457 00000 n 0000057518 00000 n 0000057579 00000 n 0000057640 00000 n 0000057701 00000 n 0000057762 00000 n 0000057823 00000 n 0000057884 00000 n 0000057945 00000 n 0000058006 00000 n 0000058066 00000 n 0000058127 00000 n 0000058187 00000 n 0000058248 00000 n 0000058308 00000 n 0000058369 00000 n 0000058430 00000 n 0000058491 00000 n 0000058552 00000 n 0000058613 00000 n 0000058674 00000 n 0000058735 00000 n 0000058796 00000 n 0000058857 00000 n 0000058918 00000 n 0000058979 00000 n 0000059040 00000 n 0000059101 00000 n 0000064272 00000 n 0000064434 00000 n 0000064600 00000 n 0000067493 00000 n 0000064124 00000 n 0000059355 00000 n 0000064766 00000 n 0000064884 00000 n 0000064941 00000 n 0000065002 00000 n 0000065063 00000 n 0000065124 00000 n 0000065185 00000 n 0000065246 00000 n 0000065307 00000 n 0000065368 00000 n 0000065428 00000 n 0000065489 00000 n 0000065549 00000 n 0000065610 00000 n 0000065670 00000 n 0000065731 00000 n 0000065792 00000 n 0000065853 00000 n 0000065914 00000 n 0000065975 00000 n 0000066036 00000 n 0000066097 00000 n 0000066158 00000 n 0000066219 00000 n 0000066280 00000 n 0000066341 00000 n 0000066402 00000 n 0000066463 00000 n 0000066524 00000 n 0000066585 00000 n 0000066646 00000 n 0000066707 00000 n 0000066768 00000 n 0000066829 00000 n 0000066890 00000 n 0000066951 00000 n 0000067012 00000 n 0000067073 00000 n 0000067134 00000 n 0000067195 00000 n 0000067256 00000 n 0000067316 00000 n 0000067433 00000 n 0000500388 00000 n 0000501200 00000 n 0000070759 00000 n 0000070910 00000 n 0000071180 00000 n 0000070619 00000 n 0000067734 00000 n 0000071062 00000 n 0000073382 00000 n 0000073588 00000 n 0000073979 00000 n 0000074244 00000 n 0000073226 00000 n 0000071382 00000 n 0000074131 00000 n 0000502502 00000 n 0000073784 00000 n 0000078187 00000 n 0000077839 00000 n 0000074459 00000 n 0000077951 00000 n 0000502647 00000 n 0000501779 00000 n 0000501635 00000 n 0000078125 00000 n 0000082796 00000 n 0000082960 00000 n 0000086459 00000 n 0000082656 00000 n 0000078430 00000 n 0000083126 00000 n 0000083300 00000 n 0000083357 00000 n 0000083418 00000 n 0000083479 00000 n 0000083540 00000 n 0000083601 00000 n 0000083662 00000 n 0000083723 00000 n 0000083784 00000 n 0000083845 00000 n 0000083906 00000 n 0000083967 00000 n 0000084028 00000 n 0000084089 00000 n 0000084150 00000 n 0000084211 00000 n 0000084272 00000 n 0000084333 00000 n 0000084394 00000 n 0000084455 00000 n 0000084516 00000 n 0000084577 00000 n 0000084638 00000 n 0000084699 00000 n 0000084760 00000 n 0000084817 00000 n 0000084878 00000 n 0000084938 00000 n 0000084999 00000 n 0000085059 00000 n 0000085120 00000 n 0000085181 00000 n 0000085242 00000 n 0000085303 00000 n 0000085364 00000 n 0000085425 00000 n 0000085486 00000 n 0000085547 00000 n 0000085608 00000 n 0000085669 00000 n 0000085730 00000 n 0000085791 00000 n 0000085852 00000 n 0000085913 00000 n 0000085974 00000 n 0000086035 00000 n 0000086096 00000 n 0000086157 00000 n 0000086218 00000 n 0000086279 00000 n 0000086339 00000 n 0000086399 00000 n 0000503874 00000 n 0000091683 00000 n 0000094279 00000 n 0000091551 00000 n 0000086583 00000 n 0000091849 00000 n 0000091911 00000 n 0000091972 00000 n 0000092033 00000 n 0000092093 00000 n 0000092154 00000 n 0000092214 00000 n 0000092275 00000 n 0000092335 00000 n 0000092396 00000 n 0000092457 00000 n 0000092518 00000 n 0000092579 00000 n 0000092640 00000 n 0000092701 00000 n 0000092762 00000 n 0000092823 00000 n 0000092884 00000 n 0000092945 00000 n 0000093006 00000 n 0000093067 00000 n 0000093128 00000 n 0000093189 00000 n 0000093250 00000 n 0000093311 00000 n 0000093372 00000 n 0000093433 00000 n 0000093494 00000 n 0000093555 00000 n 0000093616 00000 n 0000093677 00000 n 0000093738 00000 n 0000093799 00000 n 0000093859 00000 n 0000093920 00000 n 0000093980 00000 n 0000094041 00000 n 0000094101 00000 n 0000094162 00000 n 0000502357 00000 n 0000098349 00000 n 0000098512 00000 n 0000098677 00000 n 0000098843 00000 n 0000100936 00000 n 0000098193 00000 n 0000094574 00000 n 0000099007 00000 n 0000099237 00000 n 0000099294 00000 n 0000099355 00000 n 0000099416 00000 n 0000099477 00000 n 0000099538 00000 n 0000099599 00000 n 0000099660 00000 n 0000099721 00000 n 0000099782 00000 n 0000099843 00000 n 0000099904 00000 n 0000099965 00000 n 0000100026 00000 n 0000100087 00000 n 0000100148 00000 n 0000100209 00000 n 0000100270 00000 n 0000100331 00000 n 0000100392 00000 n 0000100453 00000 n 0000100514 00000 n 0000100575 00000 n 0000100636 00000 n 0000100697 00000 n 0000100757 00000 n 0000100818 00000 n 0000100876 00000 n 0000106460 00000 n 0000104271 00000 n 0000104437 00000 n 0000104603 00000 n 0000104769 00000 n 0000104932 00000 n 0000105098 00000 n 0000105263 00000 n 0000105420 00000 n 0000105582 00000 n 0000105748 00000 n 0000108870 00000 n 0000104067 00000 n 0000101073 00000 n 0000105913 00000 n 0000105975 00000 n 0000106036 00000 n 0000106097 00000 n 0000106157 00000 n 0000106218 00000 n 0000106278 00000 n 0000106339 00000 n 0000106399 00000 n 0000106521 00000 n 0000106578 00000 n 0000106639 00000 n 0000106700 00000 n 0000106761 00000 n 0000106822 00000 n 0000106883 00000 n 0000106944 00000 n 0000107005 00000 n 0000107066 00000 n 0000107127 00000 n 0000107188 00000 n 0000107248 00000 n 0000107309 00000 n 0000107369 00000 n 0000107430 00000 n 0000107599 00000 n 0000107660 00000 n 0000107717 00000 n 0000107778 00000 n 0000107839 00000 n 0000107900 00000 n 0000107961 00000 n 0000108022 00000 n 0000108083 00000 n 0000108144 00000 n 0000108203 00000 n 0000108264 00000 n 0000108324 00000 n 0000108385 00000 n 0000108445 00000 n 0000108506 00000 n 0000108567 00000 n 0000108628 00000 n 0000108689 00000 n 0000108750 00000 n 0000108810 00000 n 0000113035 00000 n 0000114063 00000 n 0000111934 00000 n 0000112097 00000 n 0000112262 00000 n 0000112427 00000 n 0000112591 00000 n 0000112757 00000 n 0000115335 00000 n 0000111762 00000 n 0000109020 00000 n 0000112922 00000 n 0000113096 00000 n 0000113153 00000 n 0000113214 00000 n 0000113275 00000 n 0000113336 00000 n 0000113397 00000 n 0000113458 00000 n 0000113519 00000 n 0000113580 00000 n 0000113641 00000 n 0000113702 00000 n 0000113763 00000 n 0000113824 00000 n 0000113885 00000 n 0000113946 00000 n 0000114123 00000 n 0000114180 00000 n 0000114240 00000 n 0000114301 00000 n 0000114361 00000 n 0000114422 00000 n 0000114481 00000 n 0000114542 00000 n 0000114603 00000 n 0000114664 00000 n 0000114725 00000 n 0000114786 00000 n 0000114847 00000 n 0000114908 00000 n 0000114969 00000 n 0000115030 00000 n 0000115091 00000 n 0000115152 00000 n 0000115213 00000 n 0000115274 00000 n 0000119044 00000 n 0000118702 00000 n 0000115459 00000 n 0000118814 00000 n 0000124448 00000 n 0000124279 00000 n 0000119259 00000 n 0000124391 00000 n 0000503992 00000 n 0000128471 00000 n 0000128241 00000 n 0000124650 00000 n 0000128353 00000 n 0000132573 00000 n 0000132236 00000 n 0000128660 00000 n 0000132348 00000 n 0000134172 00000 n 0000133941 00000 n 0000132762 00000 n 0000134053 00000 n 0000137870 00000 n 0000138034 00000 n 0000141166 00000 n 0000137730 00000 n 0000134348 00000 n 0000138198 00000 n 0000138430 00000 n 0000138487 00000 n 0000138548 00000 n 0000138609 00000 n 0000138670 00000 n 0000138731 00000 n 0000138792 00000 n 0000138853 00000 n 0000138914 00000 n 0000138975 00000 n 0000139036 00000 n 0000139097 00000 n 0000139158 00000 n 0000139219 00000 n 0000139280 00000 n 0000139341 00000 n 0000139402 00000 n 0000139463 00000 n 0000139524 00000 n 0000139585 00000 n 0000139645 00000 n 0000139706 00000 n 0000139766 00000 n 0000139827 00000 n 0000139887 00000 n 0000139948 00000 n 0000140009 00000 n 0000140070 00000 n 0000140131 00000 n 0000140192 00000 n 0000140253 00000 n 0000140314 00000 n 0000140375 00000 n 0000140436 00000 n 0000140497 00000 n 0000140558 00000 n 0000140619 00000 n 0000140680 00000 n 0000140741 00000 n 0000140802 00000 n 0000140863 00000 n 0000140924 00000 n 0000140985 00000 n 0000141046 00000 n 0000141106 00000 n 0000145333 00000 n 0000145499 00000 n 0000145663 00000 n 0000145829 00000 n 0000147321 00000 n 0000145177 00000 n 0000141396 00000 n 0000145994 00000 n 0000146056 00000 n 0000146117 00000 n 0000146178 00000 n 0000146412 00000 n 0000146469 00000 n 0000146530 00000 n 0000146591 00000 n 0000146652 00000 n 0000146713 00000 n 0000146774 00000 n 0000146835 00000 n 0000146896 00000 n 0000146957 00000 n 0000147018 00000 n 0000147079 00000 n 0000147140 00000 n 0000147201 00000 n 0000147261 00000 n 0000150683 00000 n 0000149175 00000 n 0000147549 00000 n 0000149287 00000 n 0000149344 00000 n 0000149405 00000 n 0000149466 00000 n 0000149526 00000 n 0000149587 00000 n 0000149647 00000 n 0000149708 00000 n 0000149768 00000 n 0000149829 00000 n 0000149890 00000 n 0000149951 00000 n 0000150012 00000 n 0000150073 00000 n 0000150134 00000 n 0000150195 00000 n 0000150256 00000 n 0000150317 00000 n 0000150378 00000 n 0000150439 00000 n 0000150500 00000 n 0000150561 00000 n 0000150622 00000 n 0000504110 00000 n 0000155333 00000 n 0000155497 00000 n 0000158573 00000 n 0000155193 00000 n 0000150768 00000 n 0000155663 00000 n 0000155899 00000 n 0000155956 00000 n 0000156017 00000 n 0000156078 00000 n 0000156139 00000 n 0000156200 00000 n 0000156261 00000 n 0000156322 00000 n 0000156383 00000 n 0000156444 00000 n 0000156505 00000 n 0000156566 00000 n 0000156627 00000 n 0000156688 00000 n 0000156749 00000 n 0000156810 00000 n 0000156871 00000 n 0000156932 00000 n 0000156993 00000 n 0000157054 00000 n 0000157115 00000 n 0000157176 00000 n 0000157237 00000 n 0000157298 00000 n 0000157359 00000 n 0000157420 00000 n 0000157480 00000 n 0000157541 00000 n 0000157601 00000 n 0000157662 00000 n 0000157722 00000 n 0000157783 00000 n 0000157844 00000 n 0000157905 00000 n 0000157966 00000 n 0000158027 00000 n 0000158088 00000 n 0000158149 00000 n 0000158210 00000 n 0000158271 00000 n 0000158332 00000 n 0000158393 00000 n 0000158453 00000 n 0000158513 00000 n 0000162333 00000 n 0000166142 00000 n 0000162200 00000 n 0000158803 00000 n 0000162497 00000 n 0000162555 00000 n 0000162617 00000 n 0000162679 00000 n 0000162740 00000 n 0000162802 00000 n 0000162863 00000 n 0000162925 00000 n 0000162986 00000 n 0000163048 00000 n 0000163110 00000 n 0000163172 00000 n 0000163234 00000 n 0000163296 00000 n 0000163358 00000 n 0000163420 00000 n 0000163482 00000 n 0000163544 00000 n 0000163606 00000 n 0000163668 00000 n 0000163730 00000 n 0000163792 00000 n 0000163854 00000 n 0000163916 00000 n 0000163978 00000 n 0000164040 00000 n 0000164102 00000 n 0000164164 00000 n 0000164226 00000 n 0000164288 00000 n 0000164350 00000 n 0000164412 00000 n 0000164474 00000 n 0000164535 00000 n 0000164597 00000 n 0000164658 00000 n 0000164720 00000 n 0000164781 00000 n 0000164843 00000 n 0000164905 00000 n 0000164967 00000 n 0000165029 00000 n 0000165091 00000 n 0000165153 00000 n 0000165215 00000 n 0000165277 00000 n 0000165339 00000 n 0000165401 00000 n 0000165463 00000 n 0000165525 00000 n 0000165587 00000 n 0000165649 00000 n 0000165711 00000 n 0000165773 00000 n 0000165835 00000 n 0000165897 00000 n 0000165959 00000 n 0000166020 00000 n 0000166081 00000 n 0000173035 00000 n 0000173200 00000 n 0000177863 00000 n 0000173916 00000 n 0000172890 00000 n 0000166253 00000 n 0000173365 00000 n 0000503469 00000 n 0000503300 00000 n 0000173607 00000 n 0000173666 00000 n 0000173729 00000 n 0000173792 00000 n 0000173854 00000 n 0000180855 00000 n 0000177727 00000 n 0000174242 00000 n 0000178030 00000 n 0000178089 00000 n 0000178152 00000 n 0000178215 00000 n 0000178277 00000 n 0000178340 00000 n 0000178402 00000 n 0000178465 00000 n 0000178527 00000 n 0000178590 00000 n 0000178653 00000 n 0000178716 00000 n 0000178779 00000 n 0000178842 00000 n 0000178905 00000 n 0000178968 00000 n 0000179031 00000 n 0000179094 00000 n 0000179157 00000 n 0000179220 00000 n 0000179283 00000 n 0000179346 00000 n 0000179409 00000 n 0000179472 00000 n 0000179535 00000 n 0000179598 00000 n 0000179661 00000 n 0000179724 00000 n 0000179787 00000 n 0000179850 00000 n 0000179913 00000 n 0000179976 00000 n 0000180039 00000 n 0000180101 00000 n 0000180164 00000 n 0000180226 00000 n 0000180289 00000 n 0000180351 00000 n 0000180414 00000 n 0000180477 00000 n 0000180540 00000 n 0000180603 00000 n 0000180666 00000 n 0000180729 00000 n 0000180792 00000 n 0000500027 00000 n 0000180967 00000 n 0000180993 00000 n 0000181019 00000 n 0000181226 00000 n 0000181420 00000 n 0000181617 00000 n 0000181743 00000 n 0000181915 00000 n 0000182312 00000 n 0000183000 00000 n 0000499567 00000 n 0000183643 00000 n 0000183723 00000 n 0000183871 00000 n 0000184573 00000 n 0000184779 00000 n 0000185095 00000 n 0000185770 00000 n 0000186439 00000 n 0000499196 00000 n 0000186981 00000 n 0000187459 00000 n 0000188037 00000 n 0000188391 00000 n 0000188471 00000 n 0000189375 00000 n 0000199762 00000 n 0000200400 00000 n 0000209721 00000 n 0000210111 00000 n 0000218113 00000 n 0000218411 00000 n 0000227209 00000 n 0000227588 00000 n 0000235509 00000 n 0000235798 00000 n 0000243458 00000 n 0000243714 00000 n 0000251602 00000 n 0000251842 00000 n 0000259739 00000 n 0000259977 00000 n 0000277735 00000 n 0000278132 00000 n 0000293239 00000 n 0000293582 00000 n 0000297615 00000 n 0000297861 00000 n 0000300614 00000 n 0000300835 00000 n 0000307990 00000 n 0000308232 00000 n 0000311373 00000 n 0000311598 00000 n 0000339162 00000 n 0000339525 00000 n 0000369222 00000 n 0000369718 00000 n 0000396011 00000 n 0000396348 00000 n 0000398054 00000 n 0000398282 00000 n 0000439064 00000 n 0000439677 00000 n 0000475382 00000 n 0000475844 00000 n 0000492503 00000 n 0000492730 00000 n 0000498965 00000 n 0000504222 00000 n 0000504317 00000 n 0000507384 00000 n 0000507594 00000 n 0000507821 00000 n 0000508047 00000 n 0000508269 00000 n 0000508493 00000 n 0000508722 00000 n 0000508965 00000 n 0000509200 00000 n 0000509439 00000 n 0000509682 00000 n 0000509924 00000 n 0000510167 00000 n 0000510408 00000 n 0000510648 00000 n 0000510885 00000 n 0000511128 00000 n 0000511370 00000 n 0000511613 00000 n 0000511855 00000 n 0000512098 00000 n 0000512336 00000 n 0000512575 00000 n 0000512818 00000 n 0000513060 00000 n 0000513303 00000 n 0000513544 00000 n 0000513785 00000 n 0000514028 00000 n 0000514270 00000 n 0000514513 00000 n 0000514755 00000 n 0000514998 00000 n 0000515240 00000 n 0000515483 00000 n 0000515725 00000 n 0000515968 00000 n 0000516208 00000 n 0000516451 00000 n 0000516692 00000 n 0000516933 00000 n 0000517176 00000 n 0000517411 00000 n 0000517652 00000 n 0000517889 00000 n 0000518127 00000 n 0000518370 00000 n 0000518608 00000 n 0000518845 00000 n 0000519085 00000 n 0000519320 00000 n 0000519561 00000 n 0000519802 00000 n 0000520037 00000 n 0000520277 00000 n 0000520520 00000 n 0000520761 00000 n 0000521002 00000 n 0000521245 00000 n 0000521487 00000 n 0000521730 00000 n 0000521967 00000 n 0000522200 00000 n 0000522433 00000 n 0000522668 00000 n 0000522902 00000 n 0000523137 00000 n 0000523371 00000 n 0000523603 00000 n 0000523837 00000 n 0000524080 00000 n 0000524322 00000 n 0000524565 00000 n 0000524807 00000 n 0000525046 00000 n 0000525287 00000 n 0000525533 00000 n 0000525775 00000 n 0000526018 00000 n 0000526260 00000 n 0000526503 00000 n 0000526745 00000 n 0000526993 00000 n 0000527240 00000 n 0000527489 00000 n 0000527735 00000 n 0000527981 00000 n 0000528230 00000 n 0000528477 00000 n 0000528726 00000 n 0000528973 00000 n 0000529222 00000 n 0000529470 00000 n 0000529719 00000 n 0000529967 00000 n 0000530216 00000 n 0000530463 00000 n 0000530710 00000 n 0000530959 00000 n 0000531204 00000 n 0000531436 00000 n 0000531663 00000 n 0000531894 00000 n 0000532121 00000 n 0000532351 00000 n 0000532578 00000 n 0000532810 00000 n 0000533045 00000 n 0000533278 00000 n 0000533507 00000 n 0000533738 00000 n 0000533973 00000 n 0000534207 00000 n 0000534437 00000 n 0000534700 00000 n 0000534999 00000 n 0000535298 00000 n 0000535593 00000 n 0000535793 00000 n 0000535964 00000 n 0000536133 00000 n 0000536305 00000 n 0000536468 00000 n 0000536652 00000 n 0000536864 00000 n 0000537085 00000 n 0000537311 00000 n 0000537563 00000 n 0000537767 00000 n 0000537893 00000 n 0000538027 00000 n 0000538160 00000 n 0000538294 00000 n 0000538428 00000 n 0000538562 00000 n 0000538694 00000 n 0000538827 00000 n 0000538961 00000 n 0000539095 00000 n 0000539228 00000 n 0000539361 00000 n 0000539495 00000 n 0000539629 00000 n 0000539762 00000 n 0000539896 00000 n 0000540028 00000 n 0000540159 00000 n 0000540299 00000 n 0000540431 00000 n 0000540554 00000 n 0000540653 00000 n 0000540780 00000 n 0000540913 00000 n 0000541046 00000 n 0000541164 00000 n 0000541276 00000 n 0000541316 00000 n 0000541448 00000 n trailer << /Size 1352 /Root 1350 0 R /Info 1351 0 R /ID [ ] >> startxref 541768 %%EOF gpml-matlab-v3.0-2010-07-23/doc/f6.gif000644 000765 000024 00000074517 11422240326 016632 0ustar00carlstaff000000 000000 GIF89aá‹÷ý )))85,,583338882821,0 w??DDDWWWIKUfff`eovvwPjl[\[˜¸¦..Øÿøÿ8ù44â8ÿXÿwýu.©OO¹qqôDDøVVã^^ÿww÷ww÷ffèaiÂ7^ÿxªjÿ˜ÿ¸þ”7¿ÿ?ÿØùúÜú)ÿ—Eþ™Qÿ VÿªFÿªfÿ³wþ¬rð’m‘ÿq–ÿyšÿ|þp¨þXÿôDÿõUóçLÿöeÿ÷iÿ÷vÿø{éÐh°¨C—¶22¬DD®UUµJJ±ee½hh¾bg¤TgÖÿö6ÿ79ÿ(/ý99ÅVÿ7Cÿvÿ5RêvvÄ~~ÇmmÊDPÿIUÿVYÿDLüU`ÿYdÿUpôfpÿitÿwwÿryüciúFFÑGWª©i«?ÿ¿_ÿŸ—ÿ¶ÿ4¢ÿw€ÿ|…üqóD®ÿG¦üTµÿY·ÿJ°ÿe½ÿi¾þz´æP”ßÖÿùú)ýØvÄÿ~ÇÿmÜùQÿô\ÿõJúòaÿõkþöqÿö{ÿ÷`à÷^á×I˜§µµµ¢œ¢Ì“’ú‘‘ÿºˆþ»»û±±û«ªÚ°³þ‚”ꎤÿŠ­ÿ–³üš»ÿ§¦å­•Ñ®ÿƘüÅÿù™ÿù”ÿø‰ÿÙºüͬÄÿ³Ëÿ¼Áÿ¯ÿûºþú«ÓÍŸŠŠÍ™™Ó””ѓ̪ªÚ°°Ü‰‘ÿ˜šÿ‰‹ü™ ÿ¿î»»â«±ÿ»»ÿ±¶ý¥¨÷…®ËˆÌÿ™Óÿ™Éü¼Áÿ·ÀøªÚÿ²×ú€ÿ÷‰ÿø“ÿø»âÿ ÿùªÿú²ÿúžØÐÌÌÌÝÝÝßÑÑÍÅÒüËËþÝÝîÕÐÓÿÆ×ÿËÛÿÑÏýÂÿãÌÿüÌÿìÝâÿÚÿýÝâàÄËËéÌÐÿÝÝ÷ÉÊÿÌéÿÃåÿÝàÿÑãøÁÿûÝñÿÐÿüîîîààéþîîöîíéÿãÿöîÿþîñÿìîîùîðÿîøÿàÿýøÿöðÿþÿÿÿàßî˱Ô!ù,á‹þû H° Áƒ*\È°¡Ã‡#JœH±¢Å‹3jÜȱ£Ç CŠI²¤É“(Sª\ɲ¥Ë—0cÊœI³¦Í›8sêÜɳ§ÏŸ@ƒ J´¨Ñ£H“*]Ê´©Ó§P£JJµªÕ«X³jÝʵ«×¯`Ã>}'€Ø³hÓŽEÀ¬Ú·pãú”PÖ­Ü»xó²¤ÀR[½€ îè—B¿¿ƒ+æ ïÁƒÓN{€àÁ;‹Ó<ˆx±çÏR)QÚ7 A—!Œ¦ø®@€Ë‡í ÜÚ%¾ÚùáŽhY ‚›ûQ >ñ€Èœe÷{`€ó ¢'@ ztÔ£[×^]z‚íÛ¿þw—ž]<ùQ¾ ?ž»ûëí”?ß;xûíÍë׿{þìD0Ýt ü—ß}ÓP{,°@vøÞ€ÞWÅt ÂÇuþ÷@“\x†&PA P€ªxà‹¶çß{3î—!ÞµX Š @õøœƒ P€ƒGY€¹Ü@ÄõÁ4ï8F‘PBPg%¥NùQHEηxT‰– …‘æFø ÁQ;~ôQÌmµ UЖ-ÐÀ „ÚÀ†":@—°ý6<<@‰iù%ŸÉA=ÝãÅ1!æSPÜóQ%•0ô„©ñ‹FþzÒÎEí rHX‹6T¢†öÚk®¢õSA”\lDûðÕÖ²Ìj?›ò„˜"‰º“9O̪ ù"ÅFddë¬QP `åÊP¾¶[¨º½ã%È™EIX‘%ŽõëX[ŽE¶”æ„OcŠô¬N´›à#—ÛYG”Ç°í†8Ì!8ˆÊ$ þßã ¶üÆ“'T„ ëkˆ _²‡šà£r{ ² \ÀŠWT¬ÈEòð$>„ B¨rDÁ>1aDŽÁˆ¢|+üYLÔ3€J$STˆ s¨Ã,^ñ‹&É_Lö×’šµ«z‰!‚3‰ !‰ ɇÏT²Hè &æ@Ãî8’8ÄZ ¥±èE™ò%¹È…K°¶ZH•+eB ‘càÁ!’<\JdÙ&¶kÐÙIò˜=bñ˜~ÄáeF–ð" «S£O´æHH"$—59„QòD4pÒ$ÄD(ùÈGR²$ÍT‰. Ù)Ðò'FŒñÒŽ=èr&ÉþP\J„Q†o3P[ è¯(Ђ.3&éD /0ÖNm %žÑ‚ ‚Ž=8ã&í€#IÔÁ†4Xœí¢@ *Ò?šòt(áLX’{@5aiB¤É^¨r!BËKd*_’ä•{XK*¨”t G è9I’Ð’ü‚¡"|‰B͇p!õÔ©DØfD#ì`‡n6⇒”ƒ Á€I8bÔ‘"Õ Z\êHš:’³ve„SŠ?Þ¹^ñš{ئCÜ F<â+;ž±ŒG4¢ÍpGF|’]<2&k5H JR.S r ]CrÖVÅ¥y .Ú{\5!ùøþÃ=Âg8v’];–¡ˆe\D£)GnŠÙjÇ èq—[ÐІä”&)mM* _H5'Ô=HR«ßEÒžÑÇ#ÑŒ±B$ŠˆÆB²ÛànÄ i˜hLΦ07¹ÈÝ"~ó«ç‚º$‘nMxÙ‘3ÁC'.HU¢Ù]“c a‡"¡‹ðƒ¨0Bì^Œ48Š4¡oBÚª\ƒæ´ÌD)i‘È•†î(Nxäkð.Äx„V+âŽE4C"¶È1\V“̤Š&Nî~•šâè²x+³ iQnæ`¸A!~#l«áäƒê…H ߻Ӫ5P÷(þ~•|\ÿ~d´¹Ç“µ"3¦„¡b !Ë }&`Hø¡ˆ+]Á‡ÈXäΕøÄ&frLü‘§é„§™ky‚i‚@t!n±Aˆ1惸ƒ¼#Á±A0MY‹œ5x7!j15@ëåÖúÖ¸îo“Ebi­¸¦L¡°¢Ž×d‰iÆ"Ì+Æ>¤Õ²™gbd‚,€Öȵu¶³mk]#TÅp"òUöW§´Ž!3Fˆ3ÊzM£ä‚nH²ÐiS»sµÖ/¶µ}ëƒÂÎYCZµ‚-<3… NÈ1®|ŠºèÍíCÂAq{P$Šp»+²†6ú¤Ú¹6®þ¹ÍíûºÙ#”ÞHöNäLB ŠËû!…˜'dZ{ˆ$Þ G4d®E(~àƒ ½èAÅ(¶‘füX ìÇYžOê*ß¹Þ7®OÞ‘”g„ïI‚#r&4%Áæ²A¼{$x:Bö±QxéFÏûÑ}Àw£‚ áG#ÂKtÌ[!ø0Õsrˆ<ë§5×9ðB+·HäS¶Ërtí ‡ð3A0ïƒØã?ß} tW„Ãâ Ç6D‘z£‡" YFéÕ]ê‡LËŸ=i|?컬¿ÖÇ—ü®9’‹°c%ÊQ¡ZCÒ=ê²äÔ/H8FþáƒÔ+]°gÈì»ÿWÜØç AÄE#Òë ŸøH~äÏ€åkä©[ ƒÚ•r烴 +×pRWèehó EçžÐ ˜á | áw§‡ðp7'n@ñ~¹V|¸¶“·^Gv%61„<5Bñ¡^p’dM¶uûÀ}A¡Àt±­`t Ð€[‚èí'²†7 Xk#¨•çåPöæ÷ 󰂃ý_MÑ p­P{··á@tCˆ]¥I˜g°æ~òQ¸‡È§|ߦž’…M1Bæpþ(©t e†1 zž  1~—‹i jC!| °HŠxŠñç€ò‡ñ{YAˆT!Ü5ÿ§ŸWÄ€qƒá D÷ Ý@t>h›È{DH!|@Š´æŒ©øŒQHŠS˜%k 9D1sÑ7MQI¾ÀW‰hwˆlÁ‹±ÈzóÐâ °ñø¢ž@„‘†Ôu³•Ö“VWC©x|¨èŒémUò 0@“!)#q‹çcgg’sHK«‚)ÀÐé‘ZP^à‘nð=B³‹iHÜP{Áˆõ€ H@þD€ð(1Iª ±D× [²T‡ Á ·”"æ„YJ‹²¢A³¦r””nôV”WaFé³Ð…ázVc]&Ï€~á Að?x 2™ ß°°€C á |‡{ñ »GB)Yà) C€wg~¡ËÀRµ³epMÑ„1Š÷Ù¢ø)Z2,!w2J&̩䈮„P0‹åà ýaAh浓ø–Xt©¡ F †¬g´iIÞã¢è¢-ºŠ I/œÑ•àÀ!™‘éb°Oq âHDÙ6w½·‡¤ÝÇ–ñ C ‘ QZþóà@@šHt‚xƶ[Ê¥öY¦‘ÄU\ˆ.¨rf΀hñÏŽíH1yž­I@|*ÇHî©@úô ©õ)©! ñ Χì•©V!1³›ôvª£:è§*â@€Ê¬° ™°;€;p­ôÀõ0ª*¡:Ë°H}Àl!‘PA•a¤È-J¯óŠ˜“v£q‚MÁKR` XñÞ²óTD*³Iª|·¤á§NŠà 8€:` «à ± «°<¦ —HPA E(«p‰«Ÿ4¯Ðþ².»öê²ôʼz¦ú:Š–›IA¬X¡yZIÅG§FCç Ûn9®;`ÝÚ¦;ð­AÛÛ`ê[á ½‡³ÎÙ¹Ú²0˲2›¶5[’ÙOVqN†WA «n0pÑuÈæÅ}@´1‰ô°7p ± 6`¸ñ G{©G KÐaGRQ¶3›¶1³k{%7ÛîZïà^iŸ˜VºM‰u– K¨‰Ûº {DPF¹a6WºpSñ~ikŠ™«¹/ ¾Ú>Rá?#TºU1¶°ø8aI¤¹Ìþ0é£@F€ á 6À A6P»q;ꉮìà?¶i^ ³D¡¢šÅ`¿ö{¿÷ë²°¿÷Û¹p6p˼Tp 9¡·£§cõ†A@²ѽ±­V;¾6€±‘—è:–µjºðWP¼ûË¿-Û¿2«Â,pv0 /ÅP`êÀN~yOx:žðq(¯[à+¾á 8`ûÀÁ? —ý¨Ñ;v0U¼(¬Âù›ÂøK³öwÙh<[™<ÂDØ ¹^Û´×Ä¡ ª`¦€Ì;¾1Nœ’¼{l÷»þŸ¤¿[<È'œÂ«h¼Ì9§TA aœñ”]Âpyý I ̺>PŒ‚ Ç: 6`J,Fçă3ɶdô{dZŒÅ¬,ÈöÛÂQ‚ ÆcL])¼D4vë'[óГ}z»Á;À~‚ œ  ½‘ ™Px<Fwªîà[°¡Mð+¿óùIšûÍüû¿]Lº\šÖæc ± —Ú΄ÐÆåúÃ~Ê´Ë›€ &`) )&P€ ñ°ñ(Ð ý= ÐEwŽ»ÛKÁÎWñ~ú«Âý›Ñ-È° MEH±X·\1 ›f®•ºñþe„?­C žŠ›˜0Л@Ó%ÀÌ qÄ‘­‰‰>0Ì qé QT|÷à ¾@ úw ¾à£/aÑÑù{ÑÝÑ‘P;³§WqºkçOCKº'ö`¡A“QÇ QÓ)PÐ !m²±ÉéèÓi9ÛÙ^ëðbº°¦ a®ôîD JíÔ] PÀ…—™¢(ÈV=Õ}RehS±£G¸´°i 4Ë£GYîÀÇó<°0¨a ÇŒ×ÀÏpíšP ±²+õðÉh)Áü€~‘±À ýà`qaðnòØAØR³€Ü$‘ÊÖFÙþ¿Õ‘N·šÙnjºÈºË½Ü‹`;¾­¶-®± v}úL×Á )€«`ÇI¼ÒÚ‹°rWÊ¿=(ô{…¸Sp`,ÙSÅ­ÑÖÝÍÔ$üf_¨7¼ÀIèt³¡–± Ä0Ýà;Ç!ül% Á <0·=C¢Ž q ”ßѨ áËýß/7€³¾Ñ п?ÞA~ÕöÚRq‘`ÖQÆ!+)¨­Ð}~‚ o<ä½×P×p×€ ±±Q‘zz™¨ù042Ø‘yP=·à8N=îã”ä ÎÐE}Ñçþ{f¦M!lQ3§¬¨ Ú«Û@½¤áZ7`¾ÑÚìm&ðå« Ä‰A@×<½ð3ùí)­^mïàN)¯a@äC.ä®äpç qJ–:ïp]Q‘N¡äR“ÝX¡ÛYÚÁª‘­‘å¯})€ ßÜ+Út ’õÛwðÛ»° À€Ü —i]q¾lþO ¡êÒÝêÕ½k”9=pªÁð× G/.Û@tz ¥A1p±½ì\þÞÁ DܱиjìÛÛŒ Àð | í»P¿ îQ ç&‡XÒ!!Õ Àê¯îê"/þë áAÓ3cëä!^á õ ÄãJ»% ðAðAÛmÜš©g‰ú`–üñÀòIšm,E’îÒÝ¿‡üÔ£ÀRµ¬©™9ÈKèAaÒ^t°— !;`^ò åáó‘Þý@Ì+ù´áU¿>ÀÉ Ñô'!°õ ±!?øB>ø†ø&ŸLcäP a Áö÷åL~ K2/4¡‹nå6âô`[Þö÷ÄÜž¨ÅP È –›|¸Ð=ò…ûCNø?‰vƒ6ç|J>ƒž·…žaá‹tþ_ìÕ*Q[˜àåáöÑÉ {öCߣͻ¹ó Z ±åöñQ¯û¹øëÿû!TR1BÎ}¾Ðõ¡ô|†ùÛyw _A¤ßÀ~Þp$ˆÍB† >x Ó>„ÞvDò­ß¶  (BŠ6°]„fÖ@$âJ–-Þ{âÏå̆ f‘SçÎ@øì0€æP¢'p)šTéR„0É1…UjC)2ÞCð „Ø ÔäÅ~H`!Ä®¡ lSbâÄp‡7‚CŽò!j #‚€ÚÄg"Ì{n—šƒb˜¡M— ~þôù(ÏœB+@åòæ¥P|¹þ¤ÆYôRRVæùÅÐ/A~‹z Òj`=W]dÈ)…hñÚ KU?{A|„ë÷¬™H®ýBã9ÊÓÑ4i…á¼à&å:»ïün¹ºK ÆŸfÖÌwèÝC CÝá¾3 óíA¸ìk?m>|Ø Tº†â)AÑJ`È"‚ˆ¨§ŸW‚b Eúú«ŸwÔÉ*=ZÞkIŠÏ.Cà¦î¾ãŽ§îı! ÌkñyÆ3Rœ Ìc °Ž¢ÔÂG!„ˆR‘Çëúq]l#þÙF?ô3Ck ÂF0r .~¯*É3Q^PA1Þ¬;/ pGí0ÄpË;Þ(€;ö @0=þøŽ³˜}î²~˜¡s|‚cgŸÁ¦’ã‰õ8,Åý¦~ }s÷eîá„•tÉÎÄ@HòPmÄùZXŽ š(œzEbà£×cû`pÓíñÄÒªìGtÍÅð ÆÙuZUù×A>v÷”s¾(eÚ ÓjˆÀþ!R?ý±¸Á¾Æõ{ «Ð{C<l”Øà@ D °@3è ¸Šy<‘Ò*³tqâ„É‹×â45BðíÌ«ÒúaçD€»O*«")°±WÃ…= „À°{À”ÀñP€ïsB¨ ¯•ÀƒÔ@ˆ‘`ˆcÁ5Ù²ßX ‚{Øâ'L°½2Ô„U°ÖB$Ø9(‡}PàŠ6>¡ŠMs·{¸¶¤Ä~0#7(Ä^|C¬ŽDô•ëhÄ©`•†ø„hh ˆÝ ‚}C¤†ø¡‡pÀ„¶hØùK¸hˆƒÓF£"°nà-âÓ9ÌAˆr <M;Fu±¼msÀ$DtÀ`þ¼ ¤œµv‹ªwˆFl YDˆtØ‚ KÄ‹e ¼»–%ãC4$:†X…HÁH:`" ÙwpGä#y¤Ç~°Ç†0HñjBš€¬wéG_‚@ƒ«øˆf8¿}ð Áj% OºIÑy$U ÄK(zˆ”ÉAá:‡yl|¡œ£0PHè» _ÔɹœË}”Ka b¼x`»{d §X‰\p;®½@·ÚB.tˆLП¤¸Ê¢ãP›}@q‚áà†¶WhfpNéÀ:ì]`H„[ðËþJ¥ÈÅ]ÜÇ”K»¤K¼ ½´)(ÊÔìž'ÈÍ~(‡2hg ›`†[ëÿéª~X?‡ ý ³¢ˆ}Øp?…Ê"0 éˆm¯¸A‡Þ!ˆK3—' ÅFô…òœ‰Ö,Âõ\ÏׄMÚä Û|“ ÓM¢LˆpIûPÊ~Ш “AˆT ‡¨?—ÀHPPPPXLhИ@Þ{PÉýH¥\ƒ6`ˆ}¸OÝ$‡[$ŠÖ”Mö”M»„ÏÍðÉ€¡êB̹(lŠcˆ>X†X}ð ÿÁDŒð‡`…ØÛD%#"(ÓE2¸#0Ê|þ…Ò ‚Wˆ‡g¸·~ @¢{0]hr`¾–¨Ñ¨ =¢Ð8ˆØE¹dÓØ„Í|ÑËO¹°úŠ”èÍÀ@J-Mâ4Îõñ„}SLˆ(Å–ð/‚ð°h´Ù8‚!8‹pÈÐz*¾e¨¹û@„–,ƒ=½…1±S‚À¸Ë6mÓðESEÕ8UŒ9­‘SúT— …E\‰ŒÚÓbxH‚pÿäny…†è‰LpRžã„SÐà‘ Ð!}À ÿ…}ˆFgÏø»iˆ08S;EÑ?TT-U0UR…U]³U““:}Õ•¸‡)¸ü<ʆ€6†x¢)¼È”Iþùª•`…éÄ–UØUøWˆ`…FƒÎT@ÇÀ§‚Åk„eH™CàÒÀ‡5µl×w³Q=Us%W¹4Utu‹Ý•Õ¬Ž™|0‰”¨…zˆü–á´ãsŽþÉ5„°¤™°ˆZ‚¥‹Cý9…Bf Ø`Ÿq@±Üv؃ÇóM2Ø2„ðÞ$ –å¬nµI-WqYrýZ“ ”½=:¶¬u 1XD‚ˆ\0…ÜÒ†pã|¼hŸ†Ш Úè# ø ‡P ÒV؇|H Â>À³6H|/(¬¡P[ÂM¥¸IÀ [R-WR%[7R×0A'þÝdÛ•`Bñ[‡Xå#ˆzeG8¿ÿŒ4‡0‹¢8ÐE0à˘ÂÕ!Þ„Iƒ0ÑÄ4ˆ¼—€‚oíX•Õ/ Y‘%Ù»LÂѵ‘ç½Ó}·­ÜÕEˆß”DՀǮ$—퇆ۡ‡@øa MØ" Tpnˆ ™VAé‡C؃OÉÖ=¥êX‚^pý¢`àvàvàÐ V}úôKïe:!@ñ^@ƒñ;½¨mˆGp6¶Mp˜ð$QÊà' À¤¨b<ò…Bž‰4 'vä&þÚ`¨ `÷Õ´¾Â+-Ðä6ÐZ O.Ób`àbZƒ-Ð…—D„?\w`„ÙmƒÄÛÌ•°L^[‰oàJfáM`…)‹xq‡Gh„Æûß>h¶å6pˆ0Õ"~ˆðaÍ›pà‘­f%n`‘{“J>s€ÞFô^™qKˆ@ r0ƒ]pk•Z@iÉÕÿñºc‰¤³e„úþE‚EˆpøJ±x…è|„•|6îŃn\pE†f‚¿é=†äfbrÕæ0áæñè4;=]g®Ü‡ø3P˜†È‡C8É͇F˜ÆY¶`ã–HºX€…ÜR…£5#H‚Ƈö؇n€uÔŠàbÞ~°Ö>5/P_‚Ø"1xˆvèon‰Öd`h`«–è«žd¦0Û7©É‚ Uh†F^ðtnˆµB„ˆ†E8>sh7pƒRà­Oмrƒ+nˆzPᾎá†d˜R*…ŠÑ‡eh„¢F‡>8i¦F/PHgîM¥ÐÚ¨z表¥ˆÎê°j¬ÆjI†âþ€¹dS 7€‚1`KÆî‡vðƒÆF=nˆpðÿ)…ƹ\šR y…;Ò‡EPÿÝŠ•‘°2`uÀ`Zp´» 𺇧ÆrÍ¥æÍæìªÖî­^ŠŒµÓYp‚#fˆŒEhñ-é¶WfPcÆ·üÕ×V¸â´b m… 툠}è¢æ‡Cp\¦ö‚-0ƒ¼æd¼z'èK÷`mŠWˆ~—ìÖn ÏjîV ï Zpî{Äò”Y‡À‡<@ƒ/_ëøZ†EðßAzÖ¼+~Ì™ØmðÿøO‚h‡þž·?8q`ð^@¶Êm‘¯Àiþþ"ÏÎnÏþl ·ð¤ÀpÎ0û¾wéþ¤@vˆb؃Â\JE˜]âxÖ …Š)è²JŠ:nãx…z€[Ñq„÷òÂ4‡7Àzl_p?4-O $ÿ©f‰ÍUÓ ßl&ò¢èjùjÑ8dŸ…ñf u(ƒ°„[kþgˆ6(xƒ¡øu¥Ø÷¦¹l$>“c'öcord·êeŠf¿ %o„82(_㈧£XqñiÞ2ó¨t‰o±¡§ø¦žgPbÙ|¨øOæu>GM½&gOŠNÖ«Nþh'ô•hÍçì÷y0¡wHk£w€€ Æ(~Gç,nÕr|ÀL?=dz‹W„Œ¤‡u8AíDÝEû ´p…§ÜP„ÈUpY]bèƒ'ª§õ'ƒ|M}§y¨è÷¤ù÷ô¼‰ÎþyOöŸ®êJ#8 0ÎhôÙùͨl©¸Qô$að‚ª%ˆdøƒu~›þ‰¥ºphq`öú†@¤#8Å÷]…±y³j÷ø}@ìÆC -L©'&ˆÒ˜Q™‡0Ýû•˜ÉߘDö¿'x'üÀ§˘¥Èì~x ˆþêhü¹…ÇnѸÚÞıY³¨óÍ*þ|øn"ývˆ‹‚›Îi™†…oˆ¥ó¹¢•ãH{\n„ø…;Ðv`ȪuˆYˆ(äúlCð „mjicÆ̈æRLøŽ"0`7rìèñ£G\¸@VL@²€ Kh°¤1£¥K–K–8 €àƒƒ9 Ô’BJ0ô(R‚÷Â$mêT(-(÷:æB3a;@€Ú!Ôþ×È¿ƒÝ<ù(j[EqIˆQõ­#*T(6Îõ#Ȩ„Í…=Hìί¾ýÌm1H‘Ü“Y Ìø´¢ãÇ’›†ñÕÔ$H07s†y³ßƒwûú!@@çdŽES³îÇ´5ì¡R^o,‡&Eb}Ðu}ôHÁ}Ú<ý(;*ÅXH† ‰E’ž‘¹QŒ ÄB~‹š!ÌׇXB^dÒT óÄœ®ˆ Õ·¹Ê±±ÆÖ‘cÓOXù2IÍ÷·Ðo%>MBLÓO%FÕ×Ïj Ræ „Ýó-ñR†{µó]wÌ,âAö´rW@ŒbOB±Ñ–8H!&+¬ (,QVþ5bÓ;Š€xP2{ðv:fàÑaùE…%I|°Í!laL•féÇß~þõóÎ p`?§Mƒ@˜-‚T å„›‘™9N ´Ñ=i™Ð!{puP3Œ°\+@”õC+£¤â¨ÔÓÔ4IìpPd®áÃ(ùÊB€’Ph”SÐAø@!Å•N2ÖÏ”HͧƒO8U@~\òç%¬5˜«GRðÚÚ-O¨YQ9dŒŠ:$ƒ;pG6wýJ¢Õ£ÊI¼õ˜*°PªÑ<>à%Ø3Žt×Dz¥¦ÛA†ùòÄïyŒzföÐBñýºoE²â—™­]òKTþ›'…ÁOÕâ+G¹”OBºiJ?@2±6@\úÃ+ŠöËŠÎ¥–D‹e´Y@ Í#X ‰P±ìµÑN¸Ñ†¾ª ŒýœsNÂADRYz¤Y -%ÍÒÒájð®B»¯Ô ÑÂðFk°{wà #?Ί´Õ” H”ÌÚeÔŠŸ¤p;Š ë'B¼¤‘êAfP¸Q†HÅsÕBû‹e­M×°´âI;þôÀlÞÏ-MÆId³qt¿$äÌËåóÈ2šîóŠux¡Øϵ>ÊÑ*±¯b +Þ€De´Ë0eI³‚F=NÜ:{ ßÏ{ÚçÄ°=9þØà=o”³G˜OîÂM•°ãJÓÐ=gó+yáþ`ífóII¡XmÆòÈðÐ0 AÝHÊ<Õ³œÈ­²ƒ l ð v¬Ð vŽŽ¼#wºÛÅ'âŒKCŠL1þ |xA¹Gfñ¼Å8…p!GÒw=¡øƒ}ˆÉ¼Ç4~¯†âÛWÔ¤–½RD –£ˆ0¼1®í¡á‡# "²áÚ8¶¼Ez˜8È„í¢ M$Ää,p}H£,¯ë‡"€Cg!øp_B3¶ë)&ÔéÃ_¸°h Ù‡fø=Æy¯i7üU…F‹òþðLâˆ.Öð¹Ðõ£nƒê‡+ŠƒE¥¢Q9Å p` zPd‹ÁÄV±‘n!Äx9È2èGâ ¯^8B¢r%:’j(ì1CCÔãÃèÂ`­r-,ó/-Á¤i2t&Ôµ‚IÍ‘yT ¼&I„ÃkÆû16ëx9ýGÚ(Ž’#¦L èaRR„i…uVe¤Æ;H9¶€Ëƒ„hÔ#aSŽYGÁ L¡õ ÕŽb´Ž`q½(F1:EŽš:ä„^E’w„#ëJˆ>E1gA w©V=Ž@„µ”:ÈbB⃼!sù_EŠÀµPþ± ×úð§~pp ýpd-:’õ3"íˆCÊ˃ˆTJW­OžB«Ì\t5«EÉ«D Íç+Ü„”Y‘5ð¢;€0"Aê&Ëê”Èœ”âALaƒvn¤§==ˆtl@‘oá |Uf#–JIÎ9uBýÈV?2s°'˜Á,毲›¾ð# p\FQ{VµæŠ­B“ÂeƒfÙpЃDtÆAz{€âRcŽÀÑU$Ð#Ø0 j„ä*·Fý˜[?t@·§·¥ H3˜q»– ŽÊæªz”¬&³-lHëGî¥6µƒd-¬\4ÄpT¶à¥ÈZ¼!õãþœâ†&±ÓX a[ñv@Ohâ<åbBR0Pd}Á'ÜÐ~ØmmŠŠW¼‹¯xA¾ Rƪ¥È`q•¶ô¥3mi»IÇ×»‡"¤ÂŠ æÏyc‘g9g~<‚;û¨ŽD±}|cýþëqYÁy¤¹*Є ú Lh"%à„²‚`ôC±QE*~ R0™gĈ„îþ€[‚D2! Ž-I´PbÖ”÷#£^h _ÜZDÕ)åÈ*­i‹Î›Ò7š|…6!þ*XüV ‘¤‰åCƒÚ(ÄÕŠ³áŒôH0•L”@V®–›Üår Ø Yv?n¶ýç¥H g?’¡Ò~ØÆÛ%uÊÆšr?…ÏXU7k ú”woÄò¶´½.¡sšJù9ÔË«…qäÇà¦7)æ£~„£·>xb?ðd‚W•Á† J€‰éô£§O>@.òƒðàɨþÂhøð¬|ìHdâAò,”D›!#}w±d@ksyÙç©QºŒÉ ô ÏûçEÐÑã8iŽ¶·ÔW7é6õm$üÈ©‡­¡¼Sä*PÁŽèRv.ÎÅôcW¶F¤{´È ïG> qÝœš êHˆÚ¾ûÝņ‡Òc?½ãgî1<¯ˆÏ5-ôL_úÞAótg1…O¦«‘‚-šn†Q¦}Ð<ˆ´QPä¯ÖöÆ ¼ŽxT<õ™‹)¥Ó¨‡ ,0ƒ@DøC„,ø(ŒÆ#èA"„¨$Ä,ô‘G˜Z/)_xu„÷¡[B1ßc ÞPHÚÑôþ-^Ð=„DžÔ¼‹Ò‡?Ü©d^^)‚¦¨Ž, ÁWÄ ›8X)}YÆ1—¼^ !AüC0d‚ü¡BöƒÆ,œ;4‚AÐÝAƒòÄÕ·uV y I´ àˆ¡zIÜ÷ÁÆ£9_~ Î!ÒaäOÂTüÞ5Ä Æ „Ô´`~µ‹ VR8iƒu„Âh ‚œQ&è@ÄaC @HàëD $«ÃÚÄ ?(¢ H#¸ÃD¼ÞõÃ’d óÈ\Õ` ¿Ì‚\-“G@îbÀê@ õ!AL¤ÁÆ=Tm‘ÁÓþ|„S7A8"”é@&œ’ 0á%öÃ5\œ&rbF¬­%Á\¸’,ØÃ>­ÑʵA3bSžáÑbœÐ㯠 f/îãС ƒH A+^S-8Á->ÆÕÂ$ÚyW?ƒ Ø„E´ü@^ôC:„<ÉAă ¤€`ÀSE[—‘ä¥&ìÔC'Ü@ô@ô€ô€u¨ HB_üÁăô€[GðÍ!„ÂæX =òÊá8Åö|„ðã>nÔõ)Ová5-0dR¼Ã;Є_Gä×TµÔA0é#¬0èàE*@™ àA\C p#EèþŸþ™ä™@—™@JRJ,ë&¨€Æ,&Hƒ4hÂ&‚2ôÞñL`R%áQÉEÐQRÂJ2JÆóQ@%/úã‚Dxx1>UD-Ä£G¨Ã8#UD!-DÝP$5öƒ¤ÂAxC\&'”€ä;öá=D“°&Bìç©ŒÄG‹~ýSÔMæ¦[öæ¼Ô5þHƒ,)Aƒ†}—"D’ðÒ‰eædn„òÐ}ä£Sºg{Âçzæh|î(}Ò[áC¬AŠV†-Ä9ÜB@Á#‘D±´"-mZjJ7XG[¾å5(B¨@„z Ä„Þer„_&„³eoâÓìƒ50A¨\ï%„ÞyÄŠ¾(yžæF€`â}„.î¨ÊçæhÆö!Ä1”ð!žöƒ?ÜÂDÁDêD„Á-DÔ/˜”Ι;Bþ€E„¨ƒF—rÄ5”¤&°À&l‘l‚ýç˜J('fB6²NÛØCq´BÐUB œè(Â%èÂÔn®˜È&ESæ¢ïò®Wq/IPƒ€4¯ vÄ›ÁhðãAßRwIÝ Ìöæ8 Ç耮m)Ð"gGŒo?lÙ]C°IÇ ÄÄBôÜü€:òS?h?+BHV.kfýÌ{8qPð¢ÒÜÆ-HŒ¬ïþ®Ž:0ÿ®Ä4xq<@|ñ«Œ±Ä^ð‚Ëã1#Ex^ý#\ ¤€¸AŽôÀ ôÀþëé€)DÐÖjGx Tî—YèþíÔ\4®ÄèÕCÕú€+ä§à/A”C D-ŒAòeyaò{4ñòâãdä.B€(‹2ïÒó£,•´rP€Â2Ià/ äB-_ÏáËî×AXˆ œü@œÓ,Øè%Ä•©ÞF\‚Š‚ét¬0 ì`‚àÁ&ܺB øÓAxÁþ„9,eSüïÍe2Œ>ÅØ®mÙ.jïéѤ)u“rêdq„Pì„NôÄATðdäÂ]]¬Ðü”ß ,UÝpÂ(¼C ÐC–t% 6°*_–$Žþ@G“éש@ §À鉀 °Lö€ø¨£¹4r:H%IV&gEÄ€'M‹MãÑði¬‹™í5™kL+GÈ3P?ðP“r§Î)w„€%p”%@›tSàƒU£ÁmP5¯˜C@ò¿åÃ!B’ èôÖäOèðA­öý 'A¬uI¶u—Ù%ºéqrz€q·bi‚Á>äìupØhƒJƒ$†dN:¿ØxÎÜ¿-kÀóAhvê u?°µì˜ˆÉ«Pš<@¦z€HM4¥jK¦A´/du„”C@¦lÏàÖ‰>î6XÞ þØ*Äøîs³0Hè‹s{'Ä9ƒ üÃ-, Ã'Á'àC,õÆô>ƒ†õï̱!GLIbGDa°‡²¶me¾¬;¿3I @QA u“7°|ÿG€ôÀ TÂcÀiP„–Ä€³†:üBxž¨­¢LðS°ö 'E—sù–ƒFP€®ÏÉ;ü¯ûúAô:þW;°„°ĨÃ1¸A¤äÁð‚¢“C±IµÃ-ð‚ ¬.˜Ã±S9WaèÂAÈC<‚vMâŠê8È„Áƒ f°¢&¨@µ; üú°ûq´±Wƒ<èÀ_¯ƒžP›´Â:0BXƒ¹ë‚hÁÍhÁ8ÜŒ¢¿#;±D°"‰âÈ[»c|ü>„ºµ‹¼a»Aœ|¸¼¸ïÃÌ|ȃ<Àÿ{Ëë|Ï‹ûÎû<Î}?X‚=Í<%H€®cöF0€X€ ¬zQ»@QXÀg(µR3–ëzSÔzB€½S'Œ¢ëBàþk+¹)Dõ«´‡k mÈ!>wÐC8H ýê&*(ŠìÇ~ìÏwhÄ«SÄÖ؛ƭùoDV(½Ü 0,„ XGo“*–÷šo¤ؤrq"ˆ…—¡€$úŠ€MbB»^Ë' 3p?ÌÙÝÇÀ l˜è/_G°|,Žúï#EO¯—Sʾþ˾—È>T9@ôá]¿J&T¸a•N€ ¡ 1fÔ¸þ‘cG¶´ÇÙÂ|Ì–õkÓ¯•  ì 4¢J¡S M˜øˆ $C˜¢è§ƒ•ÀTHú $»Fû ˆç—B'oz‚4L#0®µh\©plV³gÑ~Ä%mÆ ´;—n]ºêõ{÷à‚iô³´àA´XZh ÀEÛF–Y$ v Ù1b'Ш>FíëWÏ*…7N1LIèd&€µ)ð‘~Ú~ÑÆ¿~ˆˆ ö¦Mñ6p¬h1ÞFWÚ`Á¼Ó²+ºW…Ö[ÞfËö¯ýÎlW¾ÜœkóçÖš…þ­GváÛ š#}´”ÌÄ´HLÈ'·Púyd³vöȧulLÈŸ'Þiã»''Û*­(IÕ¬F•Ó5¶Øƒ‹:Z •ÖZmU.[É”sÚ|€˜²È\þM#:û(FSŒÜi3[òá‡y²C ’çР¬Q¼±¡5 íù ”}–y¦Ÿ|ö¸¬4ŽIèž'ÊSÕ5SÑBu,%Zµ'7”uœ0l¯£¸l­uW‡æ5Yj$V ¨YÙ‚“uæ:ÕHŸGÑ'¡yBa1Ú¶tÃ…LÐ$¡ØZ4+y`S¨\z„2"–}@±Öžgª¤~rÉE!)p Ê}[íŠ!3X`V˜ãŽfñ%á ˆWˆÉ¦µ×­m*ÛIæã?ˆIPd’õÉX ¥|•„P!‚žš2Áhš7²™Ã…@Œ§™Åáï~ˆHhþ<™ÇE:$¸~~AC¡ZÆ°[ßR£^’¡°žÆèj¨YuõµŠbÀ¶,ûa±%†]!µuÌ™bFêãdäå(ÚGLÖ«Ÿp<±”p2)n}"¼ðëjÍÐkt¶áñ#L…Âm.÷­˜Ch$C¡[¤P¾ŸV_תtÕ§®:#'Ñ •þÞ3ò§})âÜábt8[ï6Æ¿Žäãw‡ØC†WîðØÈ&„KÝÔš: Í Ä:ëÌJ<åÙzÖ“#¨Mm,W‹¶Ö¶Åm{D1Þ( |4ÃŽhÆmbVB¼U=ˆ0=…°âÌÄ ¸fnÄ%àD4ûá0SC°+4WøF˜J¶+_¬7!ã—²Æû>®„÷Tû“ç-ÖSÚ-Ña<öqücÇwkýŒLþÐbäW@îX†"ükÈ…„C)B+z;#AÀݤ,œUófáÄD”S<’e°pCn~°=½‹$áÅbD".äH ÁŒ?þ-Dg»8•¨¿Êó¦U  =ƒi!GšÈ3òY˜²2ñƒ¶Š€„Y5¢PlrÕ5S[‰ O&d6p.F4Qîy¤£‘‡ R`3 ”eôÇ'¼Ô Ó9AwÎóö ¥0ËX#I e ½2h¯úÐZõw˸J-~ä“ö¶¤ °ì­Í·#"ùƒ¦«äH(âÍwBœÖ „z IÅo«ƒ÷­Éîä[âûÞ+]°KoÂ{Ö€ÄJVÄcd9_ÊnëýøÈc„=¸Aèéj‚ü[†9GMUøB®0×6鬉î“F1Ú Lô þî¡ã®‡ýØ;ò†eßží®* ÿ]C§t¯L®L¤Š#`p°0ø”æÌ( O2ÔmŽŽ#ìaL¯ñ– ò$ï¢b (2‚†Ð¯û˜­f6B¤.l!*σ*Ê ªe]Ü®AÒ!œúÁØâðOÂLT¦G®ÿîü!Nð#Tΰ!0Y$¢b@ök–aù˜Ïù ª°#Taú0£La–í´oóøšÆl2/¸‚¢2¤ ¬ä4átPú̳žÀk<‚`ªfã¶#îŒÃî"CIˆðÑ㈔۞00 ]…ÜèosÐ#a¢þ$âÍ%<#ÚJ ‚bÀò²Ïß8DÖ#¶§é$¬Õ˜ d¡È€-H!îÁ ¬ž%Âγí#\‡q@–‘%cnA äd ¢ %Ñ )QN†/! $]C!áÝ‚Ê(Ä@Á 3Õb!#(o°O ìF 0AÖ¤4‚KÇpTà¯V('VHDDD”àŽ!w1"!îAþÀ“x“œ€²>bTçª #š=0²JâñëNNhÇ# àŸ0ãdûÐ=œ‚Š#¶aÔ6È%ÝÊ¢¨Otà{°K€ÌêQÂòJüþXhGâ>ýì±ÕøMŒ ì–b*¥‚þž…<çIîa"o&9b#Q!®m1N­½ a !.ß2.á.çÒ.Q²Ll)šÅ%±0&7¬r)Bá˜p2Õ2‚v„.D´éÂ5O%XÀ8ÊÀ ¶C—ÒN¤2Áqèƒh¢*s£ô Ý"t0"ÝàÄr2^gãñ,›hmn"{¯#öAä’.}Ó.Óò²JÈœ Ê—¤0Cm0ƒ`¾#ÄÁ“ú³1r"ÍB2 Ãro#®N°éx›dÂ4$¢ gÜ¥€'þ!"²ÜÀ nAÍi‰€Á  ˆ¨loTHŽ6B× ’­J¨Q!vó.sAT8± NË<øAZ¼q#¸¡y8(Ë":‰@P0"bÀ:"R`à.OÂl7Â[ƳzÀ"’úARòfòö@ ŠÁàóõ¬¢>¢_§#[ @Ûâ?ÍãXÌ@¹­7ïR.4.‡“J" @œA'CŽ@­#PÆO°L#PÍC¢Õ «N#EQpM E1JÂJcF?ãZnôV‚‚iŠ±Æ@bãДäT! mÐÛ8‹’å7"ër.ƒ8ßrJY"t2&þT+?bvhç4t!rR#Ú‘ß “+DʯM€…\(‘R#Vƒ!2Aî<å¢lT¡t4àÏ- Æ ´¢##9NÎ+2µ`’U2œ@YT!ÀI¥uZç€R2°«BÑ"O¶”#BBá91âÞB¸x|ò:U@´S#°sU[H¯6 5a€Î Æ`©ªƒXûÓX{bYIEžÎA©„Õ+fó<àa œµI§5cåÒZÑÃ"ø¡þ\ã`ò#ì!>Ã&7"l#v ¹6Bé®A9OÔ;1â T ®Á S V_µiðþ°eÚ`û €´a;¢A®ÙÌÎ8Ž5½´”jbÍC ÑUž5! . !cÃÖÆ6. @ -M¬Ú‚AS=Â>BL‘ º,#Ê%58ÏÂ2â#9DM³¢M'/£ú!‡"Ä *&¡hó4>ûA (ru Ñ ô“ëô!ú°Pm¬_zpmi f-]…Q5"Zå2lU·l¥ud–c,`Z22ÜÁ›6µÊ„6eñúΕTâÖR`g"-·SpA‚2¯ N Pá"Ä*¯¥R·Wb²­iG…x™6~–a%C{‹Ô{Ïâ` ¦kâXWc{³þ8Ö<èá m¢ô–b­4B ®0¢ÕDÔ„-¸3à %°j‚û¡6z±4µø¢¾’»´Mt7BH¯×tz≌#ŠŠƒsug%Ä×,BòbIlQ8…ËvlÛ×<$`,ð‡>¢%ˆ)ú0òt#Fµo!ì#²áHàäA4Nõ#…;O×8E&éOìðÌHa“b,Q/˜Æ˜„{‘•Em–ô=}ɶV—lËØŒÕØàAiW´#XJ¢6 ŽàÝV!D_¶¸¯'ÞF`H ªáF º³#HD2Ë»Ìý~`ohK Zþ/“¨¸È!Œ±Ö?)ƒ;W#gom6ù‹5ÇLØžÖøŒS͸…]C%?¢v_AYj6âƒ~.#3]Sèó<â/:/¡[$ˆn(û! ×,jêЇ¢ ʼn jêân–UFå/8¸8B«kõ5ˆr°`0/ï 5cêàsGâ¸J¶X2bsjiÛ#ìº'b[Œ;‚* &•SYu'á·+ û±ûXþb¹–7âRÂ#X*²ox*+[¸L ;™û@7#´š;Sþ!6Áeq¨™ñèú¡v)9»‚SÖˆ‚˜€5b.a˜a½™â»­’Rú’'^‚ Ô¡H1NYÅ+£Öµ™\ƒÓ‹t£=§§}Á­èÂâ”# µówÞMÄQ½EéJ8;‘ÂV2 …Æ…x¢†b„ÎoúÁ–‚ÏutvÙ *Zl ½Lt*·œ¿þ]œËykž vœ!$½A‹þ‹þè«]2®ý½uÔ—" 3‚ b´[!Tã1"y2³"V¼X#Âèœòâ(¨27:(Z(™×7e‚ vû<¶ùìøÙuž!†Øg‹ùÇà~Œ áèÍxð‘žèŸ–>2š¾dt37Da#|F#®IÍ{˜!:!V€UõêòU‘Å…º[!ÐEî&ƒdx|ü*S`¬ã÷smŸhþýïAáÕmFà•nº}åU‚ zõ€J g”MB½òƒ\ñMG,u¤BbBu2E×EâA"ÄOI™rŠABöƒóôóˆqýdfP.¼x†ƒBÕGß~øhŽñIq‹‘B (˜à –¥¤Qå$Fñƒ+‰cDG«ì×^FísEÒIEÝB'ö£C“£”^A}¬WAqZ•A™ß‚*•ˆRÄdþP.ˆ`”‘2¸(Ej ‰;BE„=!‹†-M¤¢QmJ5š%š¨IA¥d ^!d>øàé3—ݨYA¼ìx9Uš*E„¢LÊ*«Å²ÊêRéBÅFÛÏgÔT€VJ9)” ^ÛÏ•@IÔ(>ˆ²ÜxYæQÅ7TÅy sQ'Mý»Ÿýxñþ€{бM)e(Áקp£@=:i&øíµâRäÎ2Ç ÄMGªSŒA¿k9=.¾î¿ Þ;Âý/Zý; þ–•!OAË›ZÕ¢uµ…èCcA±Ç­‡:4œƒ|±¿Äe b^h'>°RŒXÄr?¢;RD#òç R¢7µMqQU4È2Ø!”V˜§#ß@BGt0ª‡¤`eFahöA‚ïuDé3ÈVᲺô#F^Ú—Aˆ7_á¯?Ã2\"õÓgÍg‡Td||ÑKG6)ŠQDž™ÇLÈLr¡`PD„þVt$ ›ˆG•´%U)ëHJŽDŽÑF~ðGvXsv:qAnÁ´EM 5E!ÚD~Y¥} ê†FÌÖ€ ÙLf*Ô™v˜$¢*Y£È„ AB†BHEî¢Î’N,àhŒ,er'ɯèèUAÖðÙ3*ý” @gªa)Rq9@4Ÿ™P…BÍ¡‡‚h?¦”äPækbúæD<À ² J6F@B©–­ ¬ “AâŽäÌHH: ¢‹]$ 8œ?rÓ¶¢ 3"O'ò(¡5¡QkhÒŒJ. dF“W)²Uë,$;5Š)8z?2¡î~ðÚ 6ˆ#¢ZCäD'ÊxÖͼõø8¡‘0øši>vš›ý*”x.ÿK,ÄÚ‰°P@áºmf"] JH÷ÐuˆƒwÑE‘óBy^O AC²œÕ,., Zþˆúnâ¥â{àÐÂð׊ÍÕœb™³ØÅFzG._€àòç¼Aâ;‘ò@y"âàíDèÑzŠ p"vŸGŒþ}À]ï7îÇ¢áYghíÑ)Q4MÀß«†ÄÀYüh¿na‡|3§|X“P Q šW1x@)ÅT¢a$›PC7R96+µR ·}³×ÅPlq fm€ú'ƒñA ÄóKT_p|2§ƒ:(s¨Ó'a Ð 1†EáNuô54™ gaÊe˜ ƒvx·° .ÜðN¤i×~0x‘XF¸$þ3Ȇµ87h @s=ȃÇ÷@x—<—gFlB1_>p` ±`Áq%pnñáá¦Þ&GNˆR–4eT¶xPL0Ximȉga-¨€!‡9èƒ98‡txsð °O–°@ öA @a}( F‘CÌÑ-#«˜Pö÷Kg+?@X¶dÔE;á¾°‰•ÒoØÙ<3Z£¸ƒ¥8‡ÎÓ”@ û0 p D^µ!=FÕbØØ´}aƒ('Ç‘r‡š€~ñ vßàE2¡/À&c‡pVõ7š¨©þ³0 cŠý €¦HŽ¨¨fwU0ï˜*ïðŽ¨ññ(°@/9-ï“ò2-®ðúR“¡ ²`2i©Â\@Y;Y—ÀFf“?¹”!“ôÀ %°òUDÙÅB7FEYC@”aõ”!“ ?ÅÒV9 ög•Öà I‚–L”pÉ•NÙ”J—ty—v9—zy•rÙ—x¹—o ˜É—FI˜‚Y˜~i˜‹)—êˆuˆ ™u™y9™û` ð’'Ù Psx|qЙà<ûðqš1 ¬ÈGñ@€ûDË )D_4À¨!ëþ·6)0JnFW>é§9 Án¡ÁöXr¡.haÄG‘Õi”fi‰_ÀÉß ž¤ø"Yšy\쨚¬¨žëÉŠpû@ (î¹y7øyO¶[¡Òà~FaŒ£ñ%  h¹óFÇy^Ä)l§+#ƒr¡?ÀcB|há˜fžxÕƒàI‡!ÚwXèx•è9` 븕py1*£˜÷“–7Ÿ–‹×g®àö‹÷Z(¬D%Gz¤%`‚  û° væ„ä1±þÒˆ°‚9buAUÂ:`ûÙ©þßIŠÉ‘¤èƒ¿á|0›ç‹°¡¹$FÐXA>–6ÖM¡B¡!MÎÁ q£WÜ TᎰðkZábZæ€uf: %*¢á)ªoÆGöù}*F ?€Ü÷sÞ…‚b $5¨€/zœU¶¹Ï ‘±©œ:‘e ªM‚ŠËJ¢=ˆ‡Bñ—¦}CBáNðç1ª :P«F¢Íu°Ð~í”›[Î@ Q;lÆÚ‰šly_ÐÜY¯÷J¯ùúõú¬A£3:£¸¨ªÑ§úÉ‘¡AÛÚ­ü! 6°þ•ÁkSº¾–øi©lqoØè®ÑšV!×8fà žøš²úÊæ ”°;û–a¯AF¡Œa$™`´¶NC@‚¡YØZ€¥ýPeì2õ±HŸŠiòʲûzµöj¢ Ó§ýÐœgb±AAŽ ¬"h¦` [D0¨ñNUú@©•:‘Ö³€ QÛ‰R°·Ëfµõ:ªûŠ¯üzpG?oñ…b{QAA:JöÁ 8®ѨæÓt YY³èúø ƒ|«þбU«‚K¯…;¸ÜÙ¯±‘£;Çz±º›CиAþ± 1 ¡tô ¶;0*Áó²tšôºQ ¡± q þÔ®‚r§Ö9½G!0-g²‹#¢ª‹µ"úºª»æzÛ@_¾¸ ¡«¹h«Þ 8P«ý#G¹Ûn·RK¡ Ñ´q ¿Ñ ¦¤û“µAÌeòÚ ¼º œÀ ìÀß4;ù+å k9¥BA@Ë­‹Jà` 7`:à›¿Hô[‚›´H5ozôgT+À[ˆwÀ©ëÀ«ÛÀúÚÃ]«£F1?«Bñ DkôÐU6PÂ;Å™Å%|;À ¾ /¡ÂróNÏÂfô3þxÐ¥µ5œu£¯©›¯=|¯üÆ@ŒB\Í°‹•zAÁk[,à° Œ°€»A¾]òus34ÆöD¾p‘hœu/§r¹ Ç<¼Ã ÌI¾ ¡X.‹ÅK-©‚F°ÁQÄA`»À&gÁJö§+(!k¶·ð¥%»Dn Ç…ëÀ €ÁE%Á!O!T=š»[ãz(Il¼æzüJûÒ¨®FAÃliþ·u”œÀYðÍ¿Î ,ÇgAÇ×·´3–­L©Âˆ¢ JLìÅÁ(hˆ0a±®O»†Úli´`ÀÉŽâ<οœå|ç\þî`@áN¬|`‚ÊJ"ÈÏüæâtž ʲÓ¡;ð ÐcV¬“ÜËàÜàŒÒ ¬Ò)Ò mü—¤pñr+\¨­"G ßÅ£>p¿Ñ Ø6rò6Öe×Û{h\½^šO[—½QÉâœ+ýË0-Ðsü€}¡ÑC-©0HP´µ D@EK`GËÎ~º*Øý ŽÌ?h¦@!Ò‰çÍX Ç-íÒ] ÕQÁÐøûÊF1 1òÖ!®DÀŸ¡Ñ}D€ ©œÑ€úÖÆñéli Ù<ÒRÁ{>2²2èÍ,íÀ~ýÀ‡{P‰ˆM_‹ý%F@Fþ gm°€jýØår+o­_ÝÂ:²,ŸÚòa$<ƒòªÒW ÝÐýÍÒЬ}ùðÉQ31ÌË‘hC@H  °@‚õ ß ©pÊj­ |Lš¥Ÿ±+õÉ®­{Ü4e$’Ì×ZAÝ)MÝÏ-àƒ­PÌBáÐRq­ PÏV!°  F€.ÞãÝÛBQ,û ±GvÜ€P†cȪ$¸¼l,‡âA ,ƒSWÏ-ã.ÝW-ØÖ}Ï€ÏB©q!Öà çááÕ@a© Æ-žºßVÁrW±3mHÉ-=àÝŽª nËuQþ2ÁÝ > lG°·9žÛ-(LÝä H"å.à^ãUŽåB‘/ÐåQ³A ?~(ózaÞÑpØ!â@ÁæQAoÞºà3Íâ,§ujü{ÎYÐéwÝŸ^Ý\; µ¡h®A.ž€ÓU²GßQä@±¿ºÒÿûÏ>òÔÉÆr»¾3Œ'ƒú àéÐ-ê ~åÏ3¹„Øê!Ô 1bæë#ûí«þà áÕϾRÇPÆQÁHJ’×-Çrånb@²¦½D þéïnì£N000¬^ÍÙ §Ñ?ÀêÝöÐ NÆï­N5{þÁ€†RA´,ék¾àðú'¯Ç.êð^ãy.ÂA ÷­îpêg1¢p+€Ú ÙÜ@ò€ú Û~ôçèáÃ÷ð—f÷ÀÍœèÜW}ìî^삵±íhð1B_žÐ Ônû° ­àk·ÂôþFýw#àŽ¥}ó™P@ÚØóÈnöBq—ÒÄÝì!ÛàkôE_¡Ð Û€ò 1Ü õ%¯Øf¾— ;>µþ\;ƒÃ±±â˜†é kaàâã­…•_ì•ùŽùžŽ. .WB ˆ^ÙWaÚ ÒÞ5·¡àú£Ð £þàú¢P=,+? Ý`õÁ.Üo†þ0î_ ·0пDšoùšùÎßéhßââYŽ€õg±/ä%/r¨ôEæ¢pògÑ ÐÑÑ/ÁOóVAÅoüöñþÀÞ†÷ZýùÏù>ïéÒOâùöäëWÐàA„ >kDPáCˆÃ…ãÖª•+‹]…›Ñ£Ág>tvhd¿cÇNF$÷„ÜJ˜1eΤYóä½'÷lîä¹RW•“°`ÉB´h–£D‡-: gO¢ëóÔ ;Eì¬n=H a3‘'û8ü˜²'Η\½reÛ6áZ…hÝÎ¥ifAP¥þE•æeš×)Ý•QCĕߣGüÛ|gŸ#H+ÑZ LåY(³¸6^Üy'g…P|y&ýPÝ'\a½Ôõk×JG” a gl£e-Ó#h0÷ [,ØV)atöfÞü ”ZÎg»9–@õØÙ_ÿ•^°vBÄ /{D¶{ÂföÁ, ó0®¾œÜ:_Ÿî=(´ìw&ÓÏúH¡´‹¿Þyà¦9¨€®¶¾o ¶ØQ¤™ý âg‘ed˽˸šŠå24±&¹Nt‹^üÃkÀì®pjJ(ég*@ÁilsK0…ÐÙc®fyƾgqG¦ö@t‹(Âþ0Ç&Uœ­Ê‚öqÉJ·ÖÀ§Ÿ‚b ;,Èt­µ2YûëÎî2¨’‚¨Éq. JƺŠ<Ò9=iúü‘Ìš‹) H‹KD"ljC}J4 úï#"ˆÀ‰¡ÐÔ‹L.*p"Ùúy`=QbË}T¨ÊèzF‘eô™­Ošˆ1)&Aé2GŠ(h)ÑQg!Ø­ÖPGR1ÇTvYfÿJ€³Ô 2Px«ó!V‹æ‘E¢YŒŸe.TŒ¦vª’)×¹Ö»‡(òû•ØîÞÉ ^«ðAàI=:€Y~™ µÆç,82¢bðü¶ÈG¢w«mÑ !G꣙ŽÉe¶uþ ã–w)¦w.Zù)<àø¤úuYÍ‚ÞIÁ* € €®ƒ„4}jävžÉHid™%m$bˆ¡)6œó‡–BŸÖÑzúäžÌ¡BŠÁžI‹ƒòhµ—ûå®»k#"æçÙØ«G–a§hØy&Ü¥'~ªÜšŽÁÃ>rh Š'¤xb )hñÅ—)ͦéž»¶Ü&Ó¾®m~ß–.n„)dNŸh¢F`]fF–D‘ ™‰FÖ¶öÀ˜&ù¼_#Œ'´f<Ê0"Ÿ|ò²GžÅ Í7·É „Ô†h¥®°Â5+°Û~¨ïEwŽôˆ„Túþ˜IƦÞ7÷å–ÉŸeøâ×zøÉ+WÑâõC§rÎs#Áø^ó½¥|OFöéÙHÐññü¯ ¨‚H1ÌeZ‰‚6ñ.n1<(DÁqÃ^wn)4j‚Œ‰±&z/ÄÂöhñ5‡|ÉÇÖÇB*l'nâLîñ>ûPrús‹?jÑ.åÑ&ø0CB¬÷$P{ ¬!·—Cæìð#‡¸•ƒdºÜËŒ;!‡/æ׸Æ}Ì*”‰¡HŸ5b0DøR¹x&È>büˆÏ'Eó=¥?yäJçW¨)Ô/yÓð|1<úµK ³ˆ#ib/…\Q!þ$S)¹g&,„ê< ŒÉs#½ŠD ‡¤É"­K+ù~˳äòšw\2Ò ¿P ?‚=+$S™Ë¬!3qHªî°2&ù #-‰õ5ƒ„g+»@6=ùoæÔ ePœiÃdvq™8¤ªhÂgÜá•,RŸ²†oæˆx f9 ø=e6é´!{CH˜ä£‡°æÈŠF«ØRŸ5>ÊQ&d (A=ÊÎú 4&ÎØC1ÐGF·H”¢+…?1jÎt t35èl¤¹“CôÁžŽÂR>ü`R·dPEdé&JN,ªf¦M˜)MÛÙŸúawðJÆæÙU«ˆþÒE‹^J@¦2u©MéyDZ“vSE”BË8êlU®'§G2Š&S¬z%«šPÓÒœÕ&iÄNMDŒátæ{œkc¥3L»šs¯~UfÂÊײFÓnIk€š¡‡ ‚4ãR0WjZÞC‰ÖG PÙeN–²z,in Õbüá†eNnMζ”ccV.E‹+Á50V!­õÈdÛuŠµ¶žìVœÏ6t.í8Ä>;|:V¼ÌQÇj#r׃äÕ²~]¯e/{Ùév¦ºl± gÙ:—|XWíJÇû_À”áXA¯A à^÷²—½vo|3_¨»ê»þoOÐXÏÖG¨æp[ŽAÄ‘¸ ÎepXÜ^Ë:0·íM;Š‘Ó=¢ü•Ì‹ÿc#„¨sAê‰v¼Òÿ¯ _‰1õ5[·—²íU1] Lt¼»}Ð-‘ gd9˶:DŒc\ Þj˜È&³U~ц•ˆ¸Hr‰QLÙ@st›åR;´LFbdËaVQÊÊügžàÿçUM˜àfÚ½Lh2è hžü¢Ÿ¦4Lð \BȲŠFô¢SÌÀGŸ¶4í;j'jj³•Ã¼#aîÚ­èCÃÚÓþQ¨;›Uïç¸úÜ5ØÐ0à•´"64ƒcëbšÖ©ª4þWÂÛlh?$Ã…‰šÙ¼édwúÓ!µu´a2ho7»¼3Q3‰9mO#xÙlyr¸‚&º»Ùdvµ Í|úØúÆw¿ jǹ-¼8ÇPDa€X&F^›¿ùÍodÇzÝ\i·¼ @‹7[Ð5Q3Žýq}CßßJÅ3Þ\Ÿ¼Ãh(ÇYæo˜Ç<ßÿžÍ;„˜<ãàVùxwAí™(àå3/ö¾‰NsÒ¼ÀùAX|êÞà¡å&Bu>«î(UïÄM1€ÌGÞï€ ª¤îÍ›nL®+M0ò@uë“í–ç˜iÂpˆ0@ì^9ÉcBX‚é9cþwÖ›”Kgsßêá7׆ãì$¾]ç;ßÿ%(!ž‰àw6xƒìÃR†gì~žJ0^«Ž7Û™{_½WÞëŽb‰¥ódx@A@oD€ô½ÙŤë“úá›ÍgxJ| {¯‡ŠM#NHïSpæ÷)@@€`ã€øÇoþó“ýåW?ùÙŸþöÃ_þëG?ýã_þú»?ÿãÿýßÿÿÿ³?ù Àñ+Pñ¿ý›?ýC@„?@¼¿úK@…¨¦‹C9dºÞ¼9œCÕKC=ÜÑÙKáÈ;>,DCL3>DDDWWWSMMfffkilwwvXccZYY–¶ 55Öÿ÷ÿ6ù66ÿVÿvýr-¢EE«UU¶SS¼vv³ffSNôEEøUUæMM÷vvÿwwø}}÷ffäffÆ-bÿ|¡qÿ–ÿ¶þ’6¿ÿ?ÿÖùúÛû)ÿ—DÿšJÿQîŠNÿ Vÿ«Eÿªfÿ³wÿ¹~ÿ®qí”m°ÿP‘ÿq–ÿxšÿzþp¬ûs üaþôDúõJþõTýöZèôSþöfÿ÷vù÷yèõpÔðc–œZ—¶..©DD®UUµJJ±ff½de‰Öÿö6ÿ67ÿ37ÐVÿvÿ9JýwwÄDPÿIUþVYÿCKþVaÿfpÿhrÿwwÿqxýchùWGÎV3èe¯?ÿ¿pÿvÿ®Vÿ¨—ÿ¶ÿ0œýw€ÿ|…ülöC®ÿK¨ðS´ÿY·þI±þf½ÿk°òSãÖÿùú)ýØmþÐvÄÿwÇÿtÐüQÿô[ÿõMýðbÿõrÿöoýéWèÛf—“•••µµ´™œ§Ê““ú’’ÿ½ˆÿª’Ú³³þ»»û²±ù¬«Æ¾Ÿœÿ„Žü’¥ÿ‹¦ÿ‡«ÿ•µý–ºÿ¦¬ù±“ï®ÿÆ™ÿÉŽþø‰ÿù˜ìó’ÿÙºýÍ­Æÿ¶Êý®þúªÿû»÷ê±Íèš™™Ó‰‰Í««Ú‹“ÿ˜™ÿˆ‹ý™ ÿŸ®ÿºÿ¼¼ã«±ÿ»»ÿ±¶ý¦©øˆ¬Ñ‰Ìÿ™ÓÿŽÍü»Âÿ©Úÿ³Ôú‚þ÷”þø‰ùí»ãÿ¡ÿù²üø îÑÌÌÌÝÝÝÑÑÚÈÆÑýËËýÝÝìÕÒÔÿÇÕýÎÿãÌÿüÌÿìÝÿüÝäÿÙÌÌéÀÀåÌÐÿËÙÿÝÝùÊËýÌéÿÄçÿÝáÿÂþûÐÿüÝñÿÌÿâîîîèàã÷îîþîîéÿãÿöîÿþîòÿìîîùàéðîðÿîøÿâþüøÿöðÿþÿÿÿDZÞ!ù,ã‹þý H° Áƒ*\È°¡Ã‡#JœH±¢Å‹3jÜȱ£Ç CŠI²¤É“(Sª\ɲ¥Ë—0cÊœI³¦Í›8sêÜɳ§ÏŸ@ƒ J´¨Ñ£H“*]Ê´©Ó§P£JJµªÕ«X³jÝʵ«×¯`Êõ÷N€±hÓªmúN€³kãÊË3‚Y¸tóêÝ«r€Koù LX#?Àü.̸1WjP#HÍÁïôU¤ÀÀÅŽC‹–úàÝ»¼ó÷À?ŠïX­ïèÛHû”Wðµ^Ïþ€ ôÌÏ’p‰\zx€Ð ˜y:Ö«OÇNýïÒ»þ[çþÝ;øìÞ·[OAûíß·'Ÿ ýuë ⣿ÿ¼xõÓ±gÝ{öÉ7`vúQ7ßzö‘§|ÔPà~ÓÈŸxþ•wáw j`yéX!‚Ö7`d68(`£•P€Tá@TP ˜ @zÞ5ð¡‰Gva†Šx]F. AÐÑ`©À–°å—^€åVgæ@PóŽÇ=ÄOXBPsþ Ð&W„u‰œ`Ep8YÜãÏ9køƒÎþHâL)Hˆâ òt“ƒ@ôœOA™dâR&!hJi6'¨ÀÛ&w*D¨Z€êª¬¦ºêþw<8` joŸÙæ„•X—\"¬A€®vÏùèH?ÎH‚JGÔÒ„?òäÀ?(`SP¤€ª­Æo»¬™ì@–¼s%|Fä×[ü—@ÀK¬±` ; ‚.{Ž>ƒ ÃN$¨ aÄ6LØãOݤ ‘ÐÉJÚˆ@®?üdƒ‚&¯mÒÃ7­ÜëЩòÖüê@Iö™?–à€ÂÇæÐÉÀæd¬áúÂ_USž1,PÃ8+±#©Ä6NxÓO«ä Ñ#‚Jñ” B<—–{nºë‚ãÄe:¤€ÍxÓ‹þHÖÓ.Ð=Y˜#P¾øC4ú<² E …+þ¬„@òtÀ[A%Œ|R<$ˆàù@Ÿt ‰?ô±Ã7P0a 1@D3Þò °ß~‚E›îaÑ–OâÄGâ‹ ã#îÜxÊ£þtÓÃ@.xbP'% Äéè©,ª?›ä‹7KœRı?D€ª¢Ïúé·¿¾ªzŸDg¹ãfÿK„ï/|øƒÌ"þ˜„4ŒP„RxãZôÈ=<¦‚ƒ¬Í$0ÙnTptuCa‹ÂHq#Q(‚º™™úFHBõ½Ïñ3Éüêw¿¦$ñEü'ˆDE£F8‚=–ðštÌþØ ›A2A‚’ÄC%hÛ@CÈŒŒë æ’^åvÐz4ÁŸF@‚)šŽA,#lxƒ@zú… 5¨*P‰š’@t+€K`µ‚‹3Äh ‡?Šç,’"‡tG,°Ä$„‰G&0p²±éÊÁåhaLš*³Ç„>”‘XäC 2‹þ¦ð:Û»Ö}|E þª•Ý}Å·XÁÂaø‰rÆ),Ž&x#´íÌÜæ ¢ \bt:H‹©ë '÷mÐ^÷À†a|vƒf€|Í—W¨|ЗuftÞ”­… ÈGóV¨Pó&dE-ð`{ÑòÐoÃg'è K0 ÿ9 û@„°Zhð 1qaÆŒ§QÉ—„è€Vˆ÷“ †%hpþp È% ‘0ŒB üÐväAv¤“o"¡ 0p”vàЭp ZsÓ ì þÐmÐ0[bQ„ ¡Í·„ Æ|¤öPþPQÂ`‰ãþÀ°ZS üÐr¼V–"ý0{[•:?ð üÐ M0 Œ2tÍ°X‹ ÷À‰ã€¢?8fÎWˆWqjŸ$«¦‹x8‰ƒ?ÖÐŒbÔ@ P@) Dµ$Ú3]«à=þ{­ÐJÈ4 ‡° ý i |¹°æ.>õ€æxŽæ˜|çØ„0ñ„‰½˜‡¿˜JíÆý7à3'xhïä+ƒé²døhápƒ§ §˜Z„W‚(¹@€ˆ@ˆ¨ŽêˆŽÏ‡ˆïØBñ(`&§ËiÛPN°k6jlÁD¶ôD9`p?Gþ ‘fÓ°Lí€^þÐ l |K7Ž‘1‘‰ŽiŽ¶è„ Z0½Ð?&‡‡Š"dF0 $†±0Kåb† @qA(NÀ GPE¦ð ’ÐÈ„§}px[ðÁ€Ñ´Žx‘i;1!}–WyIc:&Cqr–g‘ãPP Oó “‚‰hÀF?ð >´ð:xÒ@mþ  k`bÇv uC9”Jø|!ô»5}®˜9C3¤8Ëà H@^­@ dp¡U!™ ·'¶ „¨0 ‘Ðv|°‡þp€SHšå˜ŽxyšGþÙI©”z!˜²)œG`‘`ð& ü€Œñ™A¿ä «ùRÃä˜IEÈ´ ŽÐ™¿§àq€2„”!ˆC¹§ù€ÛÉÝésgäÐF‹u ( þ€CF€×¢h~ö›[uw7pÁ :ðCPP–g‰§ ‘ÀËxjÐ^JÇtA a ˆ æˆ <º£ ª‘ÚrE§cç`í刨"Y³l4pçWw÷¡¼ægÅD Ëe1Û Ñ ‹ŠpÅ ƒ³Þt£>µ?ê£ šÀ—0‹_±‹[a§ Q¤2¡h€Önú iµ þþà\Ñ]œÁ‰በÄOQÀ ¥x Ò€‡R›Ágç ppÚ£çȦnš‘NÈ[Cjz |*(Ððcûx“˜–È=dX(p16`I«@Hæ4X  W¢À Ò¶XªåpP€‘ 8š£ú£ŸŠ‘° Ñ §ú©êGZQí%€šx\˜˜f×­À…7R î°Ÿþã–þ° i@³Êf*a»Fˆ»lÜÆnìÆôbûԠ4·¸ÃV§Y¡ÇÀ"Á´þò(‚`<µÖB¦œö¹V+K±‹cƒ0ØT¬ªó«ŽÖõÐJ7" ¬û TxŽï €$±m·ûƪ Çhbvœt[H‰Ç<Œ¼*gþ‹;žýPÄæ„ þPµV¼Œ Ó•°@(9K@ óð:äYæ˜þ@bÂW Z@—+Ѿ °ÊàŒ»ô’3Hs—À&µÜI˵¬ë+À?ì¬(`<½ÌóðO¤ †z¨–üžÂdo–ÉZ# ²à ŽÐú@dÚo cÎúÎ)áÍá ÎD›ÆÛÎO1X‹¶}%G®¸ö°4–ÃüþE?ÄÌÎ oŒ«ë¨„‡ä Ä0]Ѫ,§/¸^¡Î\!Ô3!͈ñ?±“+G¯‘N&û˜AÔ a{L\ûÇ“ºÒðäb1Ë4 (ŠÄÛÀÓu[¿F¸¯i=¾ñÑŽµ”;½µàÀëùÏQ} ÁlA `¾rt óÐ ÍÐ7I¡÷€½0=)jˆñ«»—­»m}­ÑJËp@g–\þp‰lµ±Ô¡ñ]aVêÔa©»`ÜpP«SÂ'ÛLí›ÖºÛÛ½m¿¥êÙYË„ÆKí L-„*ÌV¬ïJ²T«1þ5oÏl1¤0,Ú \r&_‹2±Û¾=Þ»ÙÑÙÂ]É2¼IÍÝÎpŠÈY² P@ ~-ßðõ)®íȹÉíàÈ}(xhkqJ—Â5!ÞäMÞæ]tªjºX¸ kï-Ñ"àUº»9œ½¶xš:/I­0oüХΠíP¾š! k@³ïm6qÖP¿™m°ò[° à°é 랣ý¼¥¢±×J+P@¶K ~e³Ð€ÜíZ|4:qàýà>ÅÀfnæg~æiΦjäÁ=äQñŠ!³q(ºÑá!hþ^hŸ˜\dõä± ;°Ì¥ÛàÀVEÜÐÕìà?óMi šþÐ6šÞ\°hî梺ékÎNè-ç=ÁÑ°è npç©d›*µ’öÌ=Žé‚þΣÚbë†þ0Ïv¨°Š]r Cb=8\ õ`Ê3áÍ›®æŸ¾æœnæ¡>@ÝT×~s=Å}(šqg(C¹s¹¡ß´Î€C™QT¢(Yšl²‰ä‹à®Å^&»hÌ•àæm®æÍþìÓ.­É;q-±GmäÞî²ÝM}Ýö]½&;ⵞ«¡Ò1¶þP¯ÐfÙ}…8õMh0 ¡þ¾Ê^û~߬æ ð>îò£Nê81›:Æ©îËðŽàðši¹ŽâK´ Ð+°pôJ/ºÑhèƉèþ Bï‘Àå@¡9iQÓÙ‡@ÁìiŽæ_ÏæÍ.íBNóê}ó¾ ¡‡ãµµÓr„S|]S*oÑøß‘{”cž]<øúèÕ\ï‘­©7ÛõeöŠ¿øŒó1óf_ÔñÜí«þªº¡Àqß±ñ z÷”Ò=†~÷ÀOP°k²pƒ²Ð1½X=(ò”þ^>Ù)_ ŠÏæbÿõŠïøQátá+Aܳ¹ðíã˜àŠŒ›ÀÄwŸ:`ƒ3þA Ô°vd*·±ópgä{“Æ3ÆZ_£D¡ã ! €ûbû¸¯ûù‘ÿÄó ÏQLõ2%ˆéZ멳û:SsÈaD¿fôù[ö¿{lzùƒˆ/‹.ˆ-^ĘQãFŽ-"ðh±’%MšÜ°ƒÊ!]¾Ô!Lš5mÞÄ™SçNž=5ŠËr¯"6Ñý9¨ïÑ:ÛŠAUÑ -Ž:VU\áÖ°4ÖbB˜½SHŠ” ×/’3ˆ‹AT††\E]YÌùÄûäË*O®ükR@^œ2 FœXñbÆ …ئQ¤þú9b×´’¨£¸þò¸Š†M9vlꦱž)ô†U¤*~ë0ûkhD>lòUär¦ñâ½.”ô€\9‡äÌ9´ÞñÁÌèÕiV³žý%öè+¶áEùàåÌNgW¤åå*¯i-±ÅÏŸ*$HŒpó')ÒAdÚ1d׊̙H;€i¤æNZp9è ´ˆê ¤Ð¢w*Ä°¢ ‡Ã%¨ïÂóç¨ñn §>RäóÇ›%|ª'Š'‚!†QêEa a$}yËaÐ8‡.»2ä A¾Lj®ALîÁ #2J)§Ä+., P<Ën›ç#N„¨ž%êéé=Z†æ#ê3F™F(Affþ€þ"6øø!-„C‘ÊšŒ4nIå’k¤æœ¤J?e´QŒpÑ¢6 1¤ajž5ÁôGLoxj 2ù9eMS˜!&I êï uРԟj t”¦âB:.0$DBEk6Ø q¹ò»]´ô‡ËMû„‚ªÞ«¢yfDb–z€Dvìq4Ôò.a]4¤–D7]uÕ,C&—0jâMl^Å‚“”R!’D[~H±ožŠZie§zšˆ‚Ld©ïˆTüqbB¤¢>ÚÈÓŸ3„£7$`R@PBEÙPæÚÅð׎WfY'.pÉYIØ⧔€+ªE=ÞûÔ{Hþ‘”wô‰ä™aˆ9ÆR†Ò8")j™£[=R0]æH^·WUŽºk¯7ÚfˆòYCßÊú™Ù~LA¢!ò¦ …oáŠd1"oYüY’̈٣Gr€ ÷k¦î(dçÖôjtµÖŽkÃ'ú-Äö'5æ q™ú‰#ÞîtÌœZi¢g~P ?I$é§ó…üñÅ·ŠÎ(–ò‹çè\« uÝ«!Ïn:Ýmâîøë»' q*ÊG.ˆ ©LŸÛBG"œŠøY"Zº›(¢pŒm }"‰FÖþb÷Ç74¼\ù‹™quƒo’H å‡iCþ]ò_O€ò<ˆ ŽsÈ Äˆþ2ó¯ÍlÃ"MðžMPŽŠ âOÁO4ñ}¢GêHƒ0*2ÀÿU¤\TC×Ö¥B°ðqD’\ e˜!ql2þ G “ B@¤X )ÖäÀŠ446©[ø|&D$˜‚D[UçêD;5€ "¸°á ýQ€úÙÏ…-\’ CxiÑŒ‡n8¸ƒø>"DPa–½ L‰4™`E¶±&$ì˜b•Òüñ†6X_fäÝFÎ¥B22…ôñò‚ À’°ÄEbxFN2R7<‡ò„ A@Äz™ñÇ@Dˆ¸Bg49X¶7G$ˆB`Ñ€ÄüÁ#ˆ4$V–Ãœþ©‘E‚ј_L×$Ù“wÀ’p, ¯wu#ö²¦E°™bY¤vy:F·„JZ²Ò9³ Ï*2­5B>’hDr³¦sþ !'éÇ/†Œ.\&O&€\¢"—2 ‘MfÓ¡<±’EvQHˆâÙÇmüQ$˜³rƒ%dy·ØÁŽÐBþ Ë^س"¹P£5‡™‘|À¦Œ¼i#ÿiS”%Fš =(>ÆÐ2>Ô¨;ÑB.$JÑ]&Ô€L¥}Ì©¢¹¹¤Kp–Ú8Z {øC]„8ïц^  "¸sèL1RÌœ¶0§oU&c,™<ˆðÃ’ièQõêþ’—a©iþ2àHNò™SLUõH‹V‘p­u| ‘W)Ãhj^0;©Ö‹Ô´­8…«M#Yj$àÛt@jð¼T(3žQéêÐÙ¾˜’‘>!Xü1r”a—Ð3«vÏ" «%~Ô’¾íav?²â0rá<£îÓ%žÝ©hµF›>èïx2â€mî$˜ÐþŒ@‡²×%Í#`æØ[…@„ÈGVfQsú Ll @YšH>¶Ae«ÔƇ6Üð X¸¡COØ„ÖÂ9} ¢á‹`b>yÇyà¿`'Ææ;’—âk¨ÅQ1i],cþ¿Ø1†‹i|cãØ:ÎñŒƒ\ãûÈ?2’‰ yHŽÏ…aÜÊSþ³?®‘ä#šÐ‡^²’Él‰œx€ Ѐ›ºõŸ- Дq˜Ã9€{kÂwL ¡§µH^÷ºj‹ä3sÒk.:äóǃð# ü½ˆ+œàÅVu<~ô‘¶mýžäà€9¨A o€6´ÅÅIΊä2°° °ýVdxñ„ xb‰ôÚ,*«ÕÝjþAälà:Á>gD"vØëoÜŠ àÕ™¨«ûÈ–Qœ ‘eŒ@Ï3ëQ«]lÛçöµoú €—µy@ykòŽJ `¢®¦Q5nÍ‘kä“\ž¨™hLžˆ…Íp‹—¿\Tí?è)TuŽ¥W±@¶‘Á †]ä3 Ífs¨×…«ÃC‹íQã¨ö ÒS›4ß݈ª×mÔˆVäh'! Ó I`æ2wAÍ_þ Óu䪖‡,JJÄ×µ1=ÊG–^Ê…&ܨ ‡HÃ!nÓm3^ÛëbΛîTS~ìzÝ“ExÁÔpV„Øš XV*â)Dþ¤¹Å¦öC慞ѻ§?YÃ?ñ[¼p¶#Ÿ˜wXRÔ©–Àå×}¾…©‚¸(DÐ'¾5A`¢ïTê:Bð]„Tfqgo¡z]’à ÷lÔ,Rû´&HÛçGú®í€êä–„üï ‘_¶z­­Eš×åïÄŠ¼U ¶ðMÉ›·¾á²;´x˜"`3¯‚„6R0Úaƒr5‹0‡qЫGS¤Ç[?¼¶õ»¶öà ~°ùCÁŠˆ­‡ú=NjAø‚X1¥Ýò!Úш½féC¬_Kðˆ‘„6r‹ïà‹89hÁ3â@P¼ó ÁT?ö#þ#áÛ«q ®±YË*¶)ú#TºÁ/éÈ '8"¨›4¼›"8[¢O˜„xâ‡Üà-²‰‰,„ˆÛ‹B)D‰ó@®óËãC-‚”i‹‹x«/¡k£9*‚ì»(`CŒ¸*⪫²(‚$`JÀXL0„)/dˆâ;ÄÛ+¬B@´Â »C”Ÿ²ƒ‹5P¹Ê BöA…5錋@CŽðÁ‹˜›y,ˆX‡A°dÈðQªˆYá?ZdÅW´ÆXÄ{¨’ã¤ò⥪ˆÍ%Ar†(Z›òÁˆƒ©D‹PGŒØ#ûÀР8ô‡°ªeCº! n<£§3ôþ›ó HL?dŒY¤Åɹ­Špƒ$¬¨ñ‡µ¨+!*‚þê&ÀÄ‹P‘ŒL%ò!f€ƒ7„Aàƒ9 ƒ=0h5ˆÈ…3ˆ0j4¿kÈó3HÉ3Ä„ÔwÈ‚`Šø?Z¾l¾:bG¬ˆl¨W8½C‚ì{Hp~à‡“t*P‡.4$ŽÑÉ>4¿œ°”±È›T „ìÊ–qµ|0Ú´Œ8òn6Ò=¤Ä Ï)‹ø™ú¨£0DÆ4è¸ø–º²pLKxBš"ÈkH£ YAÚ¿¤[#Xk‡FLÍø’®ÒHéÈ“¬b,>ˆø*þ „3ÀÇÙÂå9ªÅ\+õ È°ËÜD¿(CQ/Ô++¹!b@È ]ú*ö‘… éMxĪˆÀHwô’ÑÉŠrÊzŸ8xŸÂéº L°Ïñ¼MÜO³L ´LÌq9ƒ=„ˆo*º8ƒµaóÀW‚Nˆ €Nä2…{臈¬·L>¢ø+ –˜LLV4OñËÏn»Âõ¤œo´8€ƒŠè<~1G€é£Œ)à‡(ˆèë¥ýB‚T z ܇?H‡±iK‹Ø ÝðtÐ5O}PôD õ´QF ?èùÉŠJ¾å’–¼³‹hNjȆXÐþè(…Ò!HÙ (eJ((Í„8È„Ô 1(CBÌ õ‡j,OuPóôÑÃÐF‡âÇÿÉ'j¸4ðB}06ŒÒ(Y8ûÈA‹ØÈŒ‡ohg¾p¸ %ý#tÈ êÑù¨Ò¨S-².sÙQ%ÏSÂR5•’ºX#—êœÕ>]b˜ÎÄZ`ˈM@JÑë7|gpùX |5`·ø"Õ5ÅQò|ÐÜlÓÕ¼ÕaÅ[ô3€ dÐE9»7ÖyÊæ°P‘O蓬0QÈÁhˆ"0ˆlƒµLg­µñ´wýTñ´eÅ‹°?v  þ­ˆ£ƒˆ õ‡fˆ§%"ŸYHG#âÖ%øOøV†)NPh:è%4hI)}%V¾Wy¥×xוW{õ‰ú;ªü‹Ž‹ ò£‰…dˆŸ¬†I>¢a®ñ1 ‹¼Ä„­‡—ƒ*¨Tˆ~@>e†à‡<=´Ê ”Õ§bYõاýX’í‰É̦&dŒî| –} qÀ‚øª,ʪ˜ÞòˆÔ›¹ˆaü âY9θŒ”: Yó„sXƒØ‹ÂÀZ-¢ÍΠ˨ýX©Ù ÕX˜ÐZ—àÚXWwû¾éiD ª5Zr¢‹XÃȆ!°!€R@`[ûÐmѨ}ñ‡þ¸¸¡§)Ü(VeÝy%Ä­ÉÉÔ½ˆÃ]Ùa° á G5hšV‘¶ƒˆŸ‘ ƒ½ˆ<Úˆº‘Ux¹|9"ÚLI…ˆÍ³À,@EÙeE‘×ë…ÚëÕ^ªå‰feWÚõˆÄÝâKS>HÂ÷9BËȹl¾þ“;ÊzX‚o°˜£æuöA £c*| ŠüÔ×êÕÞÍÞ~Wî݉9ͦLU ðÅ4à^(+î3ŽØ?è<8 Á~ †?Ý [ºuâÒ³ˆX¨Nx€ˆhxý'ˆ±]ŒŠæô‰A;ªM¥šx}ÚÞaëE`ð^ †hƒ61 â7PÙË¢` ¦à þþ‰Ø,Ó††ƒÀÿ8G#˜ÜÑs‚½Ü^'Ù8‚¬7 <6hš÷ÙÍ6à6àΉ Væñ½ˆ%¼,´c•AÈЀ-B°Y¸ˆœí/––@u¾.× d=½ÇÙScÕåØ.`í…Ze%ÜHæ:Þ;î%Hö7È“#„Öó‡xXΧTÏ•ÄBÀ|ÈÐ;‚XˆxšAÞz•{ê,h:MfcÖa`†cœ°ZkbÚÆàdðä þŽ÷1‡7ÈC€?0ƒDð8èU,†„uåž!˲Y®âæâÓÙ9‡Âì%-ƒê8f3ÂáŽ(`¦çíÕŸØþ•¡½m`š°ãµ !ʘg„ßø¹51Jý\ J,м,e Ë³@‘GÀ¹- â-‡dÖ[Œ¸-¡¾ ¸‘6à‘Ž×‘6éa¾ 9&U~àÎ<Ž‹ q#A‚§÷IåRØM=„¡áŒFÚS“‡•šEø48ˆ,àN#>bëðäãùe’6i’ÖÞ‘Vi›`i5…i >®ô‡^°n‘'”$1#…®=B J¼ˆiØ 09¥£î_ïà—^Œ¨ÖÛK€ë­j«ök{–E|Ödž0‡-˜Æ|h‡ÌÌ)ú*¢³ÙÐ5Ãke—pk‹Ø%@¹Ö¨ ô…þ­ þÿÁëŠèè¼ÐkÊéŠèk«®j”l¬&&¹¨éEtd„ð~±·…¹ #ÈÁyÙãz‰V©h3†úhóƒCðˆà!†tH—̈mˆ8mиë6¤špmïþn‘ön’ÎêšØêÂæ¥}ÏÍi‹Ú‰V¶)‚Q@ù€8Þp¥‹CU~@‡¡ ¤±)Ò+Êî¬n%6í®ž¡êEéoð¦ê(oš8oôæè¼…l¾edT¿=pÀ*­ßŒ¨…]ÃìŸE“?ba.œ^HFp˜Ðn8bg;âi“Vïïßñz%­«Û¦ðJ-ÉËW ÷‰Šþ¯èqKŠ†ƒð cˆxRð R+µRÄ)ÃäŒË†ü}ȱ5ˆPEê~^gð˜q¯Ññ§j9‡pﲸTÃŽªÃ0Ùû‹™NšV0µ CHLkœÁ*Y1C ñÃÂìÍè d€öadØ#ížÄˆwMljÚzó®Imˆ( 'u“†pïñ4‹p€ hõÅ(æNÒg)QoØ…±æ%íËŒñÙ áíµ½ÏËÂÄ=²ÏˆqË@pªa@¶^ÂèŒíh3b'b÷÷úô¨ uX¨(õi0¨Põo«…2ÁsûQÂFòŽ@ݱþÑ]ʘ¢z;ä2V2“Ðdt‹¸lˆ0SPde°]FV=^gaµMŸlïä6/!÷Š·xoó‡O˺ŠÐxQ…÷x׈`xöŠöBj=ˆ}Hà-‚,Ž¥üv.o%.6Έ†Hèpƒè/^8b§Ž6kÏtN÷ ˆgŠ·ø£GõŠ¨¸w¸8ú›—KøxóŽzWB«ˆgeqº/myG“BáÎÑÓp+­ÒÀÜPʦ0 š¢sv·5`*ï؇úž ú•1z¤¿ø Ç:­³ºÔšõ›íNb` ±œ3@‘†èÎ̤§fØÕ Tî'€‚Ÿž—!´n˜Î þœø@í ì,¶Ù)†4ȇî\ܺ§‰‡—ø—8üŠçÞá{¤¯p˜¸pÙŠi ‡çNG ”؉3Ð.iœØN È8öHþ3nFk™@®,aË@È ;è¾ùpͳF1í:Bþ¡¨?l4êOañ²éb ,îDæ‚Ké@ûå²½wQGìêü®rǀ㓮Ö<ôZ4èÏ=jì"2‚´ãÏ:°ÔTëÓD+ûãð7@а‰²›Ä2Q?SXÒ?Ð"±;Žˆu-Èlø²m·ÁL‘Ëæ ¬rÊ–‚ !ÍC§+? À oÏìÞ €Q­ÖK4ä-½S²@¶>¤!»ú#I3ûlSÂþ´Ò8Ù €Â ©ŸÞ ´°?Ê~óÃD­@a<¨D+ç:n»FäÜV·Ýä&Èò¤Æ{x̓KTþe∷ï—BG~=Fâd±/Wlœ8$4þ´- ?§¤&Ê´W·ÒðþÐÓ ¬ÈC då ‘7Mxê*E1‹ø—×d+xHÃñŠ·¿ãUJyÒ¹G|†1Ÿø4#ë:Ù%¯E{΃Π…£éƒ |È1!žGè#¥(BP!ÑC w²“<„°!|©ûa\ ×­ }ÑÅ0Ʊ„`,D›Æ0¤!@´6p0à=˜ÞÜm,È0,Å7¾]êo€‘È+ÁøLð35“Дü±€ãÔà]uÄÁñˆÇ:nÐCxœ×â éfL•K1\ˆ]éþCXy“ äT¬DYž Ʀ5ÈÉÂØ" ],áLÉŽ Ša by ÈÚð·{lV)ƒÓb–.\áº)Ây®Žw<&õ¨Ç>vÈz|æ@‚‘ôÂ{þ`LøÚ¶Žy ÃéÄÖà‡§Cˆ'ý1±„4ÁYþØÆPdضWÂFn‘÷R²†\Êd—0ébËw.çEä‚yD¦‰TÐô –´¥åŠb‘ ß”žË¨!«+´¶IäÃÃàÄ*’'D*=‚¡)¨8E.Ê€ùØB>1²F50…§Œ ¨z˜0Ïþ^…HjpÌ¥*S¡8Ð>ä̇rI yǬBŒ[Q1+ÚG$¢an’b=4<È*r° KÞIYØÀ:‡±Î(«Qˆ ‘0ŠxôÃûø]ï¹½–ì´§´*J„Ú@&!vTGMȺªÔ„&Ó˜Ì̽¨º™¹änºÀé@ø †!ãOsÆ#ö! ô ¬£S«8·ÖÖ‚\Š /  6eŽxà ™ðÄN¹Ç"$©J=1ŠSŒâËBhªáõ¤¤©èDØ„š#Ö±Z¢‚Hq©@è„|ã l 0Dú1„L ÄÎúD°ÐçV öˆöŽ¦À—Õ¬ºá%ñA â1«ª\rñ‹3¤$é&lèÃwqQIÌÃ0E82 $:¥‘ýñV$'ä<zày¼÷Ó’D4~¾0Ë¡”?íåeÍÓTdþ´RÆ¥¦ù% ì,AøQÈí¬Á;í„ÀØ il#PEþ?Z±„Z¤:‚²â ØÉNÄ.¶'æçücŸø¡|Ð$†CÐD$"ãAüçÊþ°ô¥Ë û%0äÞ4U;­.P:{Ò$Pj˜ôÒ"¾ß,ÌØ|°&šžÖ ž¢j“\ß@„ÜzºáÃFö±‹=¿XÁe£û nÃP|‚Ó°S!FégAËÁôbŸSI¡û¡ê&ØMf?Íóžn"Z¹2¹&sþX“3&S\Ó¢YY…`A×  ’No8=îG\!ÅNC@B¨À‚iLãŸ(OFŽå’[„oP¾=œ¼=ˆå7'þÕEìQ"нîv¿»Ýin›¿]8! ¤"Œî èá“";fu#pƒM¸ë@`‘ƒM„¿P©@òì`ûJH=–À©Èèºh !DåJÓäÁ ¥¬ÚrT'äÃ}'˜ ’¼óïzŸßoÿ—7@8´fI¡>ú I¸°2§àGÑÝénä ?È rye/}óéí@oôpKÓ „@vmÝõ¯_{¸­3ûyç~ è=þéþ{édVø™nJŒÁ 9¬ÁÜìC 0ÔÌڛĈ,Lä¹è@‘ ²y [xB6š÷-Ûõ „ðZ~þ¤?¬%xæƒ@ð‚€´_JÀ^ Þ tùEÜŒEì^þõÞþ=GðÙ ZdCÌÁ@ÀŒš¤{5Ë`Ý°D_äå Y h t h‚&h€&€B±u ²x¤Ô@´ÂúŒ• x‚¬Å‚'¤^ 6ŠöÔÛEÈ EÜ1ćO}®¨…9DP2¹Ù`ý „î ïX39²…ö€™¸ÙÜ胮ŽiÌ<`±ðÈûX_D`›xâÒ±yá°qÂxBy-èOÚ A)ÌCÛôƒ'è!p’]KÐáDäƒFЦE2…/,D¢ÿ ¢?Ì!úþx-â_È=(‹å^øHB$4œ‚=( ¯ Ä&ä@7L„& À Î hÀ¥N²Û*¢€‚L2Ü"€à„@Œƒ.ÚÛÞhÊh †=×sÄü•š2&€<$DF¤DFd ©AcDœšEÐ’!­@/¨X,‚œµÃ#DCZÆÕD ,èÀZA„<€B  @ h °@ ˜£?hÂ:ê×:z'È=lÿ¤7—?ƒ ´à ~[ ãëÍßBb„‘Då‹á`E8$DÁDn¥Tdp\$F&„N„öØÒ=Ìœ¥ÃŒO¸ƒÚ˜Â<È='rÞ@dþC ¼À \J쥥lÀ¤çmØ'0³@Ò6ì•A!ø‰/Äɹ_bAåAÆÐU†WpW~æCz¥©E€@D€FnHXªE.hŸø1¬ äŠi±C¶É ]bþ œ?¬ |^Ãå¤Lv ö Ä$D°}C[, Áú@ÒeDÂ:žXtÅdöÓÝ8D¥l&E(äjBDC‚ægÀsPC8€pÖ~çD^6¤ÎÈÁ ¤€!ä‚ d%„‚Ú”•fÔ?Hå9]6¨À h÷5 ¼@ꨀ:®£æ [PA4†ÁŽ]1p礄XDÐ̯¾+/ „SJ)ÄDýÙC8 €ÅV,«<À;¼ÃÄ\Lþ$ÝDÊ~§¬È@¸œUÑŠ`#$Y(i7|€» D<¨±=EÈê AåÜB(me¨B+Ý! (Jl¼¼"þS1ÒÖ¥¨ÚéAàÒ½:lü­?Ì)(ÁPÀ€à®îæ.ïC DŸúCðï8¶ÄÝ"Dò ¯L€óÊÌ;€ôJï@Do¡ „õ Äô^¯?d¯?l¯öV/õbïøvoùRõê‚Œƒd*CÐ 2 >@C$<ƒPAYMT=𰙯 µ@ ´<Ç €ï÷Š¯?´(0ùvCaâKþàCÿ%\Ã#àÅNøwþ` /ì[¤Œ„°¥À-%0ùr¯÷êêÙ°®š¯¨¼°÷º.k¯YX¯A °û?Ä÷îp ‡o£p ñ+°1+1[B8¯³ZÄDÀÝîî^1 ìŽïðv‰Ü:/òÖ­ò’±?ü‘ìv/AàÃÒð´; J$Ì‚Æ86‚†j²`Ä =ø@°yÔƒTÁ-Ð$àEÈ‘ˆ˜¬Œƒû «ÂòTÖ" ß›¡ŒdҞ쾜D$ œ2*§²*§2«x1D|1F,/A,oÞ.b.`AÃA€Ú±Á¸2L‚ýb]YAþ0Ü´Î}À9Dø¬Dh¡ ýëL ØÀ Á{•6ÂÓ "†w Í`EŠÕ‚­&[—DÐð±"dG G¬²<¯2«@@Û¾í@°­?\Âñ¢ÄÈÊ-A˜,Ê*¢Ž@ôÁMèÄ® *¼—( ‚®9‹<üÀä, ìEgE òA+ Y;• )ô•ø2há‘sNaòZ¤sB0l>°ó¸¸ó;§Ä òƒ̳NŸ2ÆjlÅÔ­%€ØsKT+&\+d+A§9dA·¨ÌMhI>ƒù(Ô%užŒcÃ\ô‚Æ(íÃ(Z€9+”OŒØCXÁ#ðþ#(ÆrÃÞä[›–Ý2…ˆSÜ!±~KuئøA¦ëÖôEÜ´)ïôøD*PÆ­Ep†¯%n<´€ Ì'Dh4FÈà ¼fÑÀõùÀ]Ù™£5Cdƒæ8ØÕMÅ°tØö†ê·ˆAq«?‚-22HPÍtK„ìŠFqEä4c˳¨ ?ÌhÞ(Þ.54¾ƒ˜ð°›…ë ƒ8‚3Ø™”Â,¸Gç-Ü@¤6 „*4ÇJ¨€ DŒ£ùÿÄ78CÁ C ,ŠrqË|ËfªA"íb;#ÈB^¦ b؆þÛ d8‡k8*ã À_‡y¢§z²§{‚7jÏŒq‡w€‡>HäÖÁ,HP®äjB‚:n e64APœ@€ ÁyÁ6HV¸Xð)¸/5ìQ8t¯]-ä9?TCv¸—¹86JTCœ§ æAD€¼é¨4 Ŀ݃/ƒò=Ã( …(¼ÂÓE‡ ó×@ø¹[ ¸¶¯Ù€ÃÕYFÛüˆD.Þ”×àK¸tË oÐ4ìÆ.A ÷Í‘rD$˜‡:†‹9[ÄÛŠr¤@¨C˜I µƒ¥‚Æظ1×?ÄB§'œ dÄ}7Œ‘Y!ADÝþõ9-ÌC´ Â<Á’?øÂRHB‡„§¥ÕAÈ´¤gúETƒ’l9Ñx:D˜ox¨“úZ€%4J÷ˆ8Ì1èÃ# ³ ¼·RÐÃÁ@ȃ ´@XWD¯ã7]qô@í µ‚=¤)0f?Дu Dˆ@„ƒˆKƒ$ƒ\:¶ßµE¼Ã¶§·'¸‹:˜“»ZÔrxô= A¤Ð>H$¤è ‚f0Òyµ?Äë~KÄ|ªÀ}í¤óÃÌÉþ–z0¦qÂVÃ¥;»„í Êpz;¿_Y¶a_Š•¯)¶'ãEô¨¹¸ü3B#>PÎ/tÇk{=à a5-0þRè]ž€¾cD't‚6ø;•Þe¾~ŸYÓƒèA‹zHÂt ‚ÀL­?èBM$„0ˆâ #ÓoŠ™p»àh¨ãàå[¾åW~æc~˜=~Y¸N  :°Ã#ÀºÿÌÂãÕÃ&^_j·€kCDȧEe Ÿ9hÁáž2èà ìJ3P‚P)๳ÄÂÈ<2DÀí—@'@„5D´@±-@ ð40§lþ ¸š…¤ˆ,gÎü;æ/rþü‰ËrþcG©}ìøFdI“'Eòz³r¥˜8b`ŠQ†åJ•(q¢¤FòÞ0Ÿ>Õüü™“hQ£G‰*8ZOS;MñØyêÓ§f5!‚V¯_ÁúË¥ÅE5|ü-„n¤TF ™ÅDŠ¿o9€tôt‚ÎN%DdŠ‡²žV$NÜŠPt¸cˆ§l­¢\‹;+Z$ÈQ´ˆQ#GœÃLÆ))Éb:š+ëšµHžB‡Å]›wïŽ&ˆ:ÕéTâVñ`õý•ërç`ÍmÉ…Q™š‹‹µk6ÉC#¯œxó¤Ç7ŒÙN¨w2^¦$àÅ@ïä òN`´ëþFZP‰«v<ƒ&Ñ.âG-Lɶ‚z¥1‚ #Ÿü¡°7 3”ÐCÖ€3ʨ†#Ñ8ªšRîâš[ÑE“tÉBŒàh#hþXFšGR¥ˆ"Na¢bÉ!–óZ@á “þ ŒŸ’<1i„LêóG…(ýYA0ªe ~DABjѧœ?ôñÇ:Ò6*ŠC5Q{q$‘,ôÃ8ãðM9ùô(Ä¢FÄ£Ž¦%ªA U±Ï’èjQï9ƒ ŽÈAƒ dŸfñ±R¦`Âz€ø>@9A““:mP²¯$lH¨Ï“:Z†nü¡g‰zHAb”N(aGŸ?òþñ'ŽKý¹'‹Õrâ'“0ìó|„FŒŸzéå «aö£=•ó€àH,´D =4¹rK¢ÀQxŸÃ‹`0jƒ|Ðɱ­T€ë”%hy–`ñ‡ŸlPPá¤L0(a0œdéÕ“T8a=Œ@82&¼A  ¡„~þ@GYfß‘ñ¨7«-I¶<±j[Ÿºýö¹wL"—ÞÏ5*uÛMQwý¹ÅŸ};ã ŒÔYC˜ˆ4•ä™`‘(e (FýˆS5é“ܬ¨Š?¥@éhø±§‰o¨ˆkJH8ÁLZ˜ƒŽƒœñejñ)æÞ ÇHŽ—Óg¦Ÿ PwþNWÐË“¦wiÈÃrVŒ„ACv9†·~r‰ZüYaŒöZõ#~Ú Ê£ÐîžZýÉfI½òè›0ÊäcâÂd2è`c7а0ðœÌ©é 1XŠi%5Ô°ÚÆ«ûí»¯‰óó}û“(á.oŸP÷¡ àIÈ7G?+]´˜18àˆè|$1‰0AkP¨Ç7z—Þµ@=%)›HrW’ú#·ˆGð³1zè€>´ B€ŒpŠfH¸0†?xÁŒÜc ‚k áN¾çÐF†ö³¡WÔ—“¡¹KPíb¡0?¦A`^7,Ê=´ð4TJíD2Øá€Åe`´þOy0â%‰¤…¹;rŽ‹dC1‹)cbø‚Ž|KiA0o,Áp!…4$á6ÂŘV³\¦•8Ñ#5´Sâ$DCDö&$F|‚C4IN’’uȼêçÈ“hDp|PÃ=ˆvHB¤ø‘)¦Ð„zi°ë€ì>1LL‚Œ[d;ñÅ bÃÈ*òâ+pÄEmIË<щ5¬0Zy!1òCÒ7ŠÔä5‘’Cœ ’Ý”ä%Ë•Ilzä ò‡9Ðà tb§KÅ~4°ZÜåkþðDÃhÙ‘²å³–º¥HD@Ë]â§#ÃÃÈà‚i@¢gjƒþ292½¯èépÖãž÷žó½EŽ“£~Š¤7½ ÎGE@ý§ã^¬™ê)=e (D¡@Øyè‚lÌ#$°ÝÙlÙÏŽhCmzÁ ÆÖ«cL…*F6 H(Ã{¸GÌÐ Œ´0,Ýè!EM“bS›(á&H»)ÒE= ¤cõGŒð•6¼Á‰ÇhG$T—# ÌÝ0/OÐFTÃV#ØTL†¹+A Êؘœ ‰É†²1ŒPLaJ@‚1( bìêÃEüÁ\°¦ÐÌÐ9tV±:¶"¬[%¤”¢ðã¬h¥¤ZûÔ¨±"Q‰»`Ã9þñ‡v8£Sñ|ÂlJ~È£a"G f)’]ÚóŒ•MŒ \Ë\y$E5*6PÐ ´ÒMàÆfþ |¤ÌmPFp¡Äyå®õÙlÇÊ3Ü>'FámàD$˜Á ^ðº\>‰Ó‘1šQuv¡@,bxeȨЄ¾þÕš8AR;Ÿ"V»mDÛ^À’4”%o X »Tm ˜½ZB8Ž€S4cV|_.xäq/DÉá 5`7I9›Áa°‚¯ìà#XÂr¢° ïÁ%öB åXFÊH‰µR<—ä‘î=}©;ˆ Ç-¯GûUŒÀüÀKªAùþ!âRd-’pƒ  B´dD ã2ÊÌ>2 ë%âò'“©l”&g#e=Iœå‡¡îò‹¾Œ¾`d_øPÃ.ú±B¬ÃPKˆé $L*Ø)yç\g(¹x±AE ŽE’$ŒÐ#ôq‚+LiwÊjAªz~"°1ìýÑÈaß™áÔìï<’ É`d¼rEDV„ië{ð‡/Ú€‘èl›ÛE9œÂ1=8°Xº\(‰I¡©»Ô¢~0—]þÎ"Ú\%‡ŠáEbG©ø‘B¼Ê×1ì×C}JÈY(¦²x6¶IBÀOä AØ„?jYÄeeBÆ"–ˆí–•¥'BÑÍObk8=—\Ó]'¹3mî’ ûÊn7Í|j¥š^aVâµù•#R!¥8ó´àwþà³ÄŽÂ°©›±Œ¸º4Ù2MÈý™ã6à‚ŠG°‚è‡:Ð,äo…¸y»Oâé¹ï·î`Á»œÎa{ÜDæ> ß±àä«›ù¦>á˵ |Ý£ h å:aŠ£S! öø“ô `‡VYÙîWHòŽÐCÁFðî"Ýþ&„Ã@ŠD4ú¥|à#ZXíËMn÷%ª§&(­ÆTÎ~‚Ï톯ø~ÃÀ@Ý’o+Ðù8G¸Ð'R&Å~Á ”Gø킬¯`ŽäýÚÆ# KØ>â¶N?´‚rlº4#rÀ'Þ€ô;Ú"LÄ„ûÀêIP€§ü¨² 'l³$+@¡ýÂâü²"Ä«Ä|©ôüÁ @d`ÚüAP+¿Æm§1Ü&)$Í­ †$<Ñ­ôN$"0ÔL±ù¾Pìs RÄ #z @iNg~$ ž º¡‚€²!I~‡+,ä§ÐUF@°î #ä¡Ùj¡ d.d¡La" #taɪ‘(¼Í1Ñî4ɱð$Âñ#ºÐ 'M2%Å0™fÕ>ÇÌ pœáê/„šXAÌ#UÄOŸÈÏC ²$ä!þ@¡2¨Wv¥–@ŽÀPÁØáÐäØDýd†%Ñ;òɶòQˆïå¾q¬FÒ#Ï$+Ð •ïÓr ?$ÞèER8‚Ø ÜpF!.Žà¹¾a¾.ªËéî(\Ž7l:¢Z0Žü¡ jA`¤r>.ä2¢M–ãäìdÒ =‚3¹2 áÅ5 ÓÂ2$1B·-U3%Y3Â:‚à óÊ,Ö1£²`µ”ÁR aFžŽ rr°(lPLæ,N…i"R4ì0J€àë±þHa¤¡ÐÄ*3 °’5Ä'xáPš²ÂgÆÒ÷JB%´±þ&vÊj#ºPùNR-ÓÒ>AEàÞ>‚ÀVä6—ÃÂü¡öEÖv$/‹`/iax¾Î^É#Ü£r^Ø:=!l‚?ü¡¤`ÐŒ`¢¡¡*E䶠èåAÖó Aó(â“(šlFMj>ý¡$UR-Yó$ƒ#ÔŠ#ÞÁ´@ÛŠOÂl üAØ@Ø©Tñ€ ˆ<ì½T€–Ze0_Äwp‚,ès=ºAüÁ˜¯3ü!Ð!j’%² tt¢8Ò+þËÉX#GMS$¢€5-”“*†ÔH‰Ô#" $ŽTBØþªOìeì Ìáú!/ÀkÁ¯ŽDRàwè!9y£ s1O‚FÀ¨J©è)g¡Œ ¡‘áЀí\fOsÂU/EFu96Í3KÎU5qPE·)H©5HU¤Hô`@`>D]dA¦ƒÒ@œ¨fñG œ€~à¾A-ô°´b;‚qb»J"xÊXÀ—³ f¡þ¢²œQM.B¢ˆ• Y‘Â=³Çzn´+3Rä4ÓzjZ?BZͪZ=Ö UüóJâRŸÃ@Á¢œ8¸îABU@žÀ6azíRLb(B_?Â=þ:`0„À:_¡þLA*Ï®¾.E¢h”OHÂÒftaŸö­–b5¶#xÔG?ÖZ_ó ‘À#LÖ9PV+Xhµö/ÃašáJñ´ô`Z‰àZ‚~ ,x¶( “IúÌ*túÃJ´DA¤!ä&ü âà=Y¢4gH$$7ycÓÔ“(˜5$·¶k«Õ-=äl‘¢œÆÁŒkŽ!¢AêÏ™€€€'ãÁ×N¥¼öPokCúb_Ë#T mî¢M›`вÓú¥Œ #¸ ‰j{OYâ#c3öj³¶(:×sƒôÝ8.C¶`µ¤tpäžá~$8¡àþ`Ž$ #^7Ðo4ñ¬Œ(ëŒÂL=B|©âÀa N!.¦¡úaôA(SŸuE\Îy«–=qt+é÷zK5‰‚k½ð<t%Dt‰b,8BÒ€`“àܬ[`bø âjco‰‚ì•Ï<½ q˜RSAŽ*%Yp#øC²qz+­6¨öM ¶%$¶ˆ)8':Wƒ©8%¡Š5ƒOV;G ¦#úà ú…ßè @¬Ø—ìpITLŽ,¢ŽêÎÈ^‰¢oÓ¦w“Rp3W‘€W¹3*"YÄzç9&vÊ"%67ŠPD²8‹µ¸k¹þ—~¼ø+r . †TábT™`…€öÂæ'~ûa9dX$äa \yòØÈ•3¡1J†‡ àôÊd–AÐà[À7<x‘?djÙ‘kcŠ­8ƒ'ùŠS’‹\;gy›¥ ú`öí|ãé"”`av[#ôP(MÂY1È•bòÔYúŒ†m•päVo’APÃYÂÓNäÄU‹¹O)™q•Ù#,8›Y‹%ÙŠ5Xš—£ YC#ð%t”ˆ®S†]ç°âÁÄ°dUéìCÐ_Qý!B oEat 2XŠ`œaiÙ€YšiŸõö‚ÁâFþ7¤ º#8ö$º°ŠùcÚ7<ø#XQ§»§_á言Va×(TvÞU9waIz÷ýãba‡oŒ÷@vÁ5¸àsŒÈšòjƒDŽï™©øŠñ:¯ñ:©{c©W(‰ìÄ ÈaAh1.’@ tmÓNåBåd•óKÌ4•ï#7!/.¦!ð¡Èˆ!õë†ðÀÒ“O*w®±÷ø ù®];¯š¯yïqÕ¤”zs"ë jÁ¦ê¡”1Âvû$ ½â‚l¸#®Ä *—áÂᇃ؆ز±É´UµÉŠµa»»3¸µe»6¼÷(æíièRþR÷mQ¸ bAb¡×¨°0 w™f‚K1EÀQÅa˜\a ž»LètºsˆÏg‚±ûÉ.6‰ë[»Å‘»[¯:¯8¼YCtËÉäjê*TBé”ßWÅnwÀ2âÐC6¨¼Á LI `ͤ°›ÜþÙÁ{ck_ûÇ)ÜÂÃe­! >‡,"¹Þö)‹ ¢`v =(ƒ«MªUjÙzþ¦µóì¢æ"ÔÁ2?2 il%®1™!ÇíÇÇõúŠõàÍóZÈmó’Kâ ¶@ÃÝ  Öi<üAŸ€r`øA¶H¾µ*C;Šýhiôüþ€­¡!4Ôà[Z_Æ*ÇŸ7ÁÙüE|<ÎáüÍE]Ô·ø|¨ù$bäs~ ûj2.²´Š¤F«Cë· MÂçp%â” z¸7õ¡ ’Ìimˆ×9=™=\>ý4!\¯ãÜÔKýÔÕÑÎ;B.ïÞ@C.ìÑHò(#H¬œs€º$®#Ðã>hÀ¿ü#AV´"™Ý†ÞýÙ9ÚwfÚý¡À ùÚG]Îçœ%O‰Vëç„¡Á°ã¢ˆ€ˆ¦³xòæ]!Oà vj ¼R†­­¥Ìð'no*W£þ|Ü\á!Ûþsþ*™i¾ŒgØÀ@!fGõ fRÅRùàBÙ0ÚX˜€Œ€aa~! 0"RjÀ ‰„O(RûêjþgxT~ç¯MΙËó°+åa¢}o‚ ²A¾Taæl9‹o–;Bp¥ôþ7€‹@Öaá–‰º ·ÞAÌ£Ö~QB]DîG¿ôwžî½b¼1bÕÄ!èR-–¡¹]×Xáý°Dèu¬ì$‡ïÂœàÎz¦!Ì@ˆ3ß‘v¿î~Oàcùù ·iôs^ô­ÿú±õµ¢~$å|Ò•þ!¨ vìÁ7†UC¥'†r0ºA ^!.TA–A óƒ$m4‘ŸÀ <ˆ0¡Â… :|ÈDz*Z¬x£F=@üø1B„÷¶äò×+º?ȉB‚$É 4ä}êJ`& @ú$èé'HÚvÀæ¯G,WSŽI)ÝnxùËÅE¨Ö­\ßøó qØ0®dËš= RâCŠ3rÔè6£[hŠ˜kK0ehx $éN#E¨°Ê±Šž òüÑQ¢nÃ’ fŠ,PEÎU?¼-1‚D”#vöñâãO\–ʬËzûPlëÙ´kTëЀÛþ‹o-ʵH׶ÂÄiÉ¥nMhƒ’)"sJõ°ê/Sˆ¢Â R&[-!½=¹ðëG=&¥yÔîº]müÝËr¯;×ðfa;”­ßAüHàB L”o%hQ‚ ê\EÐ…ùø‚†:ŒòÌ(îA…+:l’M (ÐãO'" ¡?ßqõNBD $Ï òt“=KœRŠ¬#:ä !â´èSŒB™#V“j4)Ö9 #%’þ(i¥•L¤`ƒ ^”`„öC¡?|ð±Ò1ŠÀ„RlÒC6* €häãÏþ¸dÙ“P> e•^Šém\ꑇƒ_‚y‘˜R¸†0Ç 2‰‡NQa «|rBNñ” ²L¸Ä@.!ÈO (A‰,çð$‹éœ@r–oäÀìA)ŽA”à ŒÒÅÈ2E:*r‘š*_ÊðÈNAÒ„“̃hÃ5žRˆLв…€€ŒdH<00¸ M8Bh Ašv A©Y "K‰Ë\äv É #YIÕ!Œ\yßýÁ÷ „Á{˜<]ʈ;œØb!…þSÀdÒ„QK½EKšä\$/¢I 3˜“¬dp®q€<Í–8@üÐR `™Ä![92DPg«B°¨ç B@èQ8ÇÙ¢yþó ¨A:Ù|¤S˜‡cçá‚w¼ã!ßÎÌПŒ™E "4¦Fñ€E©@=I,ä`C-±¸•D¾t¨Ð:çA|ÙÎ`Fò£q€@œ:š½ã¤dyGJ€¥ `\}Ú;äU°ä«Q«XBV„µ¬h=ë•ÜšV¬U­þà5Âó§Nè ` Å#’Qˆ{¤áðÈøæ*W³²õ­‹+b«Xºþ¶µ±p­,e/+YÆfÖ±‘Mìd7kYÐbÖ³š%-g?kÚЦv´ý¬%"ÀUª6d¨í/ËPÉ2Ô–—|ªo B 8 ¢ yÇyä.DH à¥zìàLp¥± B(JYYx]s¿ûR£Ä,oy%‰[óâ68P…ê@øáj¸Â­¯}…;‚ðÃi@~ Bðf‚EÕðÄó~ (ðèºÐøƒ>æ.è—Çln…›+[¥wÃ~¤zƒó‘Î÷Jˆ5F,K¼oÅ,†ßXÝç_ ú#ÀÍÕ•@àa£¹âmH¸Ã:̇¢Q—2j‘‡ºE {þØÃÆ­:ƒCx. ô´'.Ü© Ä aîPùå/x‚3µ``¡œñ‡tð Ë:$xßÞ‰tx©fò0i“Ò•"ÄË/}®@4¡‚nì€Opå(šAˆeœƒHº8œýOñ„¼ê¬³“/ Ib^ñ}GŸ':`8PàXBŽ I#‚h¸‡9l éXëRÒ!oÈpë ã:×¼&ƒ¦…¢â·x Ðó?í8Pz !Áo‹:a50Ê°ä$îP­=Ô ·¸&C·9œÞ]ÿú'–ö°=mlrê*O5’‚ЊéAûøC?Ðð¬3LXÖüÆ%­b€]ƒ{þ×ßNï¸[óir2O âY™ó#ù´áËjV¿/®È Dç5‡Å ©H`¢ú@[f@#pPnªpF ÐÑ«0 ã4§£ÆýakoœÛ÷5¤0òbîÏ-8ž·Qè¾ðJ0ýO,¿Tê/Mò¶yîŽ÷ÖJ4&'˜ÿu`X0ÁFÅ$æ‹ú¼ú¡äìôDÝ>Q«ã®Ût÷¶Îín÷;g)Ḝ‘uü± !8Á) D1Ô`Ÿ-à£æŒ·âÍ~÷ºã=òg ßirØ;ˆ7Ä»©Œ5ا»/½ø/yç:ò·®|e.¯Hdk¦:X L~|,*ÂþÁ0½ïiwsL~ò·f=©bKóëž8 ¾ñ„Þ|¯ýhl¿£×ªûê‡w×KöYœQÎä‘[T £ˆ„ –‡¢á$¿¯ÿß‚o÷0à]ÿà÷¶øëB~"”yò«ÐJFpÀLÇf3gØ4¨×ÞÆè!—nt¤"¼ oî¡«ÖS%h1Áú§‚8[‡$]·Hëæ€&hÞ ,EpË0ú€Ö`pFu„q7¸‚GèsYB(Bã‘'ͧG*w æ0Á€pwå´…å$w !|“·‚ׂIh%X;>Dþ'ð :}Iç ;h7e‚MÓA¿‡dPû§‡Ã÷h†±ƒ6ýàt±0p# ÏðrAæÔV‡#Tüöx)¨‚eh‰—؇að‡(Å„á“ "À+š ¥„1 ;µ l¦o‘Ø4“(k•¨‰™¨‡˜¨‚hÊGGýc+š  zT ͆]‹  õaH¤õfÚB°¨ŒÅ}C‹—x„˜h†Hˆ|ãgúã:° 7xRÁ kXèŠP‹±–‡™h‹ë˜‚¸Xð‰±jÙp@}=& ƒ€h€â€ÞUŽsŽ&‹h‰ù§ŽíˆY¤ ¨¤yÙ° ›Phìþ Ph@÷y)¼°Ao A4Œ'‹`™x’'™œøsñÈ7ü#š‘MQJ—hj ,FBCu¹‹<é…¡*©’ Y”–èŽdƒáC=á<´çyó– ŒÀE1Ñ…mç“‘•Òô…aF™‚EI–aà‚-rÓ'$]¶€S &¾À:?È‘æh(8––H”y™‰IÉi9-6Ön:` ÷xùH õA‡uùŠwùµ’‘)é—[˜ÐÒyÒž`=B…{°÷°˜Œ)‰Ž¹“)–’Y™Z¡‹´fýh®$ “аêpZÀv‹þ•Œ¹4’ ÑŒÿ¤m¸“’Ç™’“i’Ê9™«)—‰)ÛÑÖaSñFŠp›l@K:IšÝ™%Ág’É)žãÙœ I;MéòÐÀ˜ ¡ X•EÞIŸòxljšD‰ŸËéœvá’Ób'éyÚ°…9…° %ŸõÉ ö9¨ ¡á)ž}Ù’±sJ¶B'B° †™]C²  ”䢺4¢ºTœ½´œûÉœÈ ýéK8v¢ üК° ?ˆ5‰få0Xó—eÖ•¹¤¹ô•¡_ ¡Iª¤áé¢ H’* Bð£ð  Kç€ nÖ ]ª‡¤K*¦&Ù¤!ñŸþ˜Ò ,’3Ò Ós]©ÓÂ`Œâ¥ujÁ¦_¦cj’_e  \gdOP¢fú’¨T#ØP DÐ#!¸Ds½i§“Z`ʧKê§Ä&bA5C g*­y13bˆ¬€S= ÐtþÀ1œåôªÒ«Òt¢`©§z*¡{º¤ìõ[)&ò…p j%}R£<ƒt8¸ƒpjÀ¹@” ­•qsðXð`€¤Êy«a€°uíe—¨¯'¬-2tƒ#‰H‘,±(W­ñŠ`º­õj¯·Ê«þà^WâP¨!t“™ñÀ†9ðEÐ )Žt)¯ k;Byþ¯»­ ¶©qfD@g1_'f„IÛ´Ç°çy@ö¯Œ”²ŒT«F*±/ eR&Oõõ_•±”™Â/Eá2Ì€åà£ÿ Å ž¸ž!ÜÃŒÀ"¿e¡+üà‹› ŒŠƒþÈ[º`qQŒÆ|›«¹û±>À°(t[̯,ÈW`˹ÛÉáüÁÐÌ-¢À¶1tÖQP†É„0Xøk•¡ÁY4Ï"TÏ"¤ÉáÁÊ ÎÉ|ËYrδ WºyÜc@8YàÃþJ[J½H½H:¼Éþ,ΞüÏgøÌ[¶R£:`ÊìLjoà ¼ÍÜ<²ÎÞìÍ*ÍÒ/DZŠj,  9 JWú }`Ò' ÈÁ·Ò-½Ò­Ò1m3 §4ƒ*ðn§šºLaûÔ(ý ðÒD]ËYÝÒH]¡K;ë<%,=¦ºkpÚ\˜¬HlEnEÏh É<Ô,M×+íÕ˵Ñ>ñ\ô(hSv@É0K~\ÕÆ,”â\ÔCíÒWײב¡ö'9𠆶jj@ £yØAØÞlWÚ,=Ú¢ Ó¸<6V£;à JP“Ï—}ÀÙ ËþaÚ¦Ú¹]Ú§ Бí`öÅB ÓHp¥å ª!Ñ?qÏ ´ÜáÓÜáCÑú¼Û\MÚ íØ?°³±͇#P=}k  ÝÉMGämEæmEÑ-pÛÔ]ÝVÚnµ‘Â뙧zˆÒÄEKÛ@­Þþ€Õ¤Û·=Ý×½w¾½¡† pßz¼j©ÖýØWíÞîÍòíJÒ¬;@“ܼÀ°>á¡ðâ)®â*Žá`}X#?®m“0»³­pmE<B>Br­ Û¢âF¾â®Ñ¬fk(@€ºƒð mpÿHâ@ýxI®Û®Û®áþž±²P½„Z@ÌW~ÒÁ7à¸mÝ[ßòi4?n‘Q¼€vÐ… Ž¹på¬æ€œB¹â‡~äð]Â,¤LÐZA¬7­¸ì‡ ­öjòܵ“鱳鱓Ïg pè è)Nê%iDeŽŽbeŠÿ¢>À :]› ?Ý"ûfÕäá£ëo&ä`ˆèîè‚ÞcláPÖñnbœºA†ë-r³ßÅ—ðëßuíà5n¢NìˆÎÌj%³ïåÿú bL«ÞTÌ~¶†ŠjfŠ°‰YíÖgü ÛÞ\Ýþ]šÆõPá.îõ þ!öðžb-äºUþîgSêiºz~ª+êpÅ)OTÚþh OTãvâ¿âùº¯½úï° $¿Þ@ó €ópó8¿ó<Ÿó=¯óÐ0ô&U€0ð@@¼Õó@õ9ô>ÿóWõUoõZoóY¯õRïõSŸõ`¿õ_Oõb¿õd¯õf¯ójÏöiÿóhÿös/÷uöwoõ@÷x¿öv_ö~ß÷|¯ö7¿ó…_óYŒ5oóÀø„ô¯øMÅò“îó42_Æ>è›üp}ž/õ ùúJù+üRšä˜Ïù­OG ßê¤þoÁu_÷•_*fl¬ïú»/>Ô êî¦ïÁfn.6FÅÿ>þÎûËÿfíSûð3\Ì/ý4§ûÓoýVýׯýÍ•ýÛïýÕýß/þFæðDÕ>+vODU €\à¯D…  êéÿ]ŒþOþÝBœäô2tmÈ €š?„ .dØÐáCˆ%Bä€\ŒÀoâG!EŽ´`äI”)ZpQåK˜)'¼”ðRKK1uîŒHp‚?3 xäY´è¥‹9ÞðÓèS£,]B¥:)¦©Uµž$¨Ô&€¡[Å~´ÀÁB¼ŽeÛBÓ…H ´¥ ‘š€²þ&ë¶}࿬{_¬Æ0°à¶Ô.,qÝ´5ò;ü˜í;(ëµ\õw€9w¶\­ i¶*¤àjĦÀfA@áÊ´aΔœ[÷X‹ŽSÍí{¸QÁS-Û{ôr”eŠ†>öE¡«ó,þ|;ÏëµyBBãäÞ}Þ»z˜üÞMàz5ü—Ýñï´TÿÀýýNz'ÆKï¸ PpA8èƒ:>j´p:„¤ ÃQүÔ0!H€ ÆQ$‹.!Š:üÒºèE¨)FÏ:Ѥh„ÑFÞI‹DoIµ¸\2¤ð€ò‰(¡ Uh# P!!=2G'Ý”ó.° ­TÒ0AQÔóªÔÑ„ø©Æ¶M#(ÓÑÖ>Eµ!Rµ ;gpml-matlab-v3.0-2010-07-23/doc/f1.gif000644 000765 000024 00000035352 11422240326 016617 0ustar00carlstaff000000 000000 GIF89aõš÷¢ !!!"""***---///000111333555888;;;<<<>>>???@@@AAADDDIIIKKKLLLMMMNNNOOOPPPQQQRRRSSSUUUYYYZZZ[[[\\\]]]^^^___```aaabbbcccdddeeefffhhhiiijjjkkklllmmmooohhxpppssstttuuuvvvwwwxxxzzz~~~kk’QQ­UU·ÿÿÿ þ þþùýþýýý!!û''û ü""ü//ú22ù11ú44ù77ù44ú>>ø??øBB÷FF÷HH÷II÷MMöRRõUUõ\\ôïaaóccó``ôjjòkkòppñvvñzzñ}}ð………†††‡‡‡ˆˆˆ‰‰‰ŠŠŠŒŒŒ”””–––———˜˜˜™™™žžž£££¥¥¥  «¨¨¨©©©ªªª±±±´´´ººº»»»¿¿¿‚‚Ã파헗ìë™™ÿ¡¡ë««é¹¹çÀÀÀÃÃÃËËËÌÌÌÏÏÏÒÒÒÝÝÝÆÆåÔÔãÝÝÿàààîîîÿÿÿ!ù,õšþE H° Áƒ*\È°¡Ã‡#JœH±¢Å‹3jÜȱ£Ç CŠI²¤É“(Sª\ɲ¥Ë—0cÊœI³¦Í›8sêÜɳ§ÏŸ@ƒ J´¨Ñ£H“*]Ê´©Ó§P£JJµªÕ«X³jÝʵ«W¯3LK¶¬Ù³hÓª]˶­[· "¼K·®Ý»oå2`0V.Þ¿€ žAÁà¶önŸ±ãÇ#Kž™“0”$G¨9cÆׄA>SL”HôÄÐ$CiXL¹µë×;!AÒH3gϦ ¢Îý4oˆ»Cª>»¸qÉžÔ É2™Áí߃CWè{:Cég:ν{¥Ù™mþÓìl†u†‰_h^äŒ9œºË/ ˜Éœ6Ç =Ôú…þý‡P€ Í1D|ó%(Yf”€G™~3‘'à„¶äG *¨¡cn€Rß}”9·_…$–HÒ…¦ÈaƒH„7™Ï™(ãŒeâŠ8‚â†r RãˆÓeB£2ÉQ&°–ã†m̉kÊ$!TVÙP&«-©"H\ñÚògå˜d „¤’Zr׉l²©Ärmv(˜RŠYæ3b‰fš ²ØÉ—1BÇ ƒÊX(Fgò™"°ÑÓ”¿h¢¤$RJ‘žŠnèÉlÀå£vâ)ª€˜fªa#H(Qœ£0A:ê«þÖ©f‰©f« 뮿Í@­ ~‡°å§+oFš˜,‰ËFDÃ!œdlwŒæ $¯Ø~æǯÓ&ømÆ}Új¨Ù–kU(spÛ­|,zbœˆaš+ïU¡L îºÝ¹Dljû’«¹j¢À$Ü÷âËÝà†{ln–Vñ„'4ƒ '(ìŸï><ïÇKÍ`GÆ êË\¿ƒ¬rQïIK²qœØÊwþºðÊ8%òËóɆD%Ç{íoÍVXô„G¤ <ÏÇH—ÝÁ[gÎT¥G MÏWŸœ(]õ×8]µ|›þuÊ`§ý’cwÇÉ#H­kBÇ«öå/Õ›ðã­yIÜ‘¥€qkRƒŠùé,ÙË9wÊñ YX·F7Kv£n{v#¯~Ã>¦Ç%üýXéãZ—4©y*Í´îÆm  œ±¹c³¯TûíØ#z#óÆ™™}/?Yõ*]Ÿýù5νq¶rÝØ!s¤e•Oþý•0ýú‘• ôc4 Rñ Ãy…oBàz 5;‘A•— £dâ@Í1`W·¸ù'⃠kG¸Ç"8óŸñÇ‚ÔËq"”Ì·žÀdÍ`V•!_JÌ× Dþ1|Íç™ ¤AP¡×|ÈÄ(ˆ¯Ñ—ësBƒÄ.2Äû¹Ds¼ÿtÑx˜)ðÚ8†ChÖ ` ¢„‡M<ŸúÆHç=|ÑCî$çÆ“À1ŽØ ÂvèH4,'2ä …„×G“ü¨c!)Ó>ÈTñ 7Äâ ·¢À–È8œdÿfó?Ç"~á†^·I­pð<¯Ì† ÊÉ Jn™Ci’¥36²$„¤ÚêUËÉ(G £Á.BV.Q˜Ù›A!Š)C$Ѐ I¡c²h³-Buͤ&dÚUÄ…øaù%I‚‰•/žÇ¢é¥89dÉ`Ç…ÛþcLýL÷Íóé–ó|Œ­ˆ \'RçHØÙÏ•iàÅôÇ ãͨ2Cܬ[Eúµ9 3 Ž1ä%gˆ{dA`B+Š (PÀ (ÅP«tò<7õ 1AêÁ¹¯1œ*â‡Ð`+'S h€©XM«Ëé(. €çSÉSÇD.2uXfCô Œzd˜©@BV1M•£%‘ÕªÊN,²«Œù.#£L‰$U"LX(‚¼®# ƱQ¨h¯Œid¢7#š$Ø©a‹X•$"ŸCªM¯ˆ×2NƲA­õ¥ÃJE«Ð9þZ(" $»Â))eWíÈ×ʮĩ0@!6ÛYš(öŒ›Ð]¢È2Æ{“© DT{’঵ %`f;PÊHÞµ`w¿;^ñ®•¼ç5¯(Âk¦ò¶7½ï]/zå«^ööͽ÷…o~é_û‚w¾þÅïë àøÀíµkd„ëwÀý5p„¼_ó·Â‡z¬se†Ÿ@¸Â–Øn þzNø! k‹»éUÆËJ(‚ç\‰Næ’©}‰P‚¬*9ÞîRåwÎ…!O&¡úö%šÈŸcˆD›á„t¥òÐ$Wr2º|2I‚€-$Ê¢˜2þ‹-’‰PÆzŠöˆWŸ•’¯b}uC’ÂRª¬õ1g×,’9 ?fvŠ<ñú´‘J¦=Ó=ÉKK@ D0Õ‚&´KJàš¾>%œÎõik *f’d‚©eÖÀK¯›é É œnÍ&zü”d"Éþk (²g‘D/0 @ ò¬fMG„!œ,†œ¢Ê$ƒâ–®!u¤Ÿil’èat­±­™2lC–„®´AzÝÍj— mS°Œ×²]S28Î1µ?dA$Ÿ¯ñÀmu’‰dwU¯®qr©ùÙI*¤uMr«|“0ÉÇtMÁëVš;#… ókAî ÌØÙž˜a§þó¬g‹_ü"AÐ-lì`R  9ÉÐîtE:N;qŸü"O ö]“ {žàŽöÌM~ž›&Yø6ŽÀ‰rQ\U×CŸ÷Í7’€vD‘DÅs¡W\êSψ‹å³í¡°¯Ao 'l^š[íaH—å@¢Âê-›‹ðoy[ÎÞ3:‘»£L%“0‡Úk–>í¿GÊàIxkÌؘMÀ]&gïêÖ`æ¯;>îu!SöŸÜÝÙv†§^<Ð{Dƒ„ÍO‘]o)©¼«"-N ,âöò]Þõ »ì!sm¡äÜÙ_~M¢ýn¿Ø¾Z±.ÎOAa‰ß·äôIN½êyþOtà/d[‡ÌôAQzž ޹ЅMï²þ7Zßû9þdÖ'.ĉM HÃY’yž2Œ ¯± -×íçGïlÇq{Çv0\>}Î5Dî Æ}`׵ǃRu°Q nÀ0HðS!öçY’U…ž×|çA¢`h®á ¶2³ñØXÇò—wO÷«ÇzZ„€1·¨²ŒpßòS0ÖY]§zû' èHH„Vø #'l79 Ô)‡|#xËÇ|§…áf­1DÓ !w†ï3€3Ñ¥}º‡sÝWFÇE¢0x’Q%Ôú‚ƒþ@õ„:¡€I–~űƒBXnlx)þöÂrg 0‡îs:ÑmÎöÜq€WLYˆ€‰ qQ-qˆa¸H:}EÇÑwç‚“±>ón¡‰Ža XI×U‚3EÅÂx‘¨QÖ±“R”¡/Ž¶Š ãAH'”}rC@rjØzçA°sHyâøƒ³†Z‡~Šwu‡Ÿw‹1ŒŽqKa¢F,HáÈSãØ-€£¸N¥’ç›Gô›$±v‰‡ˆÜÁmׇl˜n8JfC¹æz`‡*qmIÖ…ò‘…ÅÆøp6Y7Bt˜ˆ9•þ À±FÎå'òQ’¡“Ù1À'|©88¯‘¡m !+’ уĸ/ò¡ ¹”©…4&¦5ƒ¤B Z¼D Y… Ño5–ÜÁ‰îh‹ðÈà'ÂR“‘|ŽáŠóQ på§@ÉSÐ8}-¸†H#LéOÃd®Ñh‘a”a Ò2>Iò8OýHw4 s²ó!²,™¢¤W€ªtw¦ãö«S”œ¦‡`©Bá¶]A¨,rZ ásÒèšÝÁ¢Á` %,ä¨!x+«qAq«~@W~ Á¹¡a82:«¹Š‰°›ÏUO‰9€Ù©´ò±¸×Q¤tD©8ò³…0¨–V2K³S  j0%&)P·BšÕ&ƒÑ³Ñq¦ [œð4O¯Çº ±µ”„¹© iþx©»¨P „µL– Š¹92?ÆV†áµQ‘ ’·¬ ›¶tdºBYMK¡0¹ÖSwúRJ* €¦>Æ«Kb Њ>Ã(,4Ú'ølYI™A¨“d`¨"Ñb–¯ û[Ö¥Ùºj, 0e0hÖ»b”#~°¹6µY±†²ÚÉæj¢ë¼®1ÇÂKM,’¸¢µK¯ Á vàÁÁU‘K}FÀ Á „åc¾«"˶fÃx‘1¯kM:„½(*º9¾0»À–ÄK|eàmySœ"–ð³VÇvdžÇfÇyÜÇ{¬ÇþȆaÆÇ6Œ"—$È¢À‚ƒƒŒJ?”6Èà%ˆ, ¯ZÒŽìÇ¡b˜5fU0¥ÄŠSÅV~&Å™RÅň¡¡‘!ÄÑo.£‰‘Œ ®\KÂrË r½Ø‹kŒ-ÅT`Cq,Ç™iz;¨·Ù3úhZ%½CdŒÅA®1‹Š„d+†™#™ÀaP…½k*¬ükÎÙSQ—ñÀŒñÈlKò ú E‡X½)½Ý(ªñ´C¾iR¢ÅUŽÁ0ÅÚZ¬©t)ªl¢£·'@L ôC,€9 û¬'q»Z‹­%UÐÒbcîþ‘éªÇñ² qÑä #Ã9âÒg쬖(l*ŒÕYZ:DûÜhË0kk'tµÌ£Í«(¿#1X 8#&a.|KOÖìa mð-_M?WhÐúƒ™ÜÌY}Ö¿ý«Ê´bÓÄn5õüM'~“ÀÖ­asÒD¢¢|bWÎ ar=×R–Ì8 ÍDiš¾YP1Ø"σ-B2½ó¬"™©ØqÕi²×ß„ýÙ·“‘ËÀ±9Ú4­×ð”¨õûÈ8¼$Š-›+ŒYËå ,çônˆàm;Ã)¸íÈ"dþÛvÜ¡}0IOãŴ˜” A Þ}¹#¦-$¢Ö €Å1ÔaÉ»ÛB­³6IG,âË*ß|ˆá½^m\¨&ç,LàʉLÑ7ð Ý3M9k„T ßRÛ$-ËóY; ,¿ÓPs °ÌW Wà 2ÝÀ põË0Y0ÚelÝd:çíÍ EZD™Ò~Ï¡ûÔ­1Ù º"d+O Ï©íémNàé[]Xq“áã qšóà­áº ¡º«£-Ò-ZfäQf A¹‘äZâ; I»l+Í=AK?ÜfŒ]cdH 3-žØþaÌÀVæ3Ž#S*LŸÛ‹¼`i>ÁIO‚ûÞ‰´P”„Ì:-RãSÞؘîÄ1ÜÝòž€¤Ö òUçž}³¡}Öu¸D’®0oîÑÔér#Ìàž¹nÇm)… ËâAÁÁÒQ¾=Ì?ÂÒÍÓb×7b¾P·ž"œ0è9±äR1xÄë³±hƒxH“åLÍ3.ÚÕQ€ä½NÑ®"•ž=°¨¾ŽñîÍ =­á3m ØL\ AåÞâ$#¿|.ãø²vMŽ3*ž¡;üM@Ô-ËW|³ËÌSî<ž1é]€²n]ÁéÝÂì÷j¨¢ìº þãGÉíâØ„’îJ“Ýëã \2äð‘ dÌ¡)L鄾 Àæ”ÁÏAå­a+µ-Ù ÑïÌÓô2Ï3 ßß±cå}ß-ªíCò·¦YFóce§‡z ßÔNŸ5'õ9¯0ÅGéÚ’‘Ô¸”Áºq¬GD§«#kß4u/ÕAæ§Ýî*òê41ôS‘O¿ÞdI¯ÇpÚvz\¿:e°ùMsäÐÊT‰¯0zÿÑH#>îMQ&ÏÝí†ê°6îëB ¶ß4½þ/•Ê$&sß-–ßC× :"Ó ½Ó ã nðeðô¼6¿ñCüf~ØsÐþ4H(-ÏOÔ\ØÎå‡ÏÜ6MQ Ô@‚ D˜PaÂCAD=„QâDŠ'T¤@ãFŽ ì`IqF …%M–äÔ0äJ–-]¾„SæÌ•38 Œ„äJI4Y±d²’'ž Òà$1È¡›'>Uè) ÑJM¡^Åê'O¹~•8ëX…%Àž˜ íZ¶`Stƒ¤ÌB•m!êésò 7ý葨ì`“R©F¼†Z»^íΛxpÏÇ•-_~,x $hÎþ)d™“qŸ´äb&Ò’]~"Ðõì”à eGÃF vB¡L*D˜Gî2TkPHPRœÔrˆ“žö62h⌠ÂÖí=1§•³Ì€ãÆŒ¿¦ )þ$§@ª“ßÇþl¢!7+9/IÌ43)®, âÄ6Q0°J¾§ÀsP2;/¤PÔ‹ ßB)d·~ûj¤#LHÁü0’îÄÊ2™À?Àšƒ$NÐ@âÀ…L´k©“þC¢“‚ˆ{h!H|pªØˆ$,5ó\ í$*û|‚oD$ zRÅŠ@Ô²-=J*À¸j 5øZh“) d§®@‚‘‚©þã¡6¹²$ó+)tI¸Œ(+8>t¯KG- ¥æ@“+áÄN¡61[Î)H“D ŃEÚóÔኴ%&¤h·DãSu ¿ŵ²9¨RšÞÅçÂCh1ã¨;©ç|$¨<=ܬ• T£=©¼%]…u"YC¤µV%s­PlB(Ve–¢x:®„§¸@³ÙH­\tZjMÂq­ AŠ`PŠ À­nkÍ×ÑFÃI;èµdT™i-.0ÒÌÒ² Vˆ¢D?€ÖÞ{Ù½–%?p8,þäJQ‚†Ù' è­P—æÐc³xòcŽãþ Â)e‘Í@È9d#‰¹$5-Û¤Ý4ÐÔ<ËeU‘Š¹ë—È[èY™Üh¯H;Ø2h>ˆ³¾@é+jEnZSŸK^IƒÝÖã›#È0>uŽs½&#Šz%SAáT„8É3dOzdc¸AÁ³VºëNhˆÁÛJ<£¾G×èï‘ëC8“ Æ,¤Å"´—Xh¨Ú$eãîtJX$Î~[ó©È”ó“h°™­ÐOÔZÕŠ-d??øšÞ?N$Ås ’·^”2_yRIJb*žQ×þC»¹3JpCÜ0Á 2‹Hœ8ûÚ7Ÿç)Ï%¯ÂȶZN°íJõ»Oº3@认ZÛIPç’× ëFø™ÀƒœC‰ôñ)ƒØ SžµF."ýXӌ娞d¬{!ÎLX#b„êBBš\hÄm=DÄ ÊŠ·I"HÇ7Dq«i~ŸŠþ‡ê½ð!<ɶò“øpfb ÑÃs¸Ò  `$’ÆvVÆ’‰ˆi\#GGÓì[ZB2ÑA¯…d'Ù''29,,ò L"†ä,¶aQC9ŸãyCø3ÍSUïCÈSð+œÿÎþRˆ9LA¢”Èë”ð;ZâçuzAÂø„EðÍò$(Kü"¢5Ùe~#K£rÈ»}%A˜ ‰–2=…v]W94 $w'šÜIÌ'Í+qÂ|6²¦Ó¶ºpáòTÈÓ’ÂJ‚L&y >Ú¤I Œ‰š*ŵ³\By¢vžDc%’²² Ë€&dSµtU$0«º áZšãBR L—0p ŽlË3Ÿ¢ÍÉÅe. I/‘3híÈ9$uŠRR”áshdI$õÆÒ¦Í0?,‰Öù^†Ò.‰V<I‚pâÝ*ªÈ“PÁ’:q$¦–d¦˜(HTþBª6­¢ƒdŠÓº’q$¨máåS.I‘D¬²9ÏQˆVµ¤Ó’œ©šŠ`˜V׃ÜÏ8y Éɺ)ƒž ›•™Yµ|bS‚Ôe-x„Êû Ç+P`q4Ý…’Qñü³ô,^Ï¢€€½p/Ë1ç;šØ¶ñneRX¬,‘"“ÓØíbÕõ!/B[–|â§ß»x“ÙJÝX7ë&§£%iM¨[É¢…µWq­D2 óqA!ÑíÒhàªAáÀR H-/}Ÿ U®¨§¯#ƒc~$ú]®d¿©°LÈGN’ pâçAñ×ää³@rñ5ÜHéž´ŠË•þo6-¿Ç)•;Ûë­&pÁ3¡ìU;‘˜íl媄¡’“ÓЦ{éŠ2X0’ñôIiOõâýa¢ HÉhk²øž¶XÛD@–í$$NΆ cÃ×@ÁR&—åkåÑy&CPoXHøe" ¶Õ£IbM’á-¹ z®ˆ€ ܈"¬pCq¹ˆAe ¦üŽ‘ê‰gƒÜ½8”HŽ[«K$˜@‹iXÝ$ƒD¡°Ã2Môà‡ÄÜ5"ŠœìIeä`*Si3a8CWR#žÇ4W±7Uï¤ÿÒÀ&±¿³8:l[ÅHLRæ–év,ôYSé\MþL`Óùi¹Òç™0,^³{ íäÌ`=©qÓÂY‘Õ äÄ °rÏ—Õ$цÒxÄ%I×@i÷»óˆ!<Å gúã`Æ›o}„ÑúJŽ‚°òÒMD˧âd &58ÿä¹îD¸( Ñq“0×'&ŒUqCbÜÝ¢Ð8~Œí),Ž*.yBþY˜so2örLZ.èöµ:]ˆÄ<ë±øú'+$sÍCÒ¦ÙXäÒYú¶ôîµ$}™Êe¯R¹ó:Á‹EiKü ¾M&[H V]«“ºÇ…nŠÜ ÉC„ò«ñvVY7ùÊ‹"óñÊåÂùÕXþ>)ïE êrL½°u ±&ýèM_ú‡ˆé³—}í;OªÎ‡b¾y„ZÝÖ®õ*“pÎO£Žß^÷´ç=ôm?yþ¬<3^I&42ƒs½Y#t‡â·³xäB¤ÓòÃ&À»î)>2m&®Ü4®µC ÜÅg·]ç*O®‘;[¾V;µp€˜;À+«¶ö1îÐ(p[ ý#‹Ó…s r“ °KŒú€À J?Q¨5- c7¸ðÞ2 Œ²åKˆœû Ð €6¢€—¿‡8ör c7ŒˆªÀ–3”89+Z¿ñð€g:“ô‰YÃÄ8 àþ¹·×+‰H°3ü¯ä-¾Š•@‹*2¿ º6 LkÂp*º øƒ Ž+ˆ¸ ; Ì Ì8d ó)$XzA/<ˆó ¼•¨AŒÀA4C;!¿Zy4ˆ±Š‰ûŠ-D¤ˆÏ; ä*yz"·+œlc7 P à•àQ±…P¤B4DyCä­(A@‰PÀœÛÁ·‹Ä\zˆƒ¯±‰#Ä lꔫHBŠ83ï@>íòD¯AP§MȸÑ#zÊ CˆN0ŸXŠÅ¹ÐÒ>”é>ð{ˆ]|¹­)EГ N4 C³»åʵ–GÄð)‚‚AH ¯þá=˜T´„ÈFƒð7Ø‹ó –o¤£p|‰Áë¤9¼GäÅ ª ,–ĈO" üµ›Ë@µ“ˆó‹`ˆ©³#ŠP4Pà», „˜|„…¼2rÈNFc `ꈀ¬«H€ëEU™€–<ˆžÁ ‹‰ à =#Fï 3S[Iˆ¨¸‚ð¾sŽ­|‚¾ð»’{60,¨AÛ ÈGPA $ †ÃˆßsŠ:Ø#—à±­›âÈ•¤Dœƒ½ÜK7Ä›D%~—äœ!°ƒ§°@ˆˆÊbÔ«%"’„¥ô‰II]ºË®™ƒïºÉÄHL´€¸ÇXÇ€Ò£—˜KĘšÚµþÁ)$É”…k“Ö NhÍ” !œÁ±l6Nh?pM¬ Œ(M¬¨=Œµ”Q¬JˆX:Ü ÙÌ\ Í€JƇÈK¬øÍ;úÊ d’¡ŒÅç ‰>Hé ç,@-!Ì ÌÆt J›"„ "Å8ËÁ˜ŠàÑÍÂy+õÄ Ù‚Hp¹Nk Ë•€£ÙˆÏ‡«±¼h §”ß‘ît¾ç¼MeLâM}£J–D»I£J5„Šô¡²c¬ÐÊÑ.ô4ψNuŠMÄG9ÐY"ÐpKˆû ßU~£”NNðªG)iàH`ƒ4]¡ö¢¨ås=j‘<ÑøMßò/ä«á«àŒsã³rŽ)hã2ê‰ Za­/q‚å§ÆÊ*–êÞÙ ùERˆXâ„hרèF7xÔ‘ ¯^Z9°qÙÙÖ7Ö‘&éE&‹JÝ2– QÚ“p-~ŠVŒŠr.ÖÅŠþO+â¹ìéص®=/óÓ¥fjÈÆŠ1Naژ㉠kÄÖ.Ô5(ƒ)ˆíU«©P‚³® LS‚•žD퇘ÄÆ3;˜æiꬑëñÊãž Xíï OxVAä!µÒi 2h§¡@»IèÕ‚À"uŽÆY¤½IÇ~l¶än Ä+™ã“xUt½¬8“Ž¾;Ô…ŠM H`äóÙ OpðtÆë +UýÊ×íÜLÖní¿Å ÉÖê<Áî“PÔnÝÂ/õí±8„ 8êâcçÐí2€“¢èm×°Kåçmá&}ë 'á æèoF’Ù6‰8Á«˜Z/?ŠMˆþ*ŒÉ ÜI@’MnÞÞ5UæÙê iUá»ì= ™Ö'ªÆŠ ‹ ±4ŽI^5Pƒ3°ç§ HnÆ‚UÜÌՀΕ0Fˆ!fñºž NØo§@X2¿ 3¿ l…Â[µ…sáðaœ¥^éaÝq¡àèF`î’@[…èò¿Uˆ@Ÿ ¯eSÉ2 v^`ìá+{V«1à9iéÌq\‰n°hH$h¾ðó±Ð“ýp×xâGGf¬¨(·lô(§ñµžöÍLX˜¹öõB•ñþzw­závÐuÂ@r‚@J‘põt'Wžsˆˆ×o¤Q<7ÖU‹„уnŸŒ¯+‰p×G2gâ«rëƒ x1øÄÙDØâ›Dnkwm]‚4F’M(¤Ö÷#% ,*ïCJ©9ˆ[·r˜aãXÄpp‰w®ÈöƒdZ>•g£?…xb/7‰[NŒ……OÙÚ"wÉXøä¦õf«ö\ú,ë–J`ßzG®Io&‹ñrzï`ÆZ?—=¸Çƒ·õ´õêÄ•¢–‡íS!·|f(Þðç“w×Ï´J_ë‚M0¡…km>zCâuMåö"´¿ŠZM =`€þ’Ïå„ uñˆô µq¬ö‰뚯§‰X„O# –'Ê4 ,ºz‡]7÷€`ÒÓOˆ9xn(A|ùè׫äÍtüÇÏ•#ØJ7Øô“ÕBà5“Àè«X[ ›¨½ þša‰Doñyº7ä›LË­ï\Ù'PƒNà‰§ …Ž³?ˆ¡ArïÆ|ît¡M2¦;Wo-ñˆ@AD,hð „ 2lèða &R¬hñ"ÆŒ7rìhÑÒ ˆ"GBœ@òäÉ‚<²l¹Ñ ,Ñ AR‰e$O:¹¼X§Â9C:ÒH$2S Œ›<$ØÔÓã£(9¡¬j•dÒ§Z·r­þ˜‚êÕ°MŠÉiFf€-8#P׸[ Ñ ’+œUùÎìhñWœÂ!DÌH¥€ÄéI(ùyzð,üå@±Ï/½M$3È„WÄÆÞm§'#=‰õÒYÒ4ÚᛕÌhÖéž.4!ׄL‘”ˆ¡°„¦—Ñæ/™ÆaJþ‹‚Ûh“jŠ:²„s\á:MèO–§¨Øɧ1>«æ44H]QTEBQB…ØL)M˜˜–Hͪ¡«ØQH·v4…|H²PD€©]é¨ çÇ¥ºÕ*Ð!úRn*1¥GÄ%Fn'Š ÊŒ9|hG(yjòs@Àg1ŠÈ;ÁJ—#)gÉrØ®xóŠ·óH {¬¶¡­àî (ØU®1 A¦å7¬™PŒ9œl5 YX6¹¢¨+f¾Š¡°Ž$UA.A.—†æꑬsCSˆ!„´I¶ /t/#]÷Ø!(AíFžÔ Pùq¼‰-fËOÊr‹þÍE/cQ¬ö"™ûh\K¼½‚¿–ýixE‘˜òbè¼NoYtÛ$Þ¢dµaËA<»•¿ùq¥‘ h„ŲSvYB·•(†Å²^ï(X!öUʇ8áT½˜x:ÖHSLßb(Äòs©=Zeªr:AüDX #‰p)’«bâ/ÉU±EqXfLÍVe­Ÿ°jó”Öª2˜þP–WsM²C ¼üe%GÎ6æiE°gcs>åž°ÌÄ›‹“‚qf9,˜U$–“ æ«h˜Iø; '"*Úº¤­•Ì„6á_įїAh“žcj£šMÌ°yp™Õ Np+Aþ ¢áâU‡F»öÀE³†ÒÎ*‹_„ÖÝ??úÍ×ògHê¬*ÁÖ-’ðôÓ?÷°oˆì¹rõì„âÍa|ý`ÓŸ{Ü_iÔ“`žÑ)`Í1 °`w ­èAœ`ÜÅ`ô¹[JRÇç™M’hŸþ–à÷åVçi¬™M Ä ¾à ¶Z >Åñ} ÆÀ ’œ.Dþ­™Ü%N°m'ÈYî!Åí B|[†ˆ÷ü Ö‰Ký9áªÉÒ·ô`OŒ^Öß!@ÙöI¡©Qa¶Xa/E‘ÖÒ_> Ï1‰ Ô_",_úAñ NÁ1ÉÀì¡êá€úᢠ" þa!â!*â!&" ¢#b 2¢!."Râ%Zb&n¢&*b$ö!&&b$jâ$†""ž"*rb)z"&^â*Râ+Žâ"Šb%Ö")¶"$¢,6"'2b,âb/¢0£"A#2&£2.#36£þ3>#4F£4N#5V£5^#6f£6n#7v£7~#8†£8ŽcåÊù©Z€Øù­9nŸjœß¡†jÀÍià;¶#éeB ¤ë©†1¢Àö?â#Å…®$Ö!ÏA(ëUIBš£AšaT¤Bê` С»ùAE¦‡DN¤Ñ]dÑ!ÊÞ&dF¦dN&eV¦e^&f6ãaf&gRÜf’Ägv¦h®ZhŽ¦i Yiž¦jNHº”ˆ(¬®D#‰ˆ(B‰¨ÇÀ ?žËTâÝR©ܤÇþ­&qŽDk€VÊ#®ØÚhD%®ô%Pþä¹@§ˆxaz £À§wZ%A”ÈZ¸ûeZXËÅì#wN'º„dO†‚jÐI‰áwÖ'Jç©„Bºà þ sŠÂnºgYÚØV¶Ž}¨XçªDâJ@m(u (Jìc*@w"h}çÑì'~þšc€Í€ GZ §†¦hUpèAx¨Bðæ„’¨–©G†ªhq²¨ŒN'ˆÂh{æ(I0 ®¤¡¦(Ž2(ƒ>h„¦ úèH HÖãh‘²çtâO‰ü'®'ÐezFz€a”†©‘†'ŠúhB¢§Ç¨§¨× $…¡Ûv‚…“Š©úh‰thƒÀ>ŠÈmJ§(ØAt6P ŠA¤EœH2„ ®8àç”J¨‘ú¢¨tD!\ªä¦k²NZŒÈ¢ K„jG©F*ª:Dj„V¦ª«^Ǫ¾f‰¾*­ŠD¬ÁpÖª®Ž)i î*°ç¬+±RÈëH;gpml-matlab-v3.0-2010-07-23/doc/f0.gif000644 000765 000024 00000011576 11422240326 016620 0ustar00carlstaff000000 000000 GIF89aõšõ? """***---111333555888AAADDDFFFKKKSSSUUUYYY___```dddfffhhhwww~~~ÿ88ÿHHÿXXÿllèhhÿppðttøwwÿˆˆˆ‰‰‰–––™™™£££ªªª±±±»»»  ÚŠŠÿ““õ™™ÿ©©ç££ÿ²²ó»»ÿÀÀÀÌÌÌÏÏÏÝÝÝÌÌÿÝÝÿîîîîîÿÿÿÿ!ù,õšþÀŸpH,ȤrÉl:ŸÐ¨tJ­Z¯Ø¬vËíz¿à°xL.›Ïè´zÍn»ßð¸|N¯Ûïø¼~Ïïûÿ€‚ƒ„…†‡ˆ‰Š‹ŒŽ‘’“”•–—˜™š›œžž£¤¥¦§§ ¨«¬­®¯°§±´µ¶·¸¶ ¹½¾¿À²ÁÃÄÄuŸIËT..ÎSÍÒFÈÊÕDÔÚNÐÝOÜà×tÉàBâçIßêKéÚäsæçíK.+öJõêñróàzèS"pà‘‚—l<ñ Á‡—™Xˆ-¢Å‹š&:aÇ!Æ ilÂñÇj8@¦Ä¸òÒH&%Ýœ I³¦ —Kb¶™i³§þO9óŸ°Ñn½½Aõ^Ø»ŸßO<ýû áw‘‰ø@ÿöè‰~\ð‡ßÚ‡à‚¸… FX‰ƒZ@èxah† þŠöÕGèµb'fa¡„(>R"'¦è¢"+^Ñâ‹4£3*"`;;žÓã&7V‘cDþ$C©¤GN‘ä’PÖѤOÂa;Wž“% :3¢:_~xΔQTå™o …™h¶©†šO°éæœeÀé„œ‚üŽžÚð‰‰MàIç _Ê„ „&ª…¡K ªè£UØÙŒå¨[‚Ó©6Ÿþy†XA¥ÀÙ¦}„)¦XíÀ©™ Ñì± é®Q*g=ºòj¬YfpðŽ¥Ävì³n0:Ĭ 4«ŽŸ(}„­%Œžj),X í¸m0ú-­þET €²¤0DZµïòÆû®½¶âûüBÔKï½ÿæð¾úòûƒ¿ýœ°À <°Á¬°Ä SìpÃOñÆw\1Ç{|±Å‡lòÈ £,rÉ*ŸÌòËÇü°ÌßÀîÕ²¡ƒ 4€P±ä>+í*PÁ´ÎVªÒ-mIÎn0+Ðy¸ÚÕÕ`}‰špŒRÿ@uмªé‚°c±Ôc“ )œ•V€ ¥ª€´Ûx“'¥zA¥èÞí£JÚÆŠ†4`iD#nÞ{14mGNçäGTn¹›˜¡9N;zU†{x5X –Øæ¬cÑyŸ·åþëËnû_ÿ$-ͶÎðþ‰ï“Ð.øíÄÃÔeñÈ !üÔº'Ÿüòb7¿Ìè¤g¥zî`¢>…_žhÍíñÎ;Q×£ÐÇNèøŠ–/}øM Ÿ¨ú×>…û“/ üìSAÿ øçÏDPF)ŠQö¥þù œ“AuýUoLàëžö¤ÀGxoB<à*ئj D“?¨k„½“a®×õ‘r(|aäb(üÑÐÔÃaÓXØvâ‚Ë¢ëcºnî†F “8®%nÂ~€âyh2ZN¼¢±²¨Å]q9ìD71FeBbBˆ4Fâ‹]d#Šþà¨E9JˆŽW´c„ðH )NÑ è#ebÅ.!„âãe0@ Є"<£!‡€HþIReÄä*‰L6B‘’p#7I Qš±ˆ“$‚ 9@î ”5ä¤/9ɼ–¿Q!ôØ?:–2äå‚T®Þ2•ÂDj„Å_z„¤%¸KÀ“ÌÛL­GD'à€Rߊ€ \¥‚„ÓnµÏaÕM(°@}£U1 ¢¸KýÍR¹‚f*i…wö-Íd†°÷ºÑÛ§Û°ÓƒržëRó$B¥*µ!`x mâZsEà'E…`)ЄÍ|qpf.Óþ MC¨t°ÑÁ :ú­ Œ“g»ht ,Œf”‚KRM=-eSÔ˜óNûºØå.—ÑŒdOMÌfFUŒEueWujU¡ºU©fµe`ªU»ª±ru¬h=«Z½ÚU¬¶U«iek\ße3ei@O êM•PÞ m>ý)ƒú„½baR¥jÀ`P l^±KHæ ›C Uï`ìL© Ί/že„¤Ì PÓh–\« `+_§(NnI8mÞ$›Y%•rúìh«5 è‘-^j â*·Ã¤ÎöЇV·í#lNÐ\çfºQ@.m€™JWI?*`A‘”þ[KPv;;lgÌ Dàk*o%B{ úÎW¹P`ï¾€#ùb·HÜýïq)`%¸Æ$Kð[¯%á¨,òoY›×¯Aq&’ð„S$æV·¹:•‘†7|Ç{|¸¹)yµ™ÞÎ0˜>Έ!aßûb¤ÆˆPÍ{Iâô¸Ç.Z承•a óÓ Cö[‘W¸«W ߌq„<Í+ ¾Tš1•ïC @­¿[že ,bsÓÅO¦B©–ü*v^LJ€Ï_¿,$-‡Ù=BrDë|çæõÄ8A…ìÜçê˜ ®_“!ˆÊB7 =8{½vS ×ÑÑߪþ)^Ó!QMËô€¤uã§ÓpÞI´º“/B׎f„Žb¹L„fœC™º4#¡R¦ VˆÁ jWøÁür±|l$™ØË®‘P{µ±Y0º„BÜhC¨¦hº›•Š”ë>k·'°1gJ½/TmÐѶ¹m¢ãI'9l¸Žwæi$ßÛºY63šo2ÙX?¦ÜŠØu$þ~7áƒ84h¥¥–Y¦…ðl>ë{~ÊÑx0¬ìg7ä&þ÷¥®Ý'ƒxw`p8ºm£”ß;Å,·œÌÑ°óˆ\ƶ비~ßúAGÑÉðòÙH\¯NDY¢þªþGüÏIι4 nˆ²p]×ÿA’O¥ur1RµFÙ6ûÔåC›uK?²•Û[öÙÊ°Ët^1“õQš?ÂïWs²Ë,»¸Wù k.9Üï~XjÕojÿ`žO¼g½{j›Ö«è(t-sIGÄ×ÁNÿ+ì\èA¥@or÷V .Ðï3[Ôׯàé¤=k‹†ð* ßÏMᢠÞÀ Â6ˆ›t8*à_W÷Ÿµ½xÆéq+ÁùJ]—²šÖ¯šU®kuë\á~òƒÿüâ/÷ߺþñ·ßb1ÁPD`q˜@!è7úóÿ~þeuÅ8ó8€Oå³z‚}«FþB¦× OÐQ‚fÔEkþÆzZ‚yħ3íwIÃV0,†mV…ð€ûvRÆqƒ× z[¦‚ö  ˜#1}O@ƒa'GP:ø‚ÁǃF€q®¡H+˜€ H„χmÃg¨G}(f}I°ƒF¨LQ€8Cµ|Í… µ…òtd8fˆVw!-¦yæ%ë„‚¾¶2x†i‘†vx‡?¦H˜‡E€‡.µˆ`|ˆ†z·Q‡áƒÐˆ‰h<ЇFĈ¡ˆE`‘wE’8kH”¨J—¸›¸¡h${XJfZCaqBñ‰j`‰P0z gˆ<68®þhH™µ(·ØE¹øŠ8@!”D¿Øo@ˆ»e½Øʨ#²tw°‹g@͈‰¥8ÔhM×ÈDÇ؃®ñƘb@ƒ£ˆä˜騆Ñ؈è@nà*¨;°¯ØkŒ¶:qpv©8@¬8Û¸lá}~0ÅVÑõ j ¹{Ü6XÙâdûˆ=Ó#Ñ~¸YzàDñõèmYB¹Œ<}'™2KGíx3¹-‰®“^à>Ô‹ õŽÓ”0 àëø ù<æ¸ «dqŽd%ùI¹”Ès“ˆàJ²”½’þú†•v°-[‰U9–Süèeyg©{b©m™]F1ŒÅ(VY Ò­•7ÒPM/}ÑB 8\œÊT-}ÕÍæDÑLm«5íÕ]s ]³»ce=Ôí϶ÐQ)^‘ÄmSÊ¢óŒ`=×zP×&=Ð4•ÎÃúÔ%,÷Ô$'‰E \¬ÍÕ¹TØ[Í,*PO@ÍEÀ¢MÍÏF}_sÙHðÖEàÚ†È¬È±Ý ³=gSÛœpÛèÐͺ} ¼í^býÛêÕÄͺ9ÝEA;gpml-matlab-v3.0-2010-07-23/doc/f2.gif000644 000765 000024 00000033052 11422240326 016613 0ustar00carlstaff000000 000000 GIF89aõš÷µ !!!"""&&&(((***---...///000111222333555888:::<<<>>>???@@@AAADDDKKKMMMNNNOOOPPPQQQRRRSSSUUUYYY[[[\\\]]]^^^___```aaabbbcccdddfffggnhhhiiijjjkkklllmmmnnnooopppqqqrrrssstttuuuwwwxxxzzz}}}~~~^^€WWº99Öÿÿ þýþþþýý>>ç ó!!û%%û''û ü//ú22ù11ú44ù77ù66ú==øBB÷FF÷HH÷MMöOOöRRõUUõVVõ\\ôïaaóccó``ôjjòkkòppñppòttñvvñ€€€ƒƒƒ„„„………†††‡‡‡ˆˆˆ‰‰‰ŒŒŒ•••–––———˜˜˜™™™šššœœœ’’·   £££¥¥¥¦¦¦§§§¨¨¨©©©ªªª«««®®®°°°±±±³³³´´´¶¶¶¹¹¹ººº»»»¿¿¿··Ó파ìë¡¡ë««é¹¹çÀÀÀÃÃÃÆÆÆÉÉÉÊÊÊËËËÌÌÌÏÏÏÒÒÒ×××ØØØÜÜÜÝÝÝÆÆåÈÈåÔÔãàààéééîîî÷÷÷ûûûýýýÿÿÿ!ù,õšþk H° Áƒ*\È°¡Ã‡#JœH±¢Å‹3jÜȱ£Ç CŠI²¤É“(Sª\ɲ¥Ë—0cÊœI³¦Í›8sêÜɳ§ÏŸ@ƒ J´¨Ñ£H“*]Ê´©Ó§P£JJµªÕ«X³jÝʵ«W¯'PK¶¬Ù³hÓª]˶­[·&¼K·®Ý»t(‹·¯ß¿€ËŠ-+áÔªWˆ+^̸±ãÇ#Kf< «Ç«&Ô|ÁaÒW„>>S¤DIôÄÐ/L^ͺµkÈž qYBÍ€@˜.(dwDJž}?D-‘×È“+oÜÉÈ«Ü6D#¸oâÂN²ž}!ö –þ‹ßz ’3‘mÓla蔆ì4º7¤DH>ÃøAßÏ¿ql$—a&Ýz†¼rB)ÂÅbC .˜Pƒµr\î×ÉH`!™z3±÷ (-8(âˆõI…(Š×†H´Ù*Êä¡«hãipXŠ<¾ÖHöÀtˆˆã‘6–BCL¶KL6$‰yág"YK–Gfib“`J¶…sR‰Xuh©¦|ú…é¦cKœW&;¾¢Ûšxš¶J o¾ÉJ€¶‘†lhêbc†LÙa‰ÑÀežnH}VŠØ%HL!Ù)hFebFŠV£–£Y#§XZé-JF"þlÆh•WvUë·’¨ ªVç&‘E¢[¬Ÿ&VG‘&‹•’uöÚd'²Iv‚Ä.ÊØ{Êf;•':fsšB&DšµT+Ó ³&fãڶ»”Þº‰a@Êåš©X¨^±{£¿$RrB¼`º"›']Pª¹1͸˜îF¬TwÜ$&™BÖv Ãä°bw "ñÈEµªÅ=¢‚AA¿ôqb§„HòÍ@©†ræaòXu0ëûðW¥šŠ¤’;óømˆb'PÌ.Íœ˜"Ûªf®"Ö‘nÒýa¨¡cÓuKRi3ÎhçÄ+×(²z(c¯46Ke#vAÚxÛËþÚlSøkc €ÐÜ+Õý Õy'"uôMáÒ-÷tÍî{¶V“˜ù‚=„ç8^7v ă Í銧¾R›Ÿïçvcw’á*~øæªçîQ,·¾ßß×*D{J¶ƒâ´îÈ{ĸïäA‹c«rèEbzRÁàˆÔ˜T®‹™×%Nç*’ÄŠvÊŠû´GßôŽ¢³Š ήŒ! ¹D@Næ‰QZŒ!ÈUCÊM¦ÈZdMôFG.æ‹pkáÓÁúiRƒ¬óäcæ…žÅòFt êª"JùÔò3TåbɘWRp&èbnyÊÄMbIºt $£HÉJƲ1ù+¦µ–LÇ4§ ¡|ˆG‚G»I•ÕtŒé¨MYµ}T ´NÓä2œ¼t%D¶)’n¾"dVþI§|ôÙ/„“1Otc ÑLgZ‹5 Ðß7ù‡É~Ò9Lk1é{æq¡û3Y'“9NyΓ”Ž9?1:²T:ˆ"¤/ayPÖà“*ÄNLµÂ;‡*cáRÌ#BQ°g£Ž1Iu·<›"Fe,“d&ËùÌǬt¨8óçŒÚ3¸qgŠ éH¡š-“þ“²IibN ‘žJ$>@p€Ô*˜®yiÕÔί¼³šOÌ©bÈÊÓ“¬"kÝÀ€¢YôChdÊV}³X¬4Ò¨¯eðúj’À­‰…e_@Ã^”«xëA!ÛQÅœB¡5 `‹ € puþHA‹7xA6ž„œdYOâƒÞd­ÝÚjäJÛ›ÙÖ¨Ë áU±ÚËzJ žMèTfºêZi•ìbl°X³RD°k@Õ +ÜÕ<µ¸Ê¢&iå4Éxw"á½lAJPÞÉ”‰E©¸¤ßðw¿ùpf\‹þ ä¿þ p‚¼àØÀ[Rð%aOØÂvpƒ!Œà kØæp‡3b øÄ"N±‰Uüa?øÅþ5Ùm½ 8³Æëí b”’U"€gï9^ôFÌ@UercJæ¶y“üéF¥®òÁ2M{gTí*FŠ»} pÉ«œó¥±¦A3þUQ±õ¶r1ucjI"là c®Å_1K#k‹[T$n£¾0“d®­"À¡= a¹~Ž”Wù䡹÷$ˆ2„&Øê¹¾“s¤³¥³.“©1Mv2IJX;o ÓòÈž±ëŠÆÖRÁ®QY‰@Š¼÷!«ðÖ\5¶Èéó¨ó¤^›:1@-e[ŠœB/R-4 +ýå¥N[F uA¯­¦»&ÓË„†´ªîM¥)ZÎN¼­² º9¤ùõFf]H¨¨ù3ÿ† ›gDǼ۠à æMîìH²Ü&$¿Ínm?¦Enø‘&]MtÓLÚ÷X»þ%ce#Éܺ,­$ÇÍTj'ÇÌCÁõWdî”c|‹yt¾³:e›|DÍvhÄÅðè[#³FŒÎþ‘Iô· ÇÏ}êÆÐÐ×<—Ì#ËtŠL4E<‚:D]&(W%¯³Šƒ#œÝä™xRŠ®'ˆ¬B4HUú ³¿$ö¶)Xô˜«_š&ÈVŽÏÏ|5ˆh‡ñ!#Ñ÷£Õ¦8… ÈC.3CX2ÖîzE|@)ÄÌ1”÷;K¾dT¤>Æ›Ÿó¹F~Io‹^"“àrdÍXy¢¤Ý“UuŒ!2Îù¨‚<5»=Eb¡“™Þ1€9NJíÐÁ‹5Œê®cÖ'þ#ýŸÐ¼+¢œÙé3®˜#ó{ w-rôŒ$}¬Ê—"Ο˜ñ+ÉBihëW–°‹´#‰§B gñÇ° b([`(€²/Ý×!ÎçPÁÇvn÷v"·°W€&Ìå¿HÖrW|dC{æÅt'* q\Ÿ$øç5eT¡Ó2\·n"÷y‘±xD¡z4¥¤@Ša‹q/?ñX6õ:¯wƒø¨ § D„aD|8ážgTÀã}}Ú·Ç·Ä%…á?Ö4îáS§Jó3`ˆƒÇ%g†q,q}g{8‚pTƒÁWì·}«ÁƒBñ}þ\±9.¸½—éY2ÔþgNüaˆØ@äQLø“YØ„Qƒ‚«A¤`‡*•èƒñ=6A‘ñ€¿”pü¡[¨(« {ã‘j“]H0>ö²‰dD'ü‘|s§&ªpt0Z•¸‚$á‡]‰¯Çr&H7¤˜‚IÁŠX‘+m˜‘Š+asÎFcQq(‡£H!³…Šª ‹û66¡5:èACF4÷ñ~9wŽLW{P!Ñ{0ÑŽäqä4‰M…;…ŠÀ#8ÐØDk–†.7w—ˆ¸ì2ƒ‘‘€’ÃŇ/QoFõl!‹³w¢’Ëö¥þ#498Wx5¡•y’ŽÃL×XŠÎˆ^Ýè *“)^àq%7 ;žÖ0N® ‰9˜d†±ÐÁ ó²2å æ±’×2. HøzO¹s ‰uHCéEÓx‡KY–‰á _©—‘Ã%á‘ÿô†áˆuoùm‡ÚxW¢ŠŽÁ*KДˆá ,B/±†‹ÃDÿˆ 7—ù’˜¯Q‘玉á<ô¬’•H ‰œ”„­"%xx‹2“!Fˆ‰Œá æF'À MÂ8i GŸ(‚h•´ØÑC?Ö%zØO„Àâ«þP>/±ˆÕ‹´©øhúˆjG±‘˜3Ô§ƒw ­ q2ï 2$‰ºöOiž1žñ„äAšËvŸŠñH°­±qb…éAN UÉœ,Y!ëx{òÈ3™š ŠeÐÕ)9,qœM„’ÐcõihÍI!eÈtgIÅÊ"zµ9Ô C<)K‡üIþ9*¯¹£²žˆá<Õ :x~¹WGhSSé*ˆ¤QžÐS›2ØxH¨*ŒzT²„,щ6'QŠoâIˆ£©|\´K²‘Ñ  (GÉá¥#ኃ)óéþ~Z…”­z?7ir£«a fJ¦Œq,«cT˜ò5“±œz)’•]‡†‹aViƒ§¡…‚NÒžo*šp)=ž*ty§VÈ1?,"©4¥ 1p'EŽ›Rœ¢x‚¸Ég9T€š*3œ^¤¨‰?Äš·“P´¢¤…Á¥GÔu^Dv꧷«Ú  a¢>t¥’1|AB®5Lõ„¨­ñ£—‹Œ.Êq}ÎC™‡Ã­±ªÕÔ…®˜èx°ÆVwà«d–"Ï)—jRm¦¢Ú,2£8š¬ò×8å<ƒêQøª!! »P€ áe²áþ­)ÙƒjrQùMš³«H%!Z6•Ÿ„ÊÛ„/Ðj—°ù5llõjkÅY²F¯çÄt¹B~ƒmú ¶HÞ™Læñf™ ´ D­ò ®5Fœžæ²pð;T¥‹sÔÖ˜:Y#ªá4‚Èêµ €°ñÅVX¨Vm¹h0CJ†F‹^x¨ÿÑ «ñÔÑ-])¬ØwhœYMžÉ}|~Kñµwp°Ö–µ°VÝ“g?Õ#BuˆZ"cCx¦Á¢ @z“ }@e,‚M€ã“£ Ç >­Jr²:ÏÕ«p ^$p²h‡‹¼®…¸Nû´þ—K[,ð8> „*¨ŽA$¡®D¨W›Ç»H»VÍ‹5€ë›5%0†=BC7Öbù[búËbüû¿0bìb6ÀûkÀù•Œì_n‹zËv‘@ Â_cg'qE³À<§¸bµ°£Tq“pÀ–cBÐ<¦îë¼ AL«š•hMÛ$Ši:‰›»©—¥µõ—~Q!¦ÕÄ®¹i°k0|± €5¢¶Ìd¤H"Š1/VøÛ‹}སšœŒ±¢Ô ¼Ŧ®a¡¥q„u<‰»ÆI&5xþÊgéSÅþUÀñ‘"ü?×ù±}Úƪ ¸à¹bÅ裱‹´ªú}P¼±§É€T k(z¹ â ‡ ?>«~Œl±0l$À>’<@wÌ?Àš\ܧüäûIjÜu ÀÁ›á´–«Q¨,:ó««Ë$>ÀÄß´¬zì”êxüItë§õZN„Éš|Ì!ÑZ°Gs¼EÊ´Z&Õ£8Ú A ï$¢–KhqÊà#µ­‘­ûiˆ¸+köË$Æ°¾E±Œ£ßL³  ©câžÊuB̤²äw€°ËÉôÀ­1ÒMüc]`@žÕ ÖFV 5 (:Ð A1ŽñÕèƒ&H•ºÔÓ±˜a]qar²àgAwÒQ4­!Äúú³aÏÌó ­Áز7¡MòÒFæFb*ȽJ­åuª ûÁV¹ºͯ‘͙ͩ`M>‘ÎRñGÄü 9ŠÌ™´]„[}sþN—¬˜ú¼Ïag ¿‘Öí…^÷ÙÒ9·¸¡Ó‰q³Ëd8›¯aK]¬¡Ý1lîW؆M[;|©Ðcx±'pBµÒLÑÀ&Æž$Þïý„C ÝàÑ›øMÔÆ;°‹çL׎1¦8Y+•¿˜« Û‹ýÜàñÞ½žÊQýT§Œ¡³¡¦‹àaDâ!ÄžäÓ¯± þô-M½«PÇ;ÑÖ*HàI3)m±À-ÑÐMAh­8«Bî$´¢--žª\µÃW+n¨òF˜â¨´2ËšL¬‚±¯q˜ØÓPä¼eZµYU÷æ qÕŠ äCdw¦yสÐiÔœNÖ¨þ&þQ˪ڕþBÌáÒlëÑî8×¼†¿Ùî&E-M“>ϱ3€Æv 6áãŽä쯑惘Ü}Gλ1Î=qzþøÚ ²_¼­Jè-wÀîXï&JŽLžKT'è†þÙ#qéŠ!‘$:0gÞ)÷ÜÉ¡è‹n5ìÎb~î¯åº3Ý ·‹1)%ñáŠ-Uû $Š¤펤ãÉ!ÏèNÞ 1æk;Tó^p{òµððXâØüÅgx”­Jâ.2ïëPÑñ&/¾P O눱 U4öù>È^Jõ–DE¿« …<ó-ê&¿êÂìè&Ù&ÁïÞ±1Æ=KÄê OçLâ× ¤…[ ·­iã$·)/÷‘“qwGUŽOí!AE±pcmgw zzF O2.ølÛnœcÏ÷‘¡r²žL‰L¥,ø Ájˆ8ûŠ¡øßþ÷+ÿ æ dîõ-s6Ð=`1ùO™Ny°ú €%ˆ  ±€Ž|vËÜ'ß^µžô‡nëB )qWwY’ÛäÁÆQKà@‚ $HáA@$ÀÇBŠ[z•QãFŽ=~©¡âH’%Æ2™RåJ–-kÕÁ¨Ñ ’KEºtÙÔÇ)HÚ€TèP¢E‹¢AâÅhP=8+*,y€¬ƒ± p*ðâR®@i”ÊVìX²e ‚Xµ ’NOµ0K‘˜,qÕUï^¾EY­mÛW£¸¡’°aáa_ÄÌu•¡:…-_ÆlÖÃÆNk9Öþé“¡KhFFº+R¥©+><2ñâÆY·ª^úPônÞ½–¢±±MRŽ'PúØc§ÇáSp?‡Þñ/ÛÔ†"YŽ]Q€ã­J{;ºPÝÈI‚5Ÿ^%ÌHnÄ žTpÓ7y <^?WÖ©'OÂÎ$@xá Ä1ÈöûˆRÔƒ0¬@àh¦ü^yd"õHû¨¿ŽÞc0Ä ¦»P0º3+»ŠÛ”bA„1ÄOÄ&{MBw$‰ÂLãè+}ˆ$­Ž:C‚•mdÒ#S£á2c +¯Ä2Ë–jl2£òx3L€ãˆ¸<ˆp•@š %»ì’DÜNþÒ¤,ïÄsËáüÌUÄÜ‘=<ñÉ-¸"äÐ#»ÈK88»ÃLÕ L.»¤@1QT=5êìÂ>B‹4z1£TeÉ#£³R/er•H4ì4Wõ6‰â85/–5?.‰5Ú@¶-’EVIV¡s¥§1 ë¡Î’fCˆ±ð 1’Õ>uW42…k-£UÐÔ„S>rÅ´TW}6ºá–pVµSª•U6ÅÚnÛx_9áÏpÎÌÔ¼HõÖ©¤y}ÞU›NbÁ !¶Š&Á³ãüwT1Ñ#˜7u7‹ …ÈXÂO=BU#‹'Žl&,¢3DS)¢¤ã;þH0«Ç&þ¶d£ÃâUfFÏ´t9E×rvfšùÚ°è2µ±}/£µK€=:l–J¡k£4nÞ¨hõæª'7#ž·h‘xº(³i\-ëºÉ:;p—ˆä¨ŒT Ù[ǤŸNòLâV-R¼Æ»³)!ú_PåQ`ÁãÒ ÛŒ¾¸O£¯<×ч=9rrÕ†C¢ôè8ß1Ÿ¯ùß›xûó¬„íh ˆÏ­¥÷ôÔjQO\ I©# Är;oß`kœúnÒAß·¯¨:0a¶'Â/ГÙDmæ÷bÅ´r£[åéq*¥!ë¯_ýÑHðå¾ÿ‚ZXG°Êjôù¡þ µ¾®¸ H˜Ä¢ ªàÍ%Jœž¦¦6ÿùÁÂh:’…éA™W½x”µ,o?†¨`¾pr€ äî_ãQ ;Ø’qeDk`Z-t¨#ƒJR+ Rð8]g~YÑ–mò÷(ïíЄÓÈ%Ò—*J(‹ÃÏZd‡D (‘‰ 2Df.·’ÔÆ)Øëí¬ø¹ÐµÇ' *`®t¤LŽŒìzàm¸'º$ª¡ãVÔÍQlÀÛHO0‘!媇ì2Í´þØ®åŒ!Z„§w;B¾qŠBÜŽéÈ’ ¢a™Ž w·Ê$U£Û&9cA2)”…¬Èíþp•Á¹ˆèt·#ã±r!DÝRžfB¨w@¡{¹IØ-aŒ""„(‘ ©Ž]B)i´—½)u[…ÈÂUG‡¹SŒdÅL’"AÅOœ½™D•ÌyÊG½eGŽè$¹ÐRL×Ê{f²®J\`—"ŠŠæ@X(äÎU̓‚©‘ØH,ÓV²D1'#¤(ó<Ñ)¼ÊFiìhfá!4ä˜9g—ª¡!–ô -uœ45bЕGux¥K3Ò>ONmèÉZ°ðI8e0.kDÈ*|P%H@¥ˆ•BêºÊ\óáÈ;7"ÏÅBõÔH'sÉþ*Vl¢ \‹lÚ¥ ³"€hãÏ {­Ÿ ¤e•ƒH Ï=H ”mê@0;&Ízn²žílf/ ÚÑŠ¶x¤=­i7+ÄÐr¶´®M-lWûÙ¦ž”uÛ| þD[Ù¢v¶­eíkƒÛZX–¸·³–ŬN( *Ô+:>/du-é;tŸ•Þ÷·Â,(xꃀ§$¥°Ò tF‰×v‚ì£6çV]ÑsªP bɾH,œáºhÐn_<Ñ dyAŸ×½îÆ]×åA4`-HDˆi;`FÓ[«^<á×1 ë ¨(_¾!´ºõ’IuÎRåXâõ ë„u¬aþ£Á ¼D0‚½ÐáMxjáÌ)Eð1„ )Q«Eâ‹Ÿç8°\å h DÒ¬°XB¶µcÝ^A׈ë‹Ê„H0aÝ) aü„åtÛ0†;9Á^@C&‰æRÊ° ÁÖA¢ÈØ eÌžÂÚ``À.¡ t&Êîv‹:hà| 3Ù7#ü¤nÅæƒ>¼)yÖÍquÝœã@# ĺüÈ)(-C!ù FFrckÔŒlBÆßxéM½Ü¡eÜ+NЇnÙŠ  ònN슮4pH ³PX ]úº;nÇ}ŒjŽLƉwFHDrD _˜ÏJŠ+"u—þŦ¢ƒ¢èí!B숭. NaY#U[‚¡‡Nµ+îD¾mn Ūx°IÐ ¬ díuŸ?² ^ŽAÞQ÷Zjå_óÆ©ƒ+ÁPŽìËXÚžšÌ—a.8­á»[)‚X<M%Iþd•Èë®`‚—P ½ ‚¨ %`y\NüŠ…šÎʺ*EÀnN³œƒ{!bÝŽ•Ìzõ‚™§Ô«ÇùŠ@"(¾þ\,4€Ò¢Ø êêñ÷}§t}ŽÖ[×Ï*ˆ°#'™õi{dôêöT›7p  ûRú ÇÞ\@™þMqF˜w0Nvözzhå Óþ]ñlƒJž²W¤ØÔK(/ˆr$R©’¹@Z¿#4Ýó]â_èÍs)Òß%ħ7~QNÑÖÞD!>€ý^î0{ÌH̦cC«6hÿÞFÁ· „2§£§ïã{)Ã?3îõ¡à{¡Áê-sÀ#­E‚}ß^Êæ}&B„“Y,8?¾h—ô1Š p>™„a•U» [Rš÷Ù<î¹7þ³o@KŒ,@[ - ŠT˜U¸|[ íZ¾Ýs äéˆÂ=ù úÃ@é¬q˜ðÀñ/Ü Š89³Š˜»½P¸;Ì eb6$1M;ˆ"Ì•š«þÁ¡l‰Âó­P ´‘>:>….Ú‘ 6$üˆH0Ãßé¼W8ŽÀ¿þiÀ+dÐ>Ã3„âCé¤" à·0q¿¾˜™;𪱠ƒNëˆ2@˜{… ôŸ=²Ãñ˜Ãð‹ÓèC&¡ªçY*¥ãRxÃ¥šıÀ:{ܪrÁJ¤ÞàÂÞˆ)¸ õL½éƒD, ”;Ô ©1A*d‰?ÀƒúÀ/ú:¿³9YL æÓ99‚äc•™Ä¡°¼Ìp>È3¡ 0å Ÿ9…% Zˆ… ø-z9¤-Ì¡÷“ÆÈ?L„ˆ‚1l’O$Š2þ “:Ø¿¾€§kC•Pˆ”2¸‘ƒ—«G¾ØZ À%c^ ÄTLR Ç®ÅW¨ƒ‹¬+ $xb¿òžˆD Œ¡H¤müÀWpy,‹X8ÁÔøÈW‚”l +d ƒµ0œd1KX§îcI®0›\‰Zä¿ÛtœÆ°@ÆZ€¼×)ŠA“èƒ ¤è‚ 1C÷ œ„\J¾À© œn¡™\žçQ´Dèà 60ŠX¬•ˆ…2"6ÈUpÈÏ‰Åµì »ÄGõJy ¥Ås´&z M0ŠsQÌ’ØAP NÈ‘xÞüÿRŸŒÍƒÚ¯ŠÉ˜…Þ€ÊÝ@Ω±¢XAÜP™ŠKà%©a…u‹)®…h§êƒH€Á qþNGjŽ„ѹò Ô b¼Ž@G¾ñ=UK‘ãÇ#)½3Š´HR@J˜„U`4 Svi¶ Æ’F¼RèËjdÌôl’+€ ™Œ‹!ÂK©ú´MOx #íŠS°ÔøQ¡°‹ÌÓˆƒÌÌGrÀ>õó<ÏÓT&Yw)ŠÉ³ŒB䯌8£EÁtK<¾°R ˆlQ£jQ £ÑŠ\dÉÅ¢èɸ¨ƒP}PdY.eiƒi A½Ð«^üÒ½¸§¨$ú_Õ,VæQ’ÖéÊÜd xROû Y5ÀHUÓ4 »8x¤¯Ë Un}ÁóV¼ÕxþQ廀¬¤y=¡‘6P Y効éDŽp×¢ÈÕÛtÈð—*µ×ø¶R•.e’™x¡8 XøX‚Õ6àDÔàÃWM 9™«¯$ “=­ØŒx„‹8˜\!$Á5 X¾5äSÒ'¸‹VH mB؆­O¥ÂV‰åžXH¹™5„SÈ×ã þ/}”bØ­-¸€ˆ¿ôôáˆÙb®׎xÝb&FLX®ÝJæ64f$‰[  < ô˜DðÁ]bCþ¶½A.Šêýæg‚ˆdò¬`eFŽÕ¥ äñ8ÀR–˜¼0ðŠ˜{Šl~˜ŠcNêf0e AQ;ŽåŒe批d&‘:“Hð€ˆZ …aº~•\Ø ˜(Dé”HâN™`&nËr–å­»a]Ö \ÚÙh«ãM‰ \‹ö!Nü@3‰lýçg–¬ƒGfµL,éç0„³2Š,di3„îˆGS°¬×Œ€j·^ Iímöˆ<½êYsåµÔž2žÞg9…d½€:„ Èëv´šýI¢¨ds DˆNQæ¥`vºîI¼k@@ÑÑ$g¡&éxñ€>èŽþUЀ ŠÄÞÅÊéË RxlZ Aèê„uCkœáßX'Ù%O»ÆX¯aè8Vƒð­¾…*& ‹:4Å33¬ú/¡8ºëèµh‚U6\@îŽøÁNX5ÞÚ èiW5ŠM,ƒÖí»VŽýbóÀ\ÌÐܾX¾fr1Ã瘨ŒÖˆ-‹øÝÁ„1X‚)`°Èé.DÈ쥄`ú~â ´®ÆÉ{9Ô5ê2ŒŒ‹­‘½îhmë¥$A¾ÞUx6 €h¤ûÞ‹î% áŠâëX¾XÉŠÖŒ‹a‘÷.‰žÖ•¬–Å;àꮶþŒRh£Ø€(!ïŠUp¼‘Øo¢À¥BÖ䡈Ÿ ¸ Žl¢ë†ï–FXe±¨o— ËcrµYÓ»!¿@¢€Çyë¡HY¢8bËà)‘BðseÉÍælËzcŒÂ2hal“°q¡0£= >œÉßÆ  Q$ôÍTÞíÂ8¾‘”óÞ [!s„Í$Z Á䊌ÑSx_œ8U—¼ì”8£ªîz”ÜÅ4‹íÐ!R(«xtýÀ!•0Á¥P¢åÜg­²MHó–ÈñðgBnÑ]KÂT5g ³ ;B`Éd µd!ŸøÀ•ö¢HñÝ8Âçþ°umGˆI0tiœð݈8…ÓD qrŸs£H'âê®É*H[N.[$xÁL=[[ µkO 8ŽÕ2.®„-œ„Ž;ȯŽ7ù’Gy„?y•Oùï".RDÌÉà—§yßò­S „œï€Ï ¯ £òÄ ­hN¾0âôÁ4ó}2´¥°FÖ|ß 51P>´@Ì: „oç Ÿºd‡ŽWl ÐðEîWÀ('@k³Š7Z£ÈôàôȘ¯ßq+,ZJ«iËØŽ®'ôr¼ ÷ó÷< ÞáNh2RÃäŽèÆãñ]çO©ŸnJðÄõQ‹R/ˆþSª‚Š»w 0sŽÈfWà„A[„93½ˆwó¸vÔÈ|{ˆñ´ÃÙÇn±Xô‚hôÏ—ü®8ø*;q´ òŸ,ýã¾JÚ×ú¥¼õ²w–hó€s‚{Üò• 5Kï NÓ‘R0ò¡øs§¨ö´Œ¼ˆ va'Â&ß©rd2w}7$ Wèð @÷Ô‰p£@nÚ—hªˆW,hð „ &Œä£Ĉ'R¬hñb- 7r´¸êE^¬šØÂÔ ï êè²–>*gÒ|EcÒËœ:%jU³à…’;‡-j4§‡ŸJ—2-¸ÊPŸ£G5J­ñdÓ™ ¬FŒþ¥!kSP-¸Z]•t©¡±dײe«aظr ÁÙÖ%Õ»C_ œ‹ðÔÖµ}„ø¥éA¨Þ¡u|, JqâÈ:}ø,l9+ È’)æÝÜëe‚w¢²õp*t§=ëQ™fÚµ±XÓƈèêÜ4זعwE¾º_˜Ívá5¾ÑìOÇΧË>¡ü:ÂÔ3n§xnnÞlADÂþJéîû” )V=|£ ÌÓ(ø„øA‡^Õ§Ž^±\€`úYän)" rÔÂkõ)§Öt¿uÇ_h'h&XrÊÙ`—ƒµPGJÒV)!&ÖGàI¨›!=lgávÂ…'™kþÃrŸŠµ´`B'Ò#‘­bËÊA5£Zö€‘A§ƒEÖB^B4¤g¥•g%™ÛMLêçda4œ(™$¢ÖÇj\Ö 'D(ÐmºYd §}‰š Í9G#u6†ækèiÙ)Ü)%Èé Ș؆‹U‡L{†vAw€NGæ\ïyVŠ—…y¢¥QÒu ™Ÿ©UJ‡šZ6)už:ª\u ÉZµÊ埲»h©¶ÎÅ©˜ññWqµµ©\‰{­¥ó)ë×*¸VØd_—ñ(™€s6e"¶érYG¸Û‚eH±Àé ܳYïf«|×bêÖöª¿.!ç®þ\}´Ô,|ß]VÇÁ½!Ò^V‘Œ0Åð%KðRa"¬^½çÆJ{a]ðqÅ%o§-ÆMi×)¸—é»]Ö)µ`¥&Û \¤|®zóöÖñRu³R&Ö|3WG#] "„í¼…¹¶lY]ê±çbJB¾¹t€\[‹ÌOÿ4qm>KÛî\„ÂW‡æÊ {=7k(‹=“Ð,ë'¨_d· ÂTJÓ}øZ=¸}wB†üºñ…hÇoƒ€\@-A x€/Ⳍ3ô%=Oí Ç áhIƒ’‚‘êžu¦çGŠ®RŸ¦)9X~ÃWJ³Û~¼^‡íγz; þ|VÂ#?=×y.¿PÔRÿέ!QRÿ=â˜^¯Póo§ð\©§k8ø±4­ã´Ê½ð™MÐ4½œ.ÉícÔ?Ýì&¿tÏ~§‹ àü§@¯ p€ç1Þ·¶7—¼-°‚&óô®—JP.,± +6 !`m€jóÝóü­tÕ.„i!â.&¿•¡0a,š s¨®ÊOQŒß²rŠìé°ˆ‹¢Œr o™:èKgÄ)ºé6I|EŒâs?Ö±)fò×û*F¤íŠu€`Eˆ¹HïNtáé&Ãå©ï‡6œ‹©¨G.YÏ4¬!ÇnH î±*bQOÈîtþq)>@£!#©žø9°¬Ù¢gòW"ŠáŽŠDÞEGDí¥0.Š”$*áÃCÑQN‹à*¡Rê—ÊYªƒ4˜~0¹™FÖÄ?R¤%0{C ÆŒo³#ÇÖŒ‘KÍôß3‘J±ÙÀ@º”Œ&rJþ2Ž‡Û×ø,yÉ2EœÁ<çZNkˆ—°á:ãy+^oˆìì S„P:yòSJHº^+]‰Ï¥¬0`[2â*á¿U>­ûh)™’ÀŠÁ3„”xè÷ú¸» ê'?ý‰=)ºÇI@Òvˆ\^ù¹¬°bBØãEýG ÁÝm”Mü” RGŠÑ`¤„Pà41fNÏ\3þ2¶lŠ  Ô¥2µ©N}*T£*Õ©RU©¨*V³ªÕ­rµ«M½ªWÃ*Ö±fU)¥„"²²µ©Dè'\ã*×¹Òµ®v½+^óª×½òµ¯~ý+`+ØÁ¶°†=,b«XÛÝSüHHFMä Á±‹]à,«ÃR„äÈl`¸ÀXdö²þ+  Ùfö¥©ìt8‰ ¨Ö´ß„js[DÜæ"À/ø‚ÜÖÖ¶Ó¦|PÜ÷—¸}Aáƒä.׸ԫn%Õ•("v­{¼ðZ¼ä­àyÅ‹¸ô*м« !{Õ;·ø¶Ï½à}¯|±µýò·¿þû­}Ág_8¿n".‚uK‘SO»ª›„wïkàñâׂÍÈs%Š‘ÇYx"ícübléøx I¦Øã¹ÈF>2’“¬ä%3¹ÉN~2”£,å)S™,C®2–;we£l9Ë^>q‰¿,æéuyÌffËri[‹I¤Ö»rˆA[ ÁOá-.‚AÜÀOÊ%Á™}”4Â5Ví"çáÂyΡU4~‹»_Õn ¢Qn›ºÛgAsz(„ID†ÛâZˆ/0N,R»iRGzµ(–Ep&ÚÊ­Ó¶þæ¡õ ‘X,—ÒÑn¢Y½`W‡ù"ˆ@~2|ëeãZ³´ 6¤!½Š[Ïw&öPR+„@ÌþöEm¼^àc¼¶ìµ%¢ãX„¤Ðà~÷°#B[Œ "çÅs«Õ]lŒ(Þ„7§ Mor›ßñŽöNº«ZK<àÎv·¾!]k¦ÁÙÞïUxÄw¢ÝÊn·áð÷DÈ­ky تõsuc¬\+÷»·ÀC®ÙÜžzת6Nf¡ ܦ|µÃ¬¨¦…20üåbyÆYÝæIß8"Ég½«]‰„D«Qn±M;ÕNÖèQŽyÒk í>ˆ"†NçX¨9"€:‰ž9~‘•_ºâ^¿{Ò³o¼ó#e¶%öÞ÷ÁË\/B4áßì»AŸT<äóùÉK†¤”ïH@;gpml-matlab-v3.0-2010-07-23/doc/demoClassification.m000644 000765 000024 00000006036 11422255416 021603 0ustar00carlstaff000000 000000 disp(' '); disp('clear all, close all'); clear all, close all disp(' ') disp('n1 = 80; n2 = 40; % number of data points from each class'); n1 = 80; n2 = 40; disp('S1 = eye(2); S2 = [1 0.95; 0.95 1]; % the two covariance matrices'); S1 = eye(2); S2 = [1 0.95; 0.95 1]; disp('m1 = [0.75; 0]; m2 = [-0.75; 0]; % the two means'); m1 = [0.75; 0]; m2 = [-0.75; 0]; disp(' ') disp('x1 = bsxfun(@plus, chol(S1)''*gpml_randn(0.2, 2, n1), m1);') x1 = bsxfun(@plus, chol(S1)'*gpml_randn(0.2, 2, n1), m1); disp('x2 = bsxfun(@plus, chol(S2)''*gpml_randn(0.3, 2, n2), m2);') x2 = bsxfun(@plus, chol(S2)'*gpml_randn(0.3, 2, n2), m2); disp(' '); disp('x = [x1 x2]''; y = [-ones(1,n1) ones(1,n2)]'';'); x = [x1 x2]'; y = [-ones(1,n1) ones(1,n2)]'; disp('plot(x1(1,:), x1(2,:), ''b+''); hold on;'); plot(x1(1,:), x1(2,:), 'b+', 'MarkerSize', 12); hold on; disp('plot(x2(1,:), x2(2,:), ''r+'');'); plot(x2(1,:), x2(2,:), 'r+', 'MarkerSize', 12); disp(' '); disp('[t1 t2] = meshgrid(-4:0.1:4,-4:0.1:4);'); [t1 t2] = meshgrid(-4:0.1:4,-4:0.1:4); disp('t = [t1(:) t2(:)]; n = length(t); % these are the test inputs'); t = [t1(:) t2(:)]; n = length(t); % these are the test inputs disp('tmm = bsxfun(@minus, t, m1'');'); tmm = bsxfun(@minus, t, m1'); disp('p1 = n1*exp(-sum(tmm*inv(S1).*tmm/2,2))/sqrt(det(S1));'); p1 = n1*exp(-sum(tmm*inv(S1).*tmm/2,2))/sqrt(det(S1)); disp('tmm = bsxfun(@minus, t, m2'');'); tmm = bsxfun(@minus, t, m2'); disp('p2 = n2*exp(-sum(tmm*inv(S2).*tmm/2,2))/sqrt(det(S2));'); p2 = n2*exp(-sum(tmm*inv(S2).*tmm/2,2))/sqrt(det(S2)); set(gca, 'FontSize', 24); disp('contour(t1, t2, reshape(p2./(p1+p2), size(t1)), [0.1:0.1:0.9]);'); contour(t1, t2, reshape(p2./(p1+p2), size(t1)), [0.1:0.1:0.9]); [c h] = contour(t1, t2, reshape(p2./(p1+p2), size(t1)), [0.5 0.5]); set(h, 'LineWidth', 2) colorbar grid axis([-4 4 -4 4]) %print -depsc f5.eps disp(' '); disp('Hit any key to continue...'); pause disp(' '); disp('meanfunc = @meanConst; hyp.mean = 0;'); meanfunc = @meanConst; hyp.mean = 0; disp('covfunc = @covSEard; hyp.cov = log([1 1 1]);'); covfunc = @covSEard; hyp.cov = log([1 1 1]); disp('likfunc = @likErf;'); likfunc = @likErf; disp(' '); disp('hyp = minimize(hyp, @gp, -40, @infEP, meanfunc, covfunc, likfunc, x, y);'); hyp = minimize(hyp, @gp, -40, @infEP, meanfunc, covfunc, likfunc, x, y); disp('[a b c d lp] = gp(hyp, @infEP, meanfunc, covfunc, likfunc, x, y, t, ones(n, 1));'); [a b c d lp] = gp(hyp, @infEP, meanfunc, covfunc, likfunc, x, y, t, ones(n, 1)); disp(' '); figure set(gca, 'FontSize', 24); disp('plot(x1(1,:), x1(2,:), ''b+''); hold on;'); plot(x1(1,:), x1(2,:), 'b+', 'MarkerSize', 12); hold on; disp('plot(x2(1,:), x2(2,:), ''r+'')'); plot(x2(1,:), x2(2,:), 'r+', 'MarkerSize', 12) disp('contour(t1, t2, reshape(exp(lp), size(t1)), [0.1:0.1:0.9]);'); contour(t1, t2, reshape(exp(lp), size(t1)), [0.1:0.1:0.9]); [c h] = contour(t1, t2, reshape(exp(lp), size(t1)), [0.5 0.5]); set(h, 'LineWidth', 2) colorbar grid axis([-4 4 -4 4]) %print -depsc f6.eps disp(' '); gpml-matlab-v3.0-2010-07-23/doc/f3.gif000644 000765 000024 00000031525 11422240326 016617 0ustar00carlstaff000000 000000 GIF89aõš÷  """***---...///000111333555888:::;;;<<<>>>???@@@AAADDDHHHIIIJJJMMMNNNOOOPPPQQQRRRSSSUUU[[[\\\]]]^^^```aaabbbcccdddfffggnhhhiiijjjkkklllmmmnnnpppssstttuuuvvvwwwxxxzzz~~~tt†ee—ll“ÿÿ þ þ ÿýþþþýý--à!!û%%û ü""ü((û//ú22ù11ú44ù77ù==ø>>øccÕBB÷FF÷II÷KK÷RRõUUõ\\ôjjãïaaóccó``ôjjòkkòppñppòzzñƒƒƒ………†††‡‡‡ˆˆˆ‰‰‰ŠŠŠ“““–––———˜˜˜™™™žžžŸŸŸ£££¥¥¥  «¨¨¨©©©ªªª±±±³³³´´´ººº»»»¿¿¿ïŽŽíŒŒî——ìë¡¡ë««é¹¹çÀÀÀÁÁÁÃÃÃËËËÌÌÌÏÏÏÒÒÒÜÜÜÝÝÝßßßÆÆåÔÔãàààîîîûûûÿÿÿ!ù,õšþA H° Áƒ*\È°¡Ã‡#JœH±¢Å‹3jÜȱ£Ç CŠI²¤É“(Sª\ɲ¥Ë—0cÊœI³¦Í›8sêÜɳ§ÏŸ@ƒ J´¨Ñ£H“*]Ê´©Ó§P£JJµªÕ«X³jÝʵ«W¯/HK¶¬Ù³hÓª]˶­[·ÞÊK·®]º D¸Ë·¯ß¿fÅšÁO§Ãˆ+^̸±ãÇ#7¾âec5_¼øš°gŠ‡}žè¹à¡’S«^Ͳ™"E,/ÆLSóhƒ¥o? ­bî.µN¼8cNJŠ°iÜ óæÞCWÈ{:Céw~ßÎ5š"Q8þ1wn=†u†‡úœ_ø\ § ÝãËoüv#ÇÍk·ïíiýÂþþ!`twÌg`|vE^<–ßL¶(á„yÂn·ÜkE@â y†(¢A9ø‘I†(¶ÆrÊAF„û(£„™\˜âª±!FJˆçX&Ê!t•ÌJ‘2"ÉÂ$86ÙöEdLCiåt™œàä–±(†d/ ã•d~¶$—\nÂÆšlhQ„h°¹c™„Iå˜eæ¹U&&œˆæŸõaFjSÂT¥n™™¨Œ™ô`ØŸ.H…jvŠçgŽ˜iˆ5Bú'°¡‘ êiªV=ôá)—À¶¦jþ…¾tè©´Je¡Ÿ«6¹àÌ f©µ säê$¨Jx¨Hjœ8¬nH&)#|Æâè ¬ÅêÒ¬Âv‹ÔpT{#'Q08\¥²^êíºB‘ œ¸("+kèn«.»øò”%¼(*Û²Ã9ð‰~Ö-ʨˆ9Â/†$'*qÎ<ݦ!Rì ÈÉ®>¶FªÄù†ì Úe<ߎÉWoKÜŠìrM,¼krw qŸÊϾ¬³Lžx03‚®ng‚¶,ßëU´#"íþl' ›kÜ wHóÎX·„±ÓO—{EÇÃý(\õÙ'U‚×ÄmÌ£‡ÅÍÁ‚@VƒŒöÝ'Éöþpß‘`qpÌM·Ùœ<¢áë]°÷p5<Ü%1üV7ŒÖY<¡å½ Ròâª14q„\ HA“‹‰÷é%é͹dþñðps@qéw¢n{Hàºúc­¿ÞÚ 5 D»¥·Ï‘›ïÞX&½ë¾Õ Ÿ®uJ‡X=tª+ϘçEðJôÑnüøå®}c­gö¨—ˆ¾ôö’/ÿDšŸÏ‡_¬?*”0Ñó àC²g¿Nô úK¤#<ñyqœòâ ˆ˜ç˜#ÈÿXÒ2ÎdÐ?üŠ z@ÁNpbG®sÞjzp‡lp%  AA@å¹­Ž{\òB•Äp†þ¤–ý ± #äp81 DòÕoˆÉI™q&!¸%Ú ZFºÞhrÀ¤óµNŠÅ©ÒÔü°‰ãË€ ÷Ɖ¾};…`àû˜ˆÆÛebm»ÛÄ‚\—ÀÖÄ@‹á»bç‡G­îÉÁ¡|v‘2¢äŒ\ã„$É(C`äŽÆÖH:n%„ç¥V,´º6Âæò9%I¹A PÅZœCÕÇâ0²“‚t¥ñj8³FDñˆÝ¹C %âÈ“@R—h³Û@e„(ÀM>™¨‘(IçI­2@×ä %b°F~Ͳc¨¥q.AŠÓ$ÇD&Öà 0§ù6‰§q䀊”-—êþ<gV3g¦(ž1«™O—yBK3»¡ä¹I¸ "çĈ'z€/ØT:³²JÿtT+wˆeÆù¹ý`tæz[…©«Ê„‹¡Iä#í[%ëV®äõ6†Š% U×åê¤N+4p ƒBÁû-“Úšš¢o&"W5ÊH†%rˆ­nJ|Mp‡cÜ­þþ)jJ¸ÙªÈ©ã“T´:Kõ d"cmÄh"i&=*§Ø$•`éŠ5PQÐVy:‚µŽþ{¿¢Ì-µî ÏÌ^›¬Ò@À¢(Ïle—mMo¦0ŽlU|öÙ[FvRëÄ è‚“_nå¡C†å&5_ÔÈŽAc¨8KŸ±,Žz·°L”SÓ¥©‘BÌðZúM\7²étzÒ¡D.À‚ \`ÍEItŠH±9D6Ò¦ÃuM(q™Qõ¬Eðˆ1°Ú µöÈ­•­‘÷t11&#PÚœ¡¨]Æ«ª*ª ý•4OØ !!è$·¢¨íFænô–6KëTs;$=)càpá¡û@›H*MxÕßÿöˆ ú™Ü %§(ÊÂÛœvê~³;â$ydþ&Ññq£¨FèÓ,Ü‘lwdÛMù4t:ª‡pI&~JX3Ôºòò«Ǧ³¤ù³ê…ì31åÅ/PôдµÊÄNËqË!òŒü™1¾JŒvŒˆE ÿYÉ×=ôªçíÛŠÉ:¸ŽQϪJ˜×ÏÀå—sæfo)£vÄPÕ'Ňzê4UÖêTq7‘Â4Çô1Iì‰ l>Ÿ§?>cþˆÝ7‚÷¼ý1—ïÉ{²´wÇmÞc[! Í‘Ík¤óžGˆ«£&9iAN¯BŒáwBnšm|oÜÔ<âc^"&ô¢¯"Nö.ŸF¼ [{#4Ù“=sE„ù|‡ÌîqþṙEcà\! ~øwåÚ_±XãA?(ß& ›²ô½ªL ßüÄ¿H¥³kçdÿÛñt>73Ò7}µ“!è†|‡±u9±5ñU Â9ð®—°‡€ ðD!wËã6qoñÑ:qÆ5™ðl'tÔ׊ǂ$bzqSO6qpÅÁ" ˜1h€Ä£ávÛ‘ƒ.!mÜ1«3k8|‡*xÅA„ÊÓ„Tù¸„HÇ#¬‘E b+!r`å‚‹&“ŽVˆ’Ÿa·Ñ?LÇ ˜ˆa”Ù &™;0oÄá/=¢=LƒU9Wq^—™ÙÉ1Š!Mb©˜G)†)qš‘šÿ rn‘¡¸×°Ak"wý˜ š°ñlCa¹—j¹–ž`ŒQ.fÉZU™þ…ž˜‡<«ñTP§<' Ž&›¡›B¡’)¹I‹(Ì ZP‡ß¨—&ßÚÓ™çÉ—M!”ŸÔ0øÙ„Ɉúƒ–$qt«ñ°¹;öÙŸÑ)‘½Ç";)¸÷lðoÂŒ° h)F©d«“y*ž¡ž¸Ö{¢Y„Ù :É”6xÄ B’ÊSš)꟱÷‹±#ØC á˜᎑a¡ç¢¦é£K!™£ÁåhBo²‘9Ò ™ðtõ™‚‘•¿ÉŒÚ³@,¡¢Á¢‡Ö–à/“r—ˆQ.З c¹‹y>  ¸ÔUhk¾¸–½'¤Æá+þt¥£Šá¤ !‚¬AŸÚ3¬w›±U? ¥¦V¦¨ˆ‘&Úób”XÊ£"!Lá烤GèevQsÀžìñ§XÁª_Q™ ûȂʦj«—°±™lÓ‘.qN”ð©ªzPSz e¬a©€7#» *Ú…'€s0˜ž£a!¦™z«s D#¡!/ fDÕ§á wvQdfQH媿€ §ŒQ­ýqwÀ7º3º,~i‹Š”—ìz›†®b=H=°U{&eà“ÎcÉÉš*G°)TêFÊêö²Šbº;ü¹›æ ÉúþYÕ§å“dÀgkI/ ˆ[I­kÞv\ÊÀ Æ÷—©=ª²Á²%ðY0sPf2+A.v±€ЬÐ팀j”åÂëCÏ.‹±­ŽqœÑÂÁÌ?£Ó=ÑÓn|ÒQiò¼M=£†uщÑÒŸáÜ9ýk>AÕ ±Ð4ÒW©þ"yœªœ‚¹°Ÿ×Ï]jRÖ‘Êb¹føÓÄgχq×KÆqüœÎ7š*<²Á’ÑÑ3Ðn­°w††{=ÁÀ4©Ýö„{ÌëÜEûž)ŸÆR­;]/ÔɺìÉy»‡¡íyB˜Õ‡Ê-ð”‡ÕŠáÍp¼ÇaJš1„ Ð?QÛ·íÉþ Éì\0Òf<¼ØñЊñÂŽÌ Ô¼:ιèÜÏÝÙ#ÁÖ3ÑÐOq9{ù¸Ø6}©½f뜣3²CÅ„Ð$¡Þ3$låL”5„¾ß½Ú±/—<¶8F¡Ö Aà2„ÕËYÜR|#ä.µV +Ø…mÕ“WÀEpˆáÊQs‹Ç‘Î AÙÊw}½*ͽÙ0ár¦âѼÛU×þ¤Ü ÆŒõè؇¤ú1°'¿yß¹¢¯Kaw½•ªç ¶(‹á­ÁŒñ¢çp‡€Ü!š0,~`ÞGaw„­ÛÖ •Æà‰aæíØtn‰Eð¡?Лâ ÍÆÑÜü'ÃÞ¬,eä6…M7ºÅÅ]“Á'*DÜtü~$Ðç7½áÆRoN‘màd_ÞËq•›Cû­ŒÓßM; •? èïx‘Õ2”n,Þâeìo<ÚÀÈŠÑä Ñ ©8„p1ð…«aå&ïQQï·sðzóÎm•–ÝŠ!åAåP/ÝsíOïPªtm¶þÒçñ„žêî­Á¤«ÂðRaááòãÞb®`£é!ÁK…Ž!t%Æ\öñ7Qž0½bs 4O>Žâ–~/q$´­SêŽÙÀÎìÙa¦Ë°ÙK?>{\@ßÍZÿmˆõòAêžâÌUQLžÐ¿‚О 6ì‡aßï<më‡q“Eualõvúë8âèoÿ‹`Ñ´(_ÕÇ~ÙØ P¥v7`ž#!Þ^ÍèÜÁö¢ø‹ª„|}…ïk)lhœÝ)È}|qN›˜9í#ìCÞ±ê&ýÆ•ötŠaÔ‘•r¹ö'‹&0òSû¡ûÓ(ô}þVi¯!®zýÈ›ß!ý²`'PýOQLwf1Z%éyN=”£ÀÜﶯ„¯oB O©ú\"ÃZQL†'CEݹéüDV²pàÑI @  D˜PáB† I „(pL/#²±˜±S”"5~r  MžD™RåJ†Xj A'PžÁ,ÁÒä‹:}þTèP¢Eƒ²¸4pS‘"›f"”è΄4>2RäÑG!ÙQÂIäX²2MÊ@sêZ¶ ]ªÔÀ\ºuíåÙVï^¾}ý®QQ‘(ú»RŒ^Š\á*r“’Že-‡Ì)ñfŸoSÚþoOÎ¥‰j6h¦¿¢˜ „jƒ>BbjFc×Æ„/ÿÎCmâ =G˜ùr†j™??xçÄ+EEœM;G!®`ÅZäq°VàåÞ‘ ]õqæÊÕ¿‡¿i&œ˜B‚è§ñJ?>râ66$P tj N(ƒÍ<àЊÏ4öLŠ©!˜Šr/B 7ôë¡+¢*).³aj«‹@BÃ;ËŒÃÄ&l¨B† CÒdä±Ç¡29Á5Šb+¨8&É1ÈV¼Š©F^üM8û¢¡CBç,D.ˆª¨ " Ä¿–*B7×4‚„²Š ´þÌ"¿l«Êƒ®ÄÒ®ôØOA=¨$J¢;àXâ-<â|ì;:Ǻäç-JO…lì+CNGåÑÄN8±½4òÐZ,¢ÁN$¥4¬KËŠ1R;]IƒýüuWaß[ ¢D‹(æù¾ÿ`mP‘ˆjÅïÖ±ô–(OU 6Û PóV¨¸h"Þ.!Üâ< V1ªh6(S^kAº$­pƒÚ¡>ë t_¡6X§ "r¢ˆ'Z”`íÈm—)ÂØ0Ã1¦ê½w¬\ þ©ßƒþéL¢ºå¸ä½2‰Aa]ã@"XE®Ø±©Nb¦®PÄÒŒ5ÊÄôLþ^Éc“*éÔ¡H:é¢à€¸“Á”ÀN=GËÚD1¼@ÃÞ?ÊÔa¥[*J~=zǶÉᘿö`ˆZ´ÂÙN.) :¦·Æ[J¶Q% x*AX¸ B ü˜rí½ÊÄ„ˆªûBÑ;@qœ6 ñÞ¹;o¨ï“.ÙìÓ2˜#)‹b`ÁkÐ}„c:™"¢ÔenjέÕôu“"0*€KŸê š¶È(ñÏt!BQ Þ>.öÝoýÑæÁÊ“œø¥OP_.ß¾GO‚R ç®[½õ¯‡ò,êÓ?¨oNü£Õ$Ë Ð§!ܽN²Huþ®Ã0@¢€F*ý€#›L<0}úëæ “ /!H[IÄt 0>®Û .!7ûÁ^·3 UsÆERô»Ã~dè;ÑMÅð`rLçÄ•§„øÓ'½A$;ñQ¡ %‘|í°y=,Ê!ä²¢D’3Ï‘8Œp _ãJqŠÙØ‚E¡|` é¢I(qç ”ãsÜF6YnCèãnÞØ B<1ù+z0ɹpqˆ*ñ"étæ½'sl»Ã"Ò‚ë“ÅÑC³DrCÎ\ d3ѹ€¯Y>©$/™ Ï<™‚å#iþ£I&B͉2$HX¹µ9hO˜\‰„š?é$— !&Ùƒz”ñ&(àйA©ká4«Ð‰9p~3†î¬ <שNyÖó ô¼g;Ëyê3þdç;ó)Ðz¨aû¨Ø|4ž}¨CûiІ´&e´ˆÊd ½/cù2?í)Ñ‘Zt¢…H!zR––T“Øæ —’JÌåºÅ!ìL”\Sˆ>¡Ä­U}=Ó4Ä$¤›ÎcB %S‘Á)åØ·`ƒ ™r²Ÿ$ ñI'¢Zq3LÚæFyÛ© ­È ¥W$ˆ…¼•‡+ €–‚þ€°¤«ù*BX༢¨È1«P*‡þÅçH‰^Ô’XÕdÀ*‘¹^BY¶éOd !Jþ ŠÀd ÝQíJ(´e ~øH‡Õ–F‰A£}” >bvwq4*Bú–׆ðÕ¯Aôê.â Ðoê XW ¤gçƒNkÇ(µÒ™Ò!؇ÛÝe¢d .4SRÕ«&$«[-r«\…€åÄ0‡•I§uË©„+3’¨œ ¶ê!~ãu=pö‚+©)nº';F7!£ ¬ P›YìÒÆÂ,É„eEâN'“cMUl”•žBÑ—ôÓÛB>ü8þÑI34Õ<®%S’ÉEV0ôíÑ«û‘Lè—6=èÃnŠÆLb>Êä‹ñÖ³Ÿ¥W¸:™%ŽærâR¾¤²A2 åëÍVoÇʵ§Fà ˆ!2@rÉ%ˆÁš?zX+«÷9y„ÆiX43Ç2¦ÚÌk„M È A,Î÷A‹ûLÇPY6‹´H&ÊZšŸd°å1òfìû‘Áæ˜u’'zp‚íîl–ÎlH©;ûž.’ Ñ f$Ÿý2i„À€ÀIµ_X ʈ¡Lér¿”‰\z6çv,LNó„¶FFÍ#ñ¾¨È‹æ ̨'þA„±Ùº ŽmÎ è:–¼V.oÃý¢òHÍtJv[Ô©ê)& ö†î?\ú2QísèðÝ1ïû#9ä'.ðë±@z ŸÊ²W¹äˆÜOXžÈ°=–?F\âê!B-à”/3^È©Ë#©skZ±³ Gmô"ç¢Dúk”1õ$GÿZ·÷Ò-ÏœÈ]‹±ÄÊLb*=Ç#òhw7V`²ôXæPt—gâ ©¸Õ?‚-6 uç4g»O*¡ô²¬Á7‘4Ð*}Y²°®í&‘º^‚¥u¹ƒ¸ÏÑøÊW’9ü÷'‡`ù*H.<› JËdžE2‰þž{ÐU€Þø¹õîd:94áͳ̑\:9¨ûC»Ê ±·‘u[d®\ ·ß~¤²^#>ã g]ˆÕ_å@là(<ÚdÀæ ÉH úß  ™ãšS0°ÏbPùAá¦|˜›-mÎÉ  ®—é_Šö™„ÄÜÉ„CØñ#³±ø .ˆˆÀ{›È ¸€ø>Ó{9ÔÛª>üCÄŠ;:ù?‹¸„ˆ T¸€€ƒZs*(‚5  4½Äk å81ü‰¦#ãIÀßAú!2` 082<<d åè%瓲¶¢ˆ.O˜þ.É»Œ œ¢¯¸’³SŸ ®ýK&\‹¼!üÓÂN ;ÎÀƒÅÙd=ÊXr@%Œ83œŠÈ0r¾5¤ Í8„ñÃF`½‚sr$<´2=4Š8ÞRC˘“c‹Ü#?̘qŸ÷™„ßcÄàrD ‰Ùc½5„£  u¡„ñÃBM,ÊP„GP‘Lø0„µÃ ¾2|èXDŹ™©¢P OØ»Nð¸Ž{6˜E9Dï€xJ¸×¹Æ‚E¢Ð9«Ó¨¹¾í 5Œ[ 8|*‘P/  ›a /ˆÂÆkY±® ÅM{nœÄß@¾¢±âþó­xŽ%ð‹‘*È«k¡FЙBÅ{|¦m<t´8T9£YD ¸ƒ°+ ÚÛ„dƒGÐT°ð.` '°ˆ[”H|T} ·M0g 4P9¼ ”QÅ•¸³Û±Œe%[Ʊ›¦h|#g£G¨F”C£"C ¡H¡ÉEâ„tÇXQJ‘ÀºŸØ»$/0Ä0ƒ›Gà„¯ ­‘ †´ˆ xI˜D=‹œ"H È(ÉNh„4hÇ+èÊ GÜ“ÄøÆÉ>ÙN€–À<‘vdƒÕ‹Ç[ù W{Š¨˜ËFÌG»¤jÁ-zqK‘£•¸ÍŽ®ÀþÆdŠ¦¤ššaÍ7¹Ê”ÃÈÌôªÊ£àÌÝQʈ‚ÐœzYËÈ´•àíËÈšIÉ߀/Ì`TG‘,”¸=]¤?è¸ÊëÀ ‰M¨™ÉT4 “ €ÌÂ3‚ï,ììJç Ž¸!{<‰l|Àê|ŽëäœÉ Ç„N¨Œ)‹"›B„X¸dt£` ó4¿kKð°–ñ(J?ÀµÚ”ÏöÈÍ­qŒ(ÀÏQª ½ ÕZ.B|‘fä(0`Š- ½ŒGØM°Ðõüˆ Ž¨„Ð‹É Í¶LK ”9±ŒB8h¬Á A(‘–TyM¡EHH$ ƒ Gó(¸¢”þÅͼKHŽ‚sÎ"û ¸˜ƒIèÀË $à„ =Ñ­ðL`ÑŒpŒ×ü@•€QašJ“¹M¼ Ñ[ÑOdÙ¹v+0-긂%±Z‹4]ÓˆH•¢ü.§“Ò1ªÓŸ ÏUQq6[ùÃûÐ<1ƒ!LŒ9Tˆp ˆØÅFåÅíÌÜx!¹Æ{„‰(.`ƒ'¸>µ ZÌÏýl„Pˆ³$ŒØHTE¢Gõ‰¸S:U:)¸àœ¹!96ƒŒÑÕú(Óš©œ¡®{²“8ÕæùÖ’1VˆÔ[I•ýJå?åÇ°•¡Ö÷ÑÕ<-\% @ Ì9`Ô†þ€Ï½é×pW–(×Ká&£h-Ò7:–`…×{©ÖN¨VN°É5eKßì„;$Ö©¢RÎaPk)8t]ÕgkØA•WZN3µÜð<‹°ÄŒÝž€]‰mUÙ̘ EÙÉSA_µˆ‡}X5eŠÞ´Œœ‰OtÙG‚Y•Ù­ƒNk±Ož=Í"@‚Y¤ÅXe„ª%Éx…,zIN8á%PÔè\<‰8}¤9W^KÖò°Ï®¥“…µX¼)¸4ÀÚX­Û› yÙÔN=x©/Y§ÎÖ£}ÙÝ™±ÌQ¬¤ ƒÍÈÝùÙüdƒ¬ J€•+[ˆhYÃ}¤M‰¥ŽMÈ þ{…•u[ mÚÈåœÉ;„¼g¤E6@H40Sõ\UÕ©Ìëù °• `MK¨ ×íÙšÙJ°XH‘€!,3½påÐE dÅYY¼W…\¼ÙÖtÄã͈GH4ØO/Ø˲(Ü÷4½õé5µµÍ¿»žTqÖ­Ô€ÜAëÂ_‘ðÞŒq²ÜÕݺÌh%^‘PWΩØò˜88 ¨ èèȸƒdÜß{y7ÿ-VÄ¥ÌÕÝhÍÞAeÕËP`È™Ð#ô8vÝš§ä`üa_“]Ë Ù7jÖê’:Á€p0Y#>‹;ߧ3=´^¦“§ÁÛ[)à{iþâkŠ;8`ÃÝ–a$†ö6¾¶œ_€Ëë¾[¤Áâ,ÞÝK%ΆˆT:(ÁÒ²¨<¶Ø´ãœÎMã¨S[kiS(ñ ŽZà=]9)‹1Ö‹Ž|EsêÖ•€^ЉdÝLxá €ÀÖ_¦ØÜ[`Ú¹ËpUít<Õбж½È=8. þ õ½õ‹JÐ ˜ádPMR½ ’P—âÓÇ}‘JÛËüÏ©¸?l%YîãAIZ úƒè ½,÷ÝN>…‡)Íß1\ž)ϲh°¿èfË/>ŽæTM ±qOP€ @êåîX\þ~ЃЃñúH¦ÐIË(g²ø}m‹Ï[‰v>ÜÍøž:;( ;ˆÁre‹€VÓåšaTˆ4¼Œ4`+€.в¸„!n ûšè±`g"><#ÖFÎÀ‘µ©©™¢^!a\éIæPŠ–‚&‹€.¼šæ¿Â´–þUè+âŒÑº&Z£.~xV ³ G ®ÌÇ|„îÜÑGö‹rËÚj¶ñ„CëCžÍ0jBêA#ødèÔh)Æ+^¬M7Ñ"‚0®ÊÕé²hkŠ¨ŠR)’Šhu¾Ϩ¨ÃþëÄöëÅF)Äfì–Z)—RìÆžlÈìË6)É~lͦ쪄°¾ƒþ±NŒ²ŽhmÖˆ@Ž6D ŒëªM„DxEœ´ðmG/Xk‹ +Úظ’Έ<ç¥e¾ ílþÐûpc0éL L°;ýU¼´/ È‹±èŒXæÄз[AãîIßáÞ ù¾—.íi`¶Œ“.¶_$‹‡å„F@%°Üwdƒäî8ph7ÓcàPï®>b†‡›©gˆ`¼‹„Ö  ŽkŠ†gLf‘€áâ(±‹Eño¢Þ ±ù,1 ðòž=øåG4£gnð­QmâÈA’IÎpB™æ±±fÀfP(ðú°^dëï„Ю±È`æF ü%¬Ö»"ï½°þeÏeØeÁ2m!1&àhA «±(Ó­êç $3Ç{ñ½©äKvòæõª(¥¹ ?ØoúYnùi€FŽˆ7 sЙᆸñ6F`#W _ 7;‡0¿s?^ã˨Ž„ ®þ§>cŸ&Žˆò^ gL__-öqO6‘V ¸~#Œ†ŽLºéØ0ôù;õ€gÇs`_“½þ £]‹'¼ž@”‘9¢PéÇi÷}qu.2íÞþíÓ±e‹FÛ-Ï8 ì:G7oßµD‰ŠvPÿ ?àç­Iu éH¤|!v€‡wpÇ­&2g‰[ç/ñAˆ ªˆ—x€ýã²PsàÀð½X5f"è ±ÂpùÀضÐt¥™ùa‰w…¨a%vS¾æ³–j—#V7¡'•›Oˆ.ù߈ҾªIñ‡zP0úäX[7}ß‹"²Ù ú¨ÏØ©q‘Húߘ²¸¥{’ðúµ—zŠ«ïÓ ÷‹ òš˜wg¼ßt׈±çGFcðòxz¶x°7™…ûò(„¼ã‹¦‡¹'|‰7|¿S¿¿Œù°1ñßÈöþȇúɇuHüòú¾Hgóà­®»7Ö/zŠ`G¦£rÆ&Õ?»$Ÿý ý]ûŽˆ;ÈoΠ{)ôüµŸüH}(10Ï€S‡ˆ.~¯}ÎÌÅ¥“CŽ]/ µþ¨?~Îd“뽈4=ÂÂIÜß—Ìǽϖé×gÓY){æ0ÿÕË×ï~Týþïí½ˆž@,hð „ ¬P§ˆ'‚8Äð"ÆŒ7rìèñ#È"G’,iò¤B (K²ð3Q¢—"^^Ò¬i³‰•#=](d3ÆB‡-jô(Ò¤GU*ÍØ’&£"E ݬZ³PŽ¦=‘8þqGâ¥9à̤õlQ³h×¢TËö-F¦p >•È‰“’"h¬òøCÐÜ„™àd¨ƒDЂ31ä‹#–˸®D4E”pêë僙(….mú4êÔ-~áRâ#©<Vmû6îܺO²ž‹¹S™´ùf"”u7òäÊ—Ûî 3›Í›†óýaQw%æl³k?˽ûJço]GÜ$• u¾À³oïþ=IñlëƼ’Þj¦œð÷óïÏ_þZOE7Õ}UùA– *¸ n¢Õ$çXU ¤íæƒ&a˜áHrȃg±ðGpöMhS&8$7Ù‡"±ØâG/ˆPˆZ½Ð…tþ'ÚtÉ$3úø#(ÕØÔR)¢£M~|$“M:WdP!’6éñ$–Y>9¤R0tÑY•4í¡ÜwZ.d¡™ •é$—IµðZ˜]BHšuÚ¹`›Hµ”Iœé±ä Ú^žG‘קDs º(£ÊjÔo}þ°œ‡Vjç¥@>ZT¤aNbÇr2Þ)j¤þ¸)QV G¦ºú*d¨u(¢5Àz+®‘É*”ªHª˜+°Á¢µ«N½ž8ÉÌ%¶hq.»edÆNšÉõÐè!×-zm–Ä®Dk˜™€Æ·‹fÛh¹ÐB&mã.—® ç2 o“Þ¢Änz…л[ ØöÑh¿ÝFÖþœaþÀ$\à ;¬0ÃKœðÃ+lqÃC\ñÆOÌðñÃ$k¬1Ç(gÌ1Å—Œ±Å-‹üòÆ3«¼°ÉÛü1Ë:ãLóÏ'¬rÎ) s͸LtÈ1û\ôÊPKM4ÑKóµÌA_ôÅ“ û5Øa‹=6Ùe›}6Úi«½6Ûm»ý6ÜqË=7ÝuÛ}7Þyëö šÉ À ­bùƒß{Û‰@âwV"80Àšfz¸âiVR/€×  ŽÀ‡hÐyæYêÑ9ë®®ÁA§_iæ ¬£žú“‡ô°ûívÖ>{A«¿¦»÷Ð;îX"Ÿ&¬Yþ º|òMR¯åõ×g©}õ?rÿdöž§ù}÷0’o½øÏäúå+§üñËBíþôé»Ïžíý·Žý~Ô¼5"zùÛúõ»ƒï€ R}ôºØxuŠà÷ƒÁ1®\Ç#3è£ ÂèqˆÜä*g&Šú3àG8Ãï…-¼!s¨Ãò°‡>ü!ƒ(Ä!±ˆF3, if norm(s2)>0, s2zero = 0; end, end % s2==0 ? if s2zero % log probability evaluation lp = lZ - 2*logcosh(tau*(y-mu)); s2 = 0; else % prediction lp = likSech2(hyp, y, mu, s2, 'infEP'); end ymu = {}; ys2 = {}; if nargout>1 ymu = mu; % first y moment if nargout>2 ys2 = s2 + sn.^2; % second y moment end end varargout = {lp,ymu,ys2}; else % inference mode switch inf case 'infLaplace' r = y-mu; [g,dg,d2g,d3g] = logcosh(tau.*r); % precompute derivatives if nargin<6 % no derivative mode dlp = {}; d2lp = {}; d3lp = {}; lp = lZ - 2*g; if nargout>1 % first derivatives dlp = 2*tau.*dg; if nargout>2 % 2nd derivative of log likelihood d2lp = -2*tau.^2.*d2g; if nargout>3 % 3rd derivative of log likelihood d3lp = 2*tau.^3.*d3g; end end end varargout = {sum(lp),dlp,d2lp,d3lp}; else % derivative mode lp_dhyp = 2*tau.*r.*dg - 1; % derivative w.r.t. sn d2lp_dhyp = 2*tau.^2.*(2*d2g + tau.*r.*d3g); varargout = {lp_dhyp,d2lp_dhyp}; end case 'infEP' n = max([length(y),length(mu),length(s2),length(sn)]); on = ones(n,1); y = y.*on; mu = mu.*on; s2 = s2.*on; sn = sn.*on; % vectors only fac = 1e3; % factor between the widths of the two distributions ... % ... from when one considered a delta peak, we use 3 orders of magnitude idlik = fac*sn1/(2*tau^2) & ~idup; % interesting zone is in between [g,dg,d2g] = inv_xcothx(2*tau^2*ga(id)); thg = tanh(g); h(id) = 4*logcosh(g) -2*g.*thg; h = h + y.^2./ga - 2*lZ; g1mt2 = g.*(1-thg.^2); % first derivatives dh(idup) = 4*tau^2; dh(id) = 4*tau^2*dg.*( thg - g1mt2 ); dh = dh - (y./ga).^2; d2h(id) = 8*tau^4*( g1mt2.*(2*dg.^2.*thg-d2g) +d2g.*thg ); % second derivs d2h = d2h + 2*y.^2./ga.^3; id = ga<0; h(id) = Inf; dh(id) = 0; d2h(id) = 0; % neg. var. treatment varargout = {h,b,dh,db,d2h,d2b}; else ga = s2; n = numel(ga); dhhyp = zeros(n,1); idup = 50<=ga(:); % asymptotic behavior for large gammas dhhyp(idup) = -2*pi^2/3/sn^2*ga(idup); % tau = pi/(2*sqrt(3)*sn) id = ga(:)>1/(2*tau^2) & ~idup; % interesting zone is in between [g,dg] = inv_xcothx(2*tau^2*ga(id)); thg = tanh(g); [f,df] = logcosh(g); % h = 4*f -2*g.*tanh(g) if any(id) dhhyp(id) = 8*tau^2*ga(id).*dg.*( thg - 2*df + g.*(1-thg.^2) ); end dhhyp = dhhyp + 2; % from lZ = log(pi) - log(sn) - log(4*sqrt(3)) dhhyp(ga<0) = 0; % negative variances get a special treatment varargout = {dhhyp}; % deriv. wrt hyp.lik end end end % Invert f(x) = x.*coth(x) = y, return the positive value % uses Newton's method to minimize (y-f(x))² w.r.t. x function [x,dx,d2x] = inv_xcothx(y) % f = @(x) x.*coth(x); % df = @(x) x + (1-f(x)).*coth(x); % d2f = @(x) 2*(1-f(x)).*(1-coth(x).^2); if numel(y)==0, x=[]; dx=[]; d2x=[]; return, end x = sqrt(y.^2-1); i = 1; % init ep = eps./(1+abs(x)); th=tanh(x); fx=(x+ep)./(th+ep); % function r = fx-y; % init while i==1 || (i<10 && max(abs(r))>1e-12) dfx = x + (1-fx)./(th+ep); % first derivative d2fx = 2*( 1-fx + (x-th+ep/3)./(th.*th.*th+ep) ); % second derivative x = x - r.*dfx./(dfx.^2+r.*d2fx+ep); % Newton step ep = eps./(1+abs(x)); th=tanh(x); fx=(x+ep)./(th+ep); r = fx-y; i = i+1; end % first derivative dx/dy; derivatives of inverse functions are reciprocals dx = 1./dfx; % second derivative d2x/dy2; quotient rule and chaine rule d2x = -d2fx.*dx./dfx./dfx; % numerically safe version of log(cosh(x)) = log(exp(x)+exp(-x))-log(2) function [f,df,d2f,d3f] = logcosh(x) a = exp(-2*abs(x)); % always between 0 and 1 and therefore safe to evaluate f = abs(x) + log(1+a) - log(2); df = sign(x).*( 1 - 2*a./(1+a) ); d2f = 4*a./(1+a).^2; d3f = -8*sign(x).*a.*(1-a)./(1+a).^3; % computes y = log( exp(A)*x ) in a numerically safe way by subtracting the % maximal value in each row to avoid cancelation after taking the exp function y = log_expA_x(A,x) N = size(A,2); maxA = max(A,[],2); % number of columns, max over columns y = log(exp(A-maxA*ones(1,N))*x) + maxA; % exp(A) = exp(A-max(A))*exp(max(A)) % computes y = ( (exp(A).*B)*z ) ./ ( exp(A)*x ) in a numerically safe way % The function is not general in the sense that it yields correct values for % all types of inputs. We assume that the values are close together. function y = expABz_expAx(A,x,B,z) N = size(A,2); maxA = max(A,[],2); % number of columns, max over columns A = A-maxA*ones(1,N); % subtract maximum value y = ( (exp(A).*B)*z ) ./ ( exp(A)*x ); gpml-matlab-v3.0-2010-07-23/lik/likT.m000644 000765 000024 00000021122 11422240322 016677 0ustar00carlstaff000000 000000 function [varargout] = likT(hyp, y, mu, s2, inf, i) % likT - Student's t likelihood function for regression. % The expression for the likelihood is % likT(t) = Z * ( 1 + (t-y)^2/(nu*sn^2) ).^(-(nu+1)/2), % where Z = gamma((nu+1)/2) / (gamma(nu/2)*sqrt(nu*pi)*sn) % and y is the mean (for nu>1) and nu*sn^2/(nu-2) is the variance (for nu>2). % % The hyperparameters are: % % hyp = [ log(nu-1) % log(sn) ] % % Note that the parametrisation guarantees nu>1, thus the mean always exists. % % Several modes are provided, for computing likelihoods, derivatives and moments % respectively, see likelihoods.m for the details. In general, care is taken % to avoid numerical issues when the arguments are extreme. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-21. % % See also likFunctions.m. if nargin<2, varargout = {'2'}; return; end % report number of hyperparameters numin = 1; % minimum value of nu nu = exp(hyp(1))+numin; sn2 = exp(2*hyp(2)); % extract hyperparameters lZ = loggamma(nu/2+1/2) - loggamma(nu/2) - log(nu*pi*sn2)/2; if nargin<5 % prediction mode if inf is not present if numel(y)==0, y = zeros(size(mu)); end s2zero = 1; if nargin>3, if norm(s2)>0, s2zero = 0; end, end % s2==0 ? if s2zero % log probability evaluation lp = lZ - (nu+1)*log( 1+(y-mu).^2./(nu.*sn2) )/2; else % prediction lp = likT(hyp, y, mu, s2, 'infEP'); end ymu = {}; ys2 = {}; if nargout>1 ymu = mu; % first y moment; for nu<=1 we this is the mode if nargout>2 if nu<=2 ys2 = Inf(size(s2)); else ys2 = s2 + nu*sn2/(nu-2); % second y moment end end end varargout = {lp,ymu,ys2}; else switch inf case 'infLaplace' r = y-mu; r2 = r.^2; if nargin<6 % no derivative mode dlp = {}; d2lp = {}; d3lp = {}; lp = lZ - (nu+1)*log( 1+r2./(nu.*sn2) )/2; if nargout>1 a = r2+nu*sn2; dlp = (nu+1)*r./a; % dlp, derivative of log likelihood if nargout>2 % d2lp, 2nd derivative of log likelihood d2lp = (nu+1)*(r2-nu*sn2)./a.^2; if nargout>3 % d3lp, 3rd derivative of log likelihood d3lp = (nu+1)*2*r.*(r2-3*nu*sn2)./a.^3; end end end varargout = {sum(lp),dlp,d2lp,d3lp}; else % derivative mode a3 = (r2+nu*sn2).^3; if i==1 % derivative w.r.t. nu lp_dhyp = nu*( dloggamma(nu/2+1/2)-dloggamma(nu/2) )/2 - 1/2 ... -nu*log(1+r2/(nu*sn2))/2 +(nu/2+1/2)*r2./(nu*sn2+r2); lp_dhyp = (1-numin/nu)*lp_dhyp; % correct for lower bound on nu d2lp_dhyp = nu*( r2.*(r2-3*sn2*(1+nu)) + nu*sn2^2 )./a3; d2lp_dhyp = (1-numin/nu)*d2lp_dhyp; % correct for lower bound on nu else % derivative w.r.t. sn lp_dhyp = (nu+1)*r2./(r2+nu*sn2) - 1; d2lp_dhyp = (nu+1)*2*nu*sn2*(nu*sn2-3*r2)./a3; end varargout = {lp_dhyp,d2lp_dhyp}; end case 'infEP' if nargout>1 error('infEP not supported since likT is not log-concave') end n = max([length(y),length(mu),length(s2)]); on = ones(n,1); y = y(:).*on; mu = mu(:).*on; sig = sqrt(s2(:)).*on; % vectors only % since we are not aware of an analytical expression of the integral, % we use Gaussian-Hermite quadrature N = 20; [t,w] = gauher(N); oN = ones(1,N); lZ = likT(hyp, y*oN, sig*t'+mu*oN, []); lZ = log_expA_x(lZ,w); % log( exp(lZ)*w ) varargout = {lZ}; case 'infVB' if nargin<6 % variational lower site bound % t(s) \propto (1+(s-y)^2/(nu*s2))^(-nu/2+1/2) % the bound has the form: b*s - s.^2/(2*ga) - h(ga)/2 with b=y/ga!! ga = s2; n = numel(ga); b = y./ga; y = y.*ones(n,1); db = -y./ga.^2; d2b = 2*y./ga.^3; id = ga<=sn2*nu/(nu+1); h = (nu+1)*( log(ga*(1+1/nu)/sn2) - 1 ) + (nu*sn2+y.^2)./ga; h(id) = y(id).^2./ga(id); h = h - 2*lZ; dh = (nu+1)./ga - (nu*sn2+y.^2)./ga.^2; dh(id) = -y(id).^2./ga(id).^2; d2h = -(nu+1)./ga.^2 + 2*(nu*sn2+y.^2)./ga.^3; d2h(id) = 2*y(id).^2./ga(id).^3; id = ga<0; h(id) = Inf; dh(id) = 0; d2h(id) = 0; % neg. var. treatment varargout = {h,b,dh,db,d2h,d2b}; else ga = s2; n = numel(ga); dhhyp = zeros(n,1); id = ga>sn2*nu/(nu+1); % dhhyp(~id) = 0 if i==1 % log(nu) % h = (nu+1)*log(1+1/nu) - nu + nu*sn2./ga; dhhyp(id) = nu*log(ga(id)*(1+1/nu)/sn2) - 1 - nu + nu*sn2./ga(id); % lZ = loggamma(nu/2+1/2) - loggamma(nu/2) - log(nu)/2 dhhyp = dhhyp - nu*dloggamma(nu/2+1/2) + nu*dloggamma(nu/2) + 1; % -2*lZ dhhyp = (1-numin/nu)*dhhyp; % correct for lower bound on nu else % log(sn) % h = (nu+1)*log(1/sn2) + nu*sn2./ga; dhhyp(id) = -2*(nu+1) + 2*nu*sn2./ga(id); % lZ = - log(sn2)/2 dhhyp = dhhyp + 2; % -2*lZ end dhhyp(ga<0) = 0; % negative variances get a special treatment varargout = {dhhyp}; % deriv. wrt hyp.lik end end end % Returns the log of the gamma function. Source: Pike, M.C., and % I.D. Hill, Algorithm 291, Communications of the ACM, 9,9:p.684 (Sept, 1966). % Accuracy to 10 decimal places. Uses Sterling's formula. % Derivative from Abromowitz and Stegun, formula 6.3.18 with recurrence 6.3.5. function f = loggamma(x) x = x+6; f = 1./(x.*x); f = (((-0.000595238095238*f+0.000793650793651).*f-1/360).*f+1/12)./x; f = (x-0.5).*log(x)-x+0.918938533204673+f; f = f-log(x-1)-log(x-2)-log(x-3)-log(x-4)-log(x-5)-log(x-6); function df = dloggamma(x) x = x+6; df = 1./(x.*x); df = (((df/240-0.003968253986254).*df+1/120).*df-1/120).*df; df = df+log(x)-0.5./x-1./(x-1)-1./(x-2)-1./(x-3)-1./(x-4)-1./(x-5)-1./(x-6); % computes y = log( exp(A)*x ) in a numerically safe way by subtracting the % maximal value in each row to avoid cancelation after taking the exp function y = log_expA_x(A,x) N = size(A,2); maxA = max(A,[],2); % number of columns, max over columns y = log(exp(A-maxA*ones(1,N))*x) + maxA; % exp(A) = exp(A-max(A))*exp(max(A)) % compute abscissas and weight factors for Gaussian-Hermite quadrature % % CALL: [x,w]=gauher(N) % % x = base points (abscissas) % w = weight factors % N = number of base points (abscissas) (integrates a (2N-1)th order % polynomial exactly) % % p(x)=exp(-x^2/2)/sqrt(2*pi), a =-Inf, b = Inf % % The Gaussian Quadrature integrates a (2n-1)th order % polynomial exactly and the integral is of the form % b N % Int ( p(x)* F(x) ) dx = Sum ( w_j* F( x_j ) ) % a j=1 % % this procedure uses the coefficients a(j), b(j) of the % recurrence relation % % b p (x) = (x - a ) p (x) - b p (x) % j j j j-1 j-1 j-2 % % for the various classical (normalized) orthogonal polynomials, % and the zero-th moment % % 1 = integral w(x) dx % % of the given polynomial's weight function w(x). Since the % polynomials are orthonormalized, the tridiagonal matrix is % guaranteed to be symmetric. function [x,w]=gauher(N) if N==20 % return precalculated values x=[ -7.619048541679757;-6.510590157013656;-5.578738805893203; -4.734581334046057;-3.943967350657318;-3.18901481655339 ; -2.458663611172367;-1.745247320814127;-1.042945348802751; -0.346964157081356; 0.346964157081356; 1.042945348802751; 1.745247320814127; 2.458663611172367; 3.18901481655339 ; 3.943967350657316; 4.734581334046057; 5.578738805893202; 6.510590157013653; 7.619048541679757]; w=[ 0.000000000000126; 0.000000000248206; 0.000000061274903; 0.00000440212109 ; 0.000128826279962; 0.00183010313108 ; 0.013997837447101; 0.061506372063977; 0.161739333984 ; 0.260793063449555; 0.260793063449555; 0.161739333984 ; 0.061506372063977; 0.013997837447101; 0.00183010313108 ; 0.000128826279962; 0.00000440212109 ; 0.000000061274903; 0.000000000248206; 0.000000000000126 ]; else b = sqrt( (1:N-1)/2 )'; [V,D] = eig( diag(b,1) + diag(b,-1) ); w = V(1,:)'.^2; x = sqrt(2)*diag(D); end gpml-matlab-v3.0-2010-07-23/lik/likLaplace.m000644 000765 000024 00000025525 11422240322 020050 0ustar00carlstaff000000 000000 function [varargout] = likLaplace(hyp, y, mu, s2, inf, i) % likLaplace - Laplacian likelihood function for regression. % The expression for the likelihood is % likLaplace(t) = exp(-|t-y|/b)/2*b with b = sn/sqrt(2), % where y is the mean and sn^2 is the variance. % % The hyperparameters are: % % hyp = [ log(sn) ] % % Several modes are provided, for computing likelihoods, derivatives and moments % respectively, see likelihoods.m for the details. In general, care is taken % to avoid numerical issues when the arguments are extreme. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-21. % % See also likFunctions.m. if nargin<2, varargout = {'1'}; return; end % report number of hyperparameters sn = exp(hyp); b = sn/sqrt(2); if nargin<5 % prediction mode if inf is not present if numel(y)==0, y = zeros(size(mu)); end s2zero = 1; if nargin>3, if norm(s2)>0, s2zero = 0; end, end % s2==0 ? if s2zero % log probability evaluation lp = -abs(y-mu)./b -log(2*b); s2 = 0; else % prediction lp = likLaplace(hyp, y, mu, s2, 'infEP'); end ymu = {}; ys2 = {}; if nargout>1 ymu = mu; % first y moment if nargout>2 ys2 = s2 + sn.^2; % second y moment end end varargout = {lp,ymu,ys2}; else % inference mode switch inf case 'infLaplace' if nargin<6 % no derivative mode if numel(y)==0, y=0; end ymmu = y-mu; dlp = {}; d2lp = {}; d3lp = {}; lp = -abs(ymmu)/b -log(2*b); if nargout>1 dlp = sign(ymmu)/b; % dlp, derivative of log likelihood if nargout>2 % d2lp, 2nd derivative of log likelihood d2lp = zeros(size(ymmu)); if nargout>3 % d3lp, 3rd derivative of log likelihood d3lp = zeros(size(ymmu)); end end end varargout = {sum(lp),dlp,d2lp,d3lp}; else % derivative mode lp_dhyp = abs(y-mu)/b - 1; % derivative of log likelihood w.r.t. hypers d2lp_dhyp = zeros(size(mu)); % and also of the second mu derivative varargout = {lp_dhyp,d2lp_dhyp}; end case 'infEP' n = max([length(y),length(mu),length(s2),length(sn)]); on = ones(n,1); y = y.*on; mu = mu.*on; s2 = s2.*on; sn = sn.*on; % vectors only fac = 1e3; % factor between the widths of the two distributions ... % ... from when one considered a delta peak, we use 3 orders of magnitude idlik = fac*sn1 % 1st moment % M1p = sqrt(tvar).*( zp.*exp(lM0p) - exp(-zp.*zp)/sqrt(pi) ); % M1m = sqrt(tvar).*( zm.*exp(lM0m) - exp(-zm.*zm)/sqrt(pi) ); % lZmax = max([vp-lZ(id),vm-lZ(id)],[],2); % m1m0 = exp(lZmax).*( exp(vp-lZ(id)-lZmax).*M1p ... % -exp(vm-lZ(id)-lZmax).*M1m )/2 + y(id); % dlZ(id) = (m1m0-mu(id))./s2(id); % if nargout>2 % 2nd moment % M2p = 2*tvar.*exp(vp-lZ(id)+lM0p)/2+(2*tvar+sqrt(2)*tmu) ... % .*M1p.*exp(vp-lZ(id)); % M2m = 2*tvar.*exp(vm-lZ(id)+lM0m)/2+(2*tvar-sqrt(2)*tmu) ... % .*M1m.*exp(vm-lZ(id)); % m2m0 = (M2p+M2m).*sn(id)/sqrt(8) + 2*y(id).*m1m0 - y(id).^2; % d2lZ(id) = (m2m0 - m1m0.^2)./s2(id).^2 - 1./s2(id); % end % end % an implementation based on logphi(t) = log(normcdf(t)) zp = (tmu+sqrt(2)*tvar)./sqrt(tvar); zm = (tmu-sqrt(2)*tvar)./sqrt(tvar); ap = logphi(-zp)+sqrt(2)*tmu; am = logphi( zm)-sqrt(2)*tmu; lZ(id) = logsum2exp([ap,am]) + tvar - log(sn(id)*sqrt(2)); if nargout>1 lqp = -zp.^2/2 - log(2*pi)/2 - logphi(-zp); % log( N(z)/Phi(z) ) lqm = -zm.^2/2 - log(2*pi)/2 - logphi( zm); dap = -exp(lqp-log(s2(id))/2) + sqrt(2)./sn(id); dam = exp(lqm-log(s2(id))/2) - sqrt(2)./sn(id); % ( exp(ap).*dap + exp(am).*dam )./( exp(ap) + exp(am) ) dlZ(id) = expABz_expAx([ap,am],[1;1],[dap,dam],[1;1]); if nargout>2 a = sqrt(8)./sn(id)./sqrt(s2(id)); bp = 2./sn(id).^2 - (a - zp./s2(id)).*exp(lqp); bm = 2./sn(id).^2 - (a + zm./s2(id)).*exp(lqm); % d2lZ(id) = ( exp(ap).*bp + exp(am).*bm )./( exp(ap) + exp(am) ) ... % - dlZ(id).^2; d2lZ(id) = expABz_expAx([ap,am],[1;1],[bp,bm],[1;1]) - dlZ(id).^2; end end end varargout = {lZ,dlZ,d2lZ}; else % derivative mode dlZhyp = zeros(n,1); if any(idlik) dlZhyp(idlik) = 0; end if any(idgau) dlZhyp(idgau) = ... likLaplace(log(sn(idgau)), mu(idgau), y(idgau), 'infLaplace', 1); end if any(id) % substitution to obtain unit variance, zero mean Laplacian tmu = (mu(id)-y(id))./sn(id); tvar = s2(id)./sn(id).^2; zp = (tvar+tmu/sqrt(2))./sqrt(tvar); vp = tvar+sqrt(2)*tmu; zm = (tvar-tmu/sqrt(2))./sqrt(tvar); vm = tvar-sqrt(2)*tmu; dzp = (-s2(id)./sn(id)+tmu.*sn(id)/sqrt(2)) ./ sqrt(s2(id)); dvp = -2*tvar - sqrt(2)*tmu; dzm = (-s2(id)./sn(id)-tmu.*sn(id)/sqrt(2)) ./ sqrt(s2(id)); dvm = -2*tvar + sqrt(2)*tmu; lezp = lerfc(zp); % ap = exp(vp).*ezp lezm = lerfc(zm); % am = exp(vm).*ezm vmax = max([vp+lezp,vm+lezm],[],2); % subtract max to avoid numerical pb ep = exp(vp+lezp-vmax); em = exp(vm+lezm-vmax); dap = ep.*(dvp - 2/sqrt(pi)*exp(-zp.^2-lezp).*dzp); dam = em.*(dvm - 2/sqrt(pi)*exp(-zm.^2-lezm).*dzm); dlZhyp(id) = (dap+dam)./(ep+em) - 1; end varargout = {dlZhyp}; % deriv. wrt hypers end case 'infVB' if nargin<6 % variational lower site bound % t(s) = exp(-sqrt(2)|y-s|/sn) / sqrt(2*sn²) % the bound has the form: b*s - s.^2/(2*ga) - h(ga)/2 with b=y/ga!! ga = s2; n = numel(ga); b = y./ga; y = y.*ones(n,1); db = -y./ga.^2; d2b = 2*y./ga.^3; h = 2*ga/sn^2 + log(2*sn^2) + y.^2./ga; dh = 2/sn^2 - y.^2./ga.^2; d2h = 2*y.^2./ga.^3; id = ga<0; h(id) = Inf; dh(id) = 0; d2h(id) = 0; % neg. var. treatment varargout = {h,b,dh,db,d2h,d2b}; else ga = s2; dhhyp = -4*ga/sn^2 + 2; dhhyp(ga<0) = 0; % negative variances get a special treatment varargout = {dhhyp}; % deriv. wrt hyp.lik end end end % computes y = log( sum(exp(x),2) ) in a numerically safe way by subtracting % the row maximum to avoid cancelation after taking the exp % the sum is done along the rows function [y,x] = logsum2exp(logx) N = size(logx,2); max_logx = max(logx,[],2); % we have all values in the log domain, and want to calculate a sum x = exp(logx-max_logx*ones(1,N)); y = log(sum(x,2)) + max_logx; % numerically safe implementation of f(t) = log(1-erf(t)) = log(erfc(t)) function f = lerfc(t) f = zeros(size(t)); tmin = 20; tmax = 25; ok = ttmax; % evaluate tight bound interp = ~ok & ~bd; % linearly interpolate between both of them f(~ok) = log(2/sqrt(pi)) -t(~ok).^2 -log(t(~ok)+sqrt( t(~ok).^2+4/pi )); lam = 1./(1+exp( 12*(1/2-(t(interp)-tmin)/(tmax-tmin)) )); % interp. weights f(interp) = lam.*f(interp) + (1-lam).*log(erfc( t(interp) )); f(ok) = f(ok) + log(erfc( t(ok) )); % safe eval % computes y = ( (exp(A).*B)*z ) ./ ( exp(A)*x ) in a numerically safe way % The function is not general in the sense that it yields correct values for % all types of inputs. We assume that the values are close together. function y = expABz_expAx(A,x,B,z) N = size(A,2); maxA = max(A,[],2); % number of columns, max over columns A = A-maxA*ones(1,N); % subtract maximum value y = ( (exp(A).*B)*z ) ./ ( exp(A)*x ); % safe implementation of the log of phi(x) = \int_{-\infty}^x N(f|0,1) df % logphi(z) = log(normcdf(z)) function lp = logphi(z) lp = zeros(size(z)); % allocate memory zmin = -6.2; zmax = -5.5; ok = z>zmax; % safe evaluation for large values bd = z3, if norm(s2)>0, s2zero = 0; end, end % s2==0 ? if s2zero % log probability lp = -(y-mu).^2./sn2/2-log(2*pi*sn2)/2; s2 = 0; else lp = likGauss(hyp, y, mu, s2, 'infEP'); % prediction end ymu = {}; ys2 = {}; if nargout>1 ymu = mu; % first y moment if nargout>2 ys2 = s2 + sn2; % second y moment end end varargout = {lp,ymu,ys2}; else switch inf case 'infLaplace' if nargin<6 % no derivative mode if numel(y)==0, y=0; end ymmu = y-mu; dlp = {}; d2lp = {}; d3lp = {}; lp = -ymmu.^2/(2*sn2) - log(2*pi*sn2)/2; if nargout>1 dlp = ymmu/sn2; % dlp, derivative of log likelihood if nargout>2 % d2lp, 2nd derivative of log likelihood d2lp = -ones(size(ymmu))/sn2; if nargout>3 % d3lp, 3rd derivative of log likelihood d3lp = zeros(size(ymmu)); end end end varargout = {sum(lp),dlp,d2lp,d3lp}; else % derivative mode lp_dhyp = (y-mu).^2/sn2 - 1; % derivative of log likelihood w.r.t. hypers d2lp_dhyp = 2*ones(size(mu))/sn2; % and also of the second mu derivative varargout = {lp_dhyp,d2lp_dhyp}; end case 'infEP' if nargin<6 % no derivative mode lZ = -(y-mu).^2./(sn2+s2)/2 - log(2*pi*(sn2+s2))/2; % log part function dlZ = {}; d2lZ = {}; if nargout>1 dlZ = (y-mu)./(sn2+s2); % 1st derivative w.r.t. mean if nargout>2 d2lZ = -1./(sn2+s2); % 2nd derivative w.r.t. mean end end varargout = {lZ,dlZ,d2lZ}; else % derivative mode dlZhyp = ((y-mu).^2./(sn2+s2)-1) ./ (1+s2./sn2); % deriv. w.r.t. hyp.lik varargout = {dlZhyp}; end case 'infVB' if nargin<6 % variational lower site bound % t(s) = exp(-(y-s)^2/2sn2)/sqrt(2*pi*sn2) % the bound has the form: b*s - s.^2/(2*ga) - h(ga)/2 with b=y/ga ga = s2; n = numel(ga); b = y./ga; y = y.*ones(n,1); db = -y./ga.^2; d2b = 2*y./ga.^3; h = zeros(n,1); dh = h; d2h = h; % allocate memory for return args id = ga(:)<=sn2+1e-8; % OK below noise variance h(id) = y(id).^2./ga(id) + log(2*pi*sn2); h(~id) = Inf; dh(id) = -y(id).^2./ga(id).^2; d2h(id) = 2*y(id).^2./ga(id).^3; id = ga<0; h(id) = Inf; dh(id) = 0; d2h(id) = 0; % neg. var. treatment varargout = {h,b,dh,db,d2h,d2b}; else ga = s2; n = numel(ga); dhhyp = zeros(n,1); dhhyp(ga(:)<=sn2) = 2; dhhyp(ga<0) = 0; % negative variances get a special treatment varargout = {dhhyp}; % deriv. w.r.t. hyp.lik end end end gpml-matlab-v3.0-2010-07-23/lik/likErf.m000644 000765 000024 00000013105 11422240322 017212 0ustar00carlstaff000000 000000 function [varargout] = likErf(hyp, y, mu, s2, inf, i) % likErf - Error function or cumulative Gaussian likelihood function for binary % classification or probit regression. The expression for the likelihood is % likErf(t) = (1+erf(t/sqrt(2)))/2 = normcdf(t). % % Several modes are provided, for computing likelihoods, derivatives and moments % respectively, see likelihoods.m for the details. In general, care is taken % to avoid numerical issues when the arguments are extreme. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-22. % % See also likFunctions.m. if nargin<2, varargout = {'0'}; return; end % report number of hyperparameters if nargin>1, y = sign(y); y(y==0) = 1; else y = 1; end % allow only +/- 1 values if numel(y)==0, y = 1; end if nargin<5 % prediction mode if inf is not present y = y.*ones(size(mu)); % make y a vector s2zero = 1; if nargin>3, if norm(s2)>0, s2zero = 0; end, end % s2==0 ? if s2zero % log probability evaluation [p,lp] = cumGauss(y,mu); else % prediction lp = likErf(hyp, y, mu, s2, 'infEP'); p = exp(lp); end ymu = {}; ys2 = {}; if nargout>1 ymu = 2*p-1; % first y moment if nargout>2 ys2 = 4*p.*(1-p); % second y moment end end varargout = {lp,ymu,ys2}; else % inference mode switch inf case 'infLaplace' if nargin<6 % no derivative mode f = mu; yf = y.*f; % product latents and labels dlp = {}; d2lp = {}; d3lp = {}; % return arguments [p,lp] = cumGauss(y,f); if nargout>1 % derivative of log likelihood n_p = gauOverCumGauss(yf,p); dlp = y.*n_p; % derivative of log likelihood if nargout>2 % 2nd derivative of log likelihood d2lp = -n_p.^2 - yf.*n_p; if nargout>3 % 3rd derivative of log likelihood d3lp = 2*y.*n_p.^3 +3*f.*n_p.^2 +y.*(f.^2-1).*n_p; end end end varargout = {sum(lp),dlp,d2lp,d3lp}; else % derivative mode varargout = {[]}; % derivative w.r.t. hypers end case 'infEP' if nargin<6 % no derivative mode z = mu./sqrt(1+s2); dlZ = {}; d2lZ = {}; [junk,lZ] = cumGauss(y,z); % log part function if numel(y)>0, z=z.*y; end if nargout>1 if numel(y)==0, y=1; end n_p = gauOverCumGauss(z,exp(lZ)); dlZ = y.*n_p./sqrt(1+s2); % 1st derivative wrt mean if nargout>2 d2lZ = -n_p.*(z+n_p)./(1+s2); % 2nd derivative wrt mean end end varargout = {lZ,dlZ,d2lZ}; else % derivative mode varargout = {[]}; % deriv. wrt hyp.lik end case 'infVB' % naive variational lower site bound based on asymptotical properties of lik % normcdf(t) -> -(t²-2dt+c)/2 for t->-oo (tight lower bound) d = 0.158482605320942; c = -1.785873318175113; ga = s2; n = numel(ga); b = d*y.*ones(n,1); db = zeros(n,1); d2b = db; h = -2*c*ones(n,1); h(ga>1) = Inf; dh = zeros(n,1); d2h = dh; varargout = {h,b,dh,db,d2h,d2b}; end end function [p,lp] = cumGauss(y,f) if numel(y)>0, yf = y.*f; else yf = f; end % product of latents and labels p = (1+erf(yf/sqrt(2)))/2; % likelihood if nargout>1, lp = logphi(yf,p); end % log likelihood % safe implementation of the log of phi(x) = \int_{-\infty}^x N(f|0,1) df % logphi(z) = log(normcdf(z)) function lp = logphi(z,p) lp = zeros(size(z)); % allocate memory zmin = -6.2; zmax = -5.5; ok = z>zmax; % safe evaluation for large values bd = z-5; % naive evaluation for large values of f n_p(ok) = (exp(-f(ok).^2/2)/sqrt(2*pi)) ./ p(ok); bd = f<-6; % tight upper bound evaluation n_p(bd) = sqrt(f(bd).^2/4+1)-f(bd)/2; interp = ~ok & ~bd; % linearly interpolate between both of them tmp = f(interp); lam = -5-f(interp); n_p(interp) = (1-lam).*(exp(-tmp.^2/2)/sqrt(2*pi))./p(interp) + ... lam .*(sqrt(tmp.^2/4+1)-tmp/2); gpml-matlab-v3.0-2010-07-23/lik/likLogistic.m000644 000765 000024 00000017346 11422240322 020266 0ustar00carlstaff000000 000000 function [varargout] = likLogistic(hyp, y, mu, s2, inf, i) % likLogistic - logistic function for binary classification or logit regression. % The expression for the likelihood is % likLogistic(t) = 1./(1+exp(-t)). % % Several modes are provided, for computing likelihoods, derivatives and moments % respectively, see likelihoods.m for the details. In general, care is taken % to avoid numerical issues when the arguments are extreme. The moments % \int f^k likLogistic(y,f) N(f|mu,var) df are calculated via a cumulative % Gaussian scale mixture approximation. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-22. % % See also likFunctions.m. if nargin<2, varargout = {'0'}; return; end % report number of hyperparameters if nargin>1, y = sign(y); y(y==0) = 1; else y = 1; end % allow only +/- 1 values if numel(y)==0, y = 1; end if nargin<5 % prediction mode if inf is not present y = y.*ones(size(mu)); % make y a vector s2zero = 1; if nargin>3, if norm(s2)>0, s2zero = 0; end, end % s2==0 ? if s2zero % log probability evaluation yf = y.*mu; % product latents and labels lp = yf; ok = -351 p = exp(lp); ymu = 2*p-1; % first y moment if nargout>2 ys2 = 4*p.*(1-p); % second y moment end end varargout = {lp,ymu,ys2}; else % inference mode switch inf case 'infLaplace' if nargin<6 % no derivative mode f = mu; yf = y.*f; s = -yf; % product latents and labels dlp = {}; d2lp = {}; d3lp = {}; % return arguments ps = max(0,s); lp = -(ps+log(exp(-ps)+exp(s-ps))); % lp = -(log(1+exp(s))) if nargout>1 % first derivatives s = min(0,f); p = exp(s)./(exp(s)+exp(s-f)); % p = 1./(1+exp(-f)) dlp = (y+1)/2-p; % derivative of log likelihood if nargout>2 % 2nd derivative of log likelihood d2lp = -exp(2*s-f)./(exp(s)+exp(s-f)).^2; if nargout>3 % 3rd derivative of log likelihood d3lp = 2*d2lp.*(0.5-p); end end end varargout = {sum(lp),dlp,d2lp,d3lp}; else % derivative mode varargout = {[]}; % derivative w.r.t. hypers end case 'infEP' if nargin<6 % no derivative mode y = y.*ones(size(mu)); % make y a vector % likLogistic(t) \approx 1/2 + \sum_{i=1}^5 (c_i/2) erf(lam_i/sqrt(2)t) lam = sqrt(2)*[0.44 0.41 0.40 0.39 0.36]; % approx coeffs lam_i and c_i c = [1.146480988574439e+02; -1.508871030070582e+03; 2.676085036831241e+03; -1.356294962039222e+03; 7.543285642111850e+01 ]; [lZc,dlZc,d2lZc] = likErf([], y*ones(1,5), mu*lam, s2*(lam.^2), inf); lZ = log_expA_x(lZc,c); % A=lZc, B=dlZc, d=c.*lam', lZ=log(exp(A)*c) dlZ = expABz_expAx(lZc, c, dlZc, c.*lam'); % ((exp(A).*B)*d)./(exp(A)*c) % d2lZ = ((exp(A).*Z)*e)./(exp(A)*c) - dlZ.^2 where e = c.*(lam.^2)' d2lZ = expABz_expAx(lZc, c, dlZc.^2+d2lZc, c.*(lam.^2)') - dlZ.^2; % The scale mixture approximation does not capture the correct asymptotic % behavior; we have linear decay instead of quadratic decay as suggested % by the scale mixture approximation. By observing that for large values % of -f*y ln(p(y|f)) of likLogistic is linear in f with slope y, we are % able to analytically integrate the tail region; there is no contribution % to the second derivative val = abs(mu)-196/200*s2-4; % empirically determined bound at val==0 lam = 1./(1+exp(-10*val)); % interpolation weights lZtail = min(s2/2-abs(mu),-.1); % apply the same to p(y|f) = 1 - p(-y|f) dlZtail = -sign(mu); id = y.*mu>0; lZtail(id) = log(1-exp(lZtail(id))); % label and mean agree dlZtail(id) = 0; lZ = (1-lam).* lZ + lam.* lZtail; % interpolate between scale mixture .. dlZ = (1-lam).*dlZ + lam.*dlZtail; % .. and tail approximation varargout = {lZ,dlZ,d2lZ}; else % derivative mode varargout = {[]}; % deriv. wrt hyp.lik end case 'infVB' % variational lower site bound % using -log(1+exp(-s)) = s/2 -log( 2*cosh(s/2) ); % the bound has the form: b*s - s.^2/(2*ga) - h(ga)/2 with b=1/2 ga = s2; n = numel(ga); b = (y/2).*ones(n,1); db = zeros(n,1); d2b = db; h = zeros(n,1); dh = h; d2h = h; % allocate memory for return argument idlo = ga(:)<=4; h(idlo) = 2*log(2); % constant value below 4 idup = 50<=ga(:); zn = zeros(n,1); % linear behavior for large gammas h(idup) = ga(idup)/4; dh(idup) = zn(idup)+1/4; d2h(idup) = zn(idup); id = ~idlo & ~idup; % interesting zone is in between [g,dg,d2g] = inv_xcothx(1/4*ga(id)); thg = tanh(g); h(id) = 2*logcosh(g) +2*log(2) -g.*thg; g1mt2 = g.*(1-thg.^2); dh(id) = dg.*( thg - g1mt2 )/4; d2h(id) = ( g1mt2.*(2*dg.^2.*thg-d2g) + d2g.*thg )/16; varargout = {h,b,dh,db,d2h,d2b}; end end % numerically safe version of log(cosh(x)) = log(exp(x)+exp(-x))-log(2) function f = logcosh(x) f = abs(x) + log(1+exp(-2*abs(x))) - log(2); % Invert f(x) = x.*coth(x) = y, return the positive value % uses Newton's method to minimize (y-f(x))² w.r.t. x function [x,dx,d2x] = inv_xcothx(y) % f = @(x) x.*coth(x); % df = @(x) x + (1-f(x)).*coth(x); % d2f = @(x) 2*(1-f(x)).*(1-coth(x).^2); if numel(y)==0, x=[]; dx=[]; d2x=[]; return, end x = sqrt(y.^2-1); i = 1; % init ep = eps./(1+abs(x)); th=tanh(x); fx=(x+ep)./(th+ep); % function r = fx-y; % init while i==1 || (i<10 && max(abs(r))>1e-12) dfx = x + (1-fx)./(th+ep); % first derivative d2fx = 2*( 1-fx + (x-th+ep/3)./(th.*th.*th+ep) ); % second derivative x = x - r.*dfx./(dfx.^2+r.*d2fx+ep); % Newton step ep = eps./(1+abs(x)); th=tanh(x); fx=(x+ep)./(th+ep); r = fx-y; i = i+1; end % first derivative dx/dy % derivatives of inverse functions are reciprocals dx = 1./dfx; % second derivative d2x/dy2 % quotient rule and chaine rule d2x = -d2fx.*dx./dfx./dfx; % computes y = log( exp(A)*x ) in a numerically safe way by subtracting the % maximal value in each row to avoid cancelation after taking the exp function y = log_expA_x(A,x) N = size(A,2); maxA = max(A,[],2); % number of columns, max over columns y = log(exp(A-maxA*ones(1,N))*x) + maxA; % exp(A) = exp(A-max(A))*exp(max(A)) % computes y = ( (exp(A).*B)*z ) ./ ( exp(A)*x ) in a numerically safe way % The function is not general in the sense that it yields correct values for % all types of inputs. We assume that the values are close together. function y = expABz_expAx(A,x,B,z) N = size(A,2); maxA = max(A,[],2); % number of columns, max over columns A = A-maxA*ones(1,N); % subtract maximum value y = ( (exp(A).*B)*z ) ./ ( exp(A)*x ); gpml-matlab-v3.0-2010-07-23/inf/infEP.m000644 000765 000024 00000013621 11422240326 017003 0ustar00carlstaff000000 000000 function [post nlZ dnlZ] = infEP(hyp, mean, cov, lik, x, y) % Expectation Propagation approximation to the posterior Gaussian Process. % The function takes a specified covariance function (see covFunction.m) and % likelihood function (see likFunction.m), and is designed to be used with % gp.m. See also infFunctions.m. In the EP algorithm, the sites are % updated in random order, for better performance when cases are ordered % according to the targets. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch 2010-02-25. persistent last_ttau last_tnu % keep tilde parameters between calls tol = 1e-4; max_sweep = 10; min_sweep = 2; % tolerance to stop EP iterations if ~ischar(lik), lik = func2str(lik); end % make likelihood variable a string inf = 'infEP'; n = size(x,1); K = feval(cov{:}, hyp.cov, x); % evaluate the covariance matrix m = feval(mean{:}, hyp.mean, x); % evaluate the mean vector % A note on naming: variables are given short but descriptive names in % accordance with Rasmussen & Williams "GPs for Machine Learning" (2006): mu % and s2 are mean and variance, nu and tau are natural parameters. A leading t % means tilde, a subscript _ni means "not i" (for cavity parameters), or _n % for a vector of cavity parameters. % marginal likelihood for ttau = tnu = zeros(n,1); equals n*log(2) for likCum* nlZ0 = -sum(feval(lik, hyp.lik, y, m, diag(K), inf)); if any(size(last_ttau) ~= [n 1]) % find starting point for tilde parameters ttau = zeros(n,1); % initialize to zero if we have no better guess tnu = zeros(n,1); Sigma = K; % initialize Sigma and mu, the parameters of .. mu = zeros(n,1); % .. the Gaussian posterior approximation nlZ = nlZ0; else ttau = last_ttau; % try the tilde values from previous call tnu = last_tnu; [Sigma, mu, nlZ, L] = epComputeParams(K, y, ttau, tnu, lik, hyp, m, inf); if nlZ > nlZ0 % if zero is better .. ttau = zeros(n,1); % .. then initialize with zero instead tnu = zeros(n,1); Sigma = K; % initialize Sigma and mu, the parameters of .. mu = zeros(n,1); % .. the Gaussian posterior approximation nlZ = nlZ0; end end nlZ_old = Inf; sweep = 0; % converged, max. sweeps or min. sweeps? while (abs(nlZ-nlZ_old) > tol && sweep < max_sweep) || sweep2 % do we want derivatives? dnlZ = hyp; % allocate space for derivatives ssi = sqrt(ttau); V = L'\(repmat(ssi,1,n).*K); Sigma = K - V'*V; mu = Sigma*tnu; tau_n = 1./diag(Sigma)-ttau; % compute the log marginal likelihood nu_n = mu./diag(Sigma)-tnu; % vectors of cavity parameters F = alpha*alpha'-repmat(sW,1,n).*solve_chol(L,diag(sW)); % covariance hypers for j=1:length(hyp.cov) dK = feval(cov{:}, hyp.cov, x, j); dnlZ.cov(j) = -sum(sum(F.*dK))/2; end for i = 1:numel(hyp.lik) % likelihood hypers dlik = feval(lik, hyp.lik, y, nu_n./tau_n+m, 1./tau_n, inf, i); dnlZ.lik(i) = -sum(dlik); end [junk,dlZ] = feval(lik, hyp.lik, y, nu_n./tau_n+m, 1./tau_n, inf); % mean hyps for i = 1:numel(hyp.mean) dmean = feval(mean{:}, hyp.mean, x, i); dnlZ.mean(i) = -dlZ'*dmean; end end % function to compute the parameters of the Gaussian approximation, Sigma and % mu, and the negative log marginal likelihood, nlZ, from the current site % parameters, ttau and tnu. Also returns L (useful for predictions). function [Sigma mu nlZ L] = epComputeParams(K, y, ttau, tnu, lik, hyp, m, inf) n = length(y); % number of training cases ssi = sqrt(ttau); % compute Sigma and mu L = chol(eye(n)+ssi*ssi'.*K); % L'*L=B=eye(n)+sW*K*sW V = L'\(repmat(ssi,1,n).*K); Sigma = K - V'*V; mu = Sigma*tnu; tau_n = 1./diag(Sigma)-ttau; % compute the log marginal likelihood nu_n = mu./diag(Sigma)-tnu+m.*tau_n; % vectors of cavity parameters lZ = feval(lik, hyp.lik, y, nu_n./tau_n, 1./tau_n, inf); nlZ = sum(log(diag(L))) -sum(lZ) -tnu'*Sigma*tnu/2 ... -(nu_n-m.*tau_n)'*((ttau./tau_n.*(nu_n-m.*tau_n)-2*tnu)./(ttau+tau_n))/2 ... +sum(tnu.^2./(tau_n+ttau))/2-sum(log(1+ttau./tau_n))/2; gpml-matlab-v3.0-2010-07-23/inf/infVB.m000644 000765 000024 00000013463 11422240326 017012 0ustar00carlstaff000000 000000 function [post, nlZ, dnlZ] = infVB(hyp, mean, cov, lik, x, y) % Variational approximation to the posterior Gaussian process with MKL % covariance function hyperparameter optimisation. % The function takes a likelihood function (see likFunction.m), and is designed % to be used with gp.m. See also infFunctions.m. % % Minimisation of an upper bound on the negative marginal likelihood % \int N(f|0,K) p(y|f) df \ge Psi(ga,theta) with hyperparameters % theta=cov.hyp and likelihood lik(y,f) = p(y|f). % % Psi(ga,theta) = nlZ = % = (ln|K+ga|-ln|ga| + h(ga) - b'*inv(inv(K)+inv(ga)))*b)/2 % = (ln|A| + ln|K| + h(ga) - b'*inv(A)*b)/2 with A = inv(K)+inv(ga) % % The map (ga,K) |-> Psi(ga,theta) - ln|K| is jointly convex and ln|K| is % concave in its linear parameters theta. % % We optimise the convex Psi(ga,theta)-ln|K|+z'*theta/2 + B(t,theta) criterion % in the outer loop using the barrier function B(t,theta) = -sum(log(theta))/t % to enforce positivity w.r.t. theta. % The linear term z'*theta/2 is an upper bound on the concave ln|K| term. % We use an interleaved optimisation with inner loops doing Newton steps in ga % and an outer loop doing joint Newton updates in (ga,theta). Line searches are % done using derivative-free 'Brent's minimum' search. % % Copyright (c) by Hannes Nickisch 2010-07-22. maxitinner = 20; % number of inner Newton steps in ga % some less important parameters ep = 1e-8; % small constant for the ridge added to the stab. Newton direction smax = 5; Nline = 15; thr = 1e-4; % line search parameters tol = 1e-7; % tolerance for when to stop the Newton iterations inf = 'infVB'; [n,D] = size(x); K = feval(cov{:}, hyp.cov, x); % evaluate the covariance matrix m = feval(mean{:}, hyp.mean, x); % evaluate the mean vector if ~ischar(lik), lik = func2str(lik); end if (norm(m)>1e-10 || numel(hyp.mean)>0) ... && (strcmp(lik,'likErf')||strcmp(lik,'likLogistic')) error('only meanZero implemented for classification') end y = y-m; % no we have either zero mean or non-classification likelihoods if strcmp(lik,'likGauss') ga = exp(2*hyp.lik)*ones(n,1); % best ga is known for Gaussian likelihood elseif strcmp(lik,'likErf') ga = ones(n,1); % use a fixed ga for errf likelihood due to asymptotics else % INNER compute the Newton direction of ga ga = ones(n,1); % initial values itinner = 0; nlZ_new = 1e100; nlZ_old = Inf; % make sure while loop starts while nlZ_old-nlZ_new>tol && itinner0 Psi_s = @(s,ddga,ga,K,inf,hyp,lik,y) Psi(ga+s*ddga,K,inf,hyp,lik,y); [s,nlZ_new] = brentmin(0,s,Nline,thr, Psi_s,0,ddga,ga,K,inf,hyp,lik,y); ga = abs(ga + s*ddga); % update variational parameters end end [nlZ,dnlZ,d2nlZ,b] = Psi(ga,K,inf,hyp,lik,y); % upp bd on neg log marg lik W = 1./ga; sW = sqrt(W); % return the posterior parameters L = chol(eye(n)+sW*sW'.*K); % L'*L = B =eye(n)+sW*K*sW iKtil = repmat(sW,1,n).*solve_chol(L,diag(sW)); % sW*B^-1*sW=inv(K+inv(W)) alpha = b - iKtil*(K*b); post.alpha = alpha; post.sW = sW; post.L = L; if nargout>2 % do we want derivatives? dnlZ = hyp; % allocate space for derivatives for j=1:length(hyp.cov) % covariance hypers dK = feval(cov{:}, hyp.cov, x, j); if j==1, v = iKtil*(b./W); end dnlZ.cov(j) = sum(sum(iKtil.*dK))/2 - (v'*dK*v)/2; % implicit derivative = 0 end if ~strcmp(lik,'likGauss') % likelihood hypers for j=1:length(hyp.lik) dhhyp = feval(lik,hyp.lik,y,[],ga,inf,j); dnlZ.lik(j) = sum(dhhyp)/2; % implicit derivative = 0 end else % special treatment for the Gaussian case dnlZ.lik = sum(sum( (L'\eye(n)).^2 )) - exp(2*hyp.lik)*(alpha'*alpha); end for j=1:length(hyp.mean) % mean hypers dm = feval(mean{:}, hyp.mean, x, j); dnlZ.mean(j) = -alpha'*dm; % implicit derivative = 0 end end % variational lower bound along with derivatives w.r.t. ga and theta=hyp.cov % psi = (ln|K+ga|-ln|ga| + h(ga) - b'*inv(inv(K)+inv(ga)))*b)/2 % = (ln|A| + ln|K| + h(ga) - b'*inv(A)*b)/2 with A = inv(K)+inv(ga) % the code is numerically stable function [nlZ,dga,d2ga,b] = Psi(ga,K,inf,hyp,lik,y) n = size(K,1); [h,b,dh,db,d2h,d2b] = feval(lik,hyp.lik,y,[],ga,inf); W = 1./ga; sW = sqrt(W); L = chol(eye(n)+sW*sW'.*K); % sum(log(diag(L))) = log|B|/2 C = L'\(repmat(sW,1,n).*K); t = C*b; % t'*t-b'*K*b = -b'*inv(inv(K)+diag(1./ga))*b nlZ = sum(log(diag(L))) + ( sum(h) + t'*t-b'*K*b )/2; if nargout>1 % Hessian w.r.t. variational parameters gamma iKtil = repmat(sW,1,n).*solve_chol(L,diag(sW)); % sW*inv(B)*sW=inv(K+inv(W)) Khat = K-C'*C; v = Khat*b; % K-K*sW*inv(B)*sW*K=inv(inv(K)+W) dga = ( diag(iKtil)-1./ga + dh - (v./ga).^2 - 2*v.*db )/2; if nargout>2 % gradient w.r.t. variational parameters gamma w = v./ga.^2; d2ga = ( -iKtil.^2+diag(1./ga.^2) + diag(d2h) )/2 ... -Khat.*(w*(w+2*db)') -Khat.*(db*db') +diag(w.^2.*ga)-diag(v.*d2b); d2ga = (d2ga+d2ga')/2; % symmetrise end end gpml-matlab-v3.0-2010-07-23/inf/infLaplace.m000644 000765 000024 00000015215 11422240326 020041 0ustar00carlstaff000000 000000 function [post nlZ dnlZ] = infLaplace(hyp, mean, cov, lik, x, y) % Laplace approximation to the posterior Gaussian process. % The function takes a specified covariance function (see covFunction.m) and % likelihood function (see likFunction.m), and is designed to be used with % gp.m. See also infFunctions.m. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch 2010-02-23. persistent last_alpha % copy of the last alpha tol = 1e-6; % tolerance for when to stop the Newton iterations smax = 2; Nline = 10; thr = 1e-4; % line search parameters maxit = 20; % max number of Newton steps in f if ~ischar(lik), lik = func2str(lik); end % make likelihood variable a string inf = 'infLaplace'; n = size(x,1); K = feval(cov{:}, hyp.cov, x); % evaluate the covariance matrix m = feval(mean{:}, hyp.mean, x); % evaluate the mean vector Psi_old = Inf; % make sure while loop starts by the largest old objective val if any(size(last_alpha)~=[n,1]) % find a good starting point for alpha and f alpha = zeros(n,1); f = K*alpha+m; % start at mean if sizes do not match [lp,dlp,d2lp] = feval(lik,hyp.lik,y,f,[],inf); W = -d2lp; Psi_new = -lp; else alpha = last_alpha; f = K*alpha+m; % try last one [lp,dlp,d2lp] = feval(lik,hyp.lik,y,f,[],inf); W = -d2lp; Psi_new = alpha'*(f-m)/2 - lp; % objective for last alpha Psi_def = -feval(lik,hyp.lik,y,m,[],inf); % objective for default init f==m if Psi_def < Psi_new % if default is better, we use it alpha = zeros(n,1); f = K*alpha+m; [lp,dlp,d2lp] = feval(lik,hyp.lik,y,f,[],inf); W = -d2lp; Psi_new = -lp; end end isWneg = any(W<0); % flag indicating whether we found negative values of W it = 0; % this happens for the Student's t likelihood while Psi_old - Psi_new > tol && it2 % do we want derivatives? dnlZ = hyp; % allocate space for derivatives if isWneg % switch between Cholesky and LU decomposition mode Z = -post.L; % inv(K+inv(W)) g = sum(iA.*K,2)/2; % deriv. of ln|B| wrt W; g = diag(inv(inv(K)+diag(W)))/2 else Z = repmat(sW,1,n).*solve_chol(post.L,diag(sW)); %sW*inv(B)*sW=inv(K+inv(W)) C = post.L'\(repmat(sW,1,n).*K); % deriv. of ln|B| wrt W g = (diag(K)-sum(C.^2,1)')/2; % g = diag(inv(inv(K)+W))/2 end dfhat = g.*d3lp; % deriv. of nlZ wrt. fhat: dfhat=diag(inv(inv(K)+W)).*d3lp/2 for j=1:length(hyp.cov) % covariance hypers dK = feval(cov{:}, hyp.cov, x, j); dnlZ.cov(j) = sum(sum(Z.*dK))/2 - alpha'*dK*alpha/2; % explicit part b = dK*dlp; % b-K*(Z*b) = inv(eye(n)+K*diag(W))*b dnlZ.cov(j) = dnlZ.cov(j) - dfhat'*( b-K*(Z*b) ); % implicit part end for j=1:length(hyp.lik) % likelihood hypers [lp_dhyp,d2lp_dhyp] = feval(lik,hyp.lik,y,f,[],inf,j); dnlZ.lik(j) = -g'*d2lp_dhyp - sum(lp_dhyp); % explicit part, implicit is 0 end for j=1:length(hyp.mean) % mean hypers dm = feval(mean{:}, hyp.mean, x, j); dnlZ.mean(j) = -alpha'*dm; % explicit part dnlZ.mean(j) = dnlZ.mean(j) - dfhat'*(dm-K*(Z*dm)); % implicit part end end % criterion Psi at alpha + s*dalpha for line search function [Psi,alpha,f,dlp,W] = Psi_line(s,dalpha,alpha,hyp,K,m,lik,y,inf) alpha = alpha + s*dalpha; f = K*alpha+m; [lp,dlp,d2lp] = feval(lik,hyp.lik,y,f,[],inf); W = -d2lp; Psi = alpha'*(f-m)/2 - lp; % Compute the log determinant ldA and the inverse iA of a square nxn matrix % A = eye(n) + K*diag(w) from its LU decomposition; for negative definite A, we % return ldA = Inf. We also return mwiA = -diag(w)*inv(A). function [ldA,iA,mwiA] = logdetA(K,w) [m,n] = size(K); if m~=n, error('K has to be nxn'), end A = eye(n)+K.*repmat(w',n,1); [L,U,P] = lu(A); u = diag(U); % compute LU decomposition, A = P'*L*U signU = prod(sign(u)); % sign of U detP = 1; % compute sign (and det) of the premutation matrix P p = P*(1:n)'; for i=1:n if i~=p(i) detP = -detP; j = find(p==i); p([i,j]) = p([j,i]); % swap entries end end if signU~=detP % log becomes complex for negative values, encoded by infinity ldA = Inf; else % det(L) = 1 and U triangular => det(A) = det(P)*prod(diag(U)) ldA = sum(log(abs(u))); end if nargout>1, iA = inv(U)*inv(L)*P; end % return the inverse, as well if nargout>2, mwiA = -repmat(w,1,n).*iA; end gpml-matlab-v3.0-2010-07-23/inf/infExact.m000644 000765 000024 00000003453 11422313220 017537 0ustar00carlstaff000000 000000 function [post nlZ dnlZ] = infExact(hyp, mean, cov, lik, x, y) % Exact inference for a GP with Gaussian likelihood. Compute a parametrization % of the posterior, the negative log marginal likelihood and its derivatives % w.r.t. the hyperparameters. See also "help infMethods". % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-23 likstr = lik; if ~ischar(lik), likstr = func2str(lik); end if ~strcmp(likstr,'likGauss') % NOTE: no explicit call to likGauss error('Exact inference only possible with Gaussian likelihood'); end [n, D] = size(x); K = feval(cov{:}, hyp.cov, x); % evaluate covariance matrix m = feval(mean{:}, hyp.mean, x); % evaluate mean vector sn2 = exp(2*hyp.lik); % noise variance of likGauss L = chol(K/sn2+eye(n)); % Cholesky factor of covariance with noise alpha = solve_chol(L,y-m)/sn2; post.alpha = alpha; % return the posterior parameters post.sW = ones(n,1)/sqrt(sn2); % sqrt of noise precision vector post.L = L; % L = chol(eye(n)+sW*sW'.*K) if nargout>1 % do we want the marginal likelihood nlZ = (y-m)'*alpha/2 + sum(log(diag(L))) + n*log(2*pi*sn2)/2; % -log marg lik if nargout>2 % do we want derivatives? dnlZ = hyp; % allocate space for derivatives Q = L\(L'\eye(n))/sn2 - alpha*alpha'; % precompute for convenience for i = 1:numel(hyp.cov) dnlZ.cov(i) = sum(sum(Q.*feval(cov{:}, hyp.cov, x, i)))/2; end dnlZ.lik = sn2*trace(Q); for i = 1:numel(hyp.mean), dnlZ.mean(i) = -feval(mean{:}, hyp.mean, x, i)'*alpha; end end end gpml-matlab-v3.0-2010-07-23/util/binaryEPGP.m000644 000765 000024 00000005753 11422240326 020152 0ustar00carlstaff000000 000000 function varargout = binaryEPGP(hyper, covfunc, x, y, xstar) % binaryEPGP - The Expectation Propagation approximation for binary Gaussian % process classification. Two modes are possible: training or testing: if no % test cases are supplied, then the approximate negative log marginal % likelihood and its partial derivatives wrt the hyperparameters is computed; % this mode is used to fit the hyperparameters. If test cases are given, then % the test set predictive probabilities are returned. The program is flexible % in allowing a multitude of covariance functions. % % usage: [nlZ, dnlZ ] = binaryEPGP(hyper, covfunc, x, y); % or: [p, mu, s2, nlZ] = binaryEPGP(hyper, covfunc, x, y, xstar); % % where: % % hyper is a (column) vector of hyperparameters % covfunc is the name of the covariance function (see below) % lik is the name of the likelihood function (see below) % x is a n by D matrix of training inputs % y is a (column) vector (of size n) of binary +1/-1 targets % xstar is a nn by D matrix of test inputs % nlZ is the returned value of the negative log marginal likelihood % dnlZ is a (column) vector of partial derivatives of the negative % log marginal likelihood wrt each log hyperparameter % p is a (column) vector (of length nn) of predictive probabilities % mu is a (column) vector (of length nn) of predictive latent means % s2 is a (column) vector (of length nn) of predictive latent variances % % The length of the vector of hyperparameters depends on the covariance % function, as specified by the "covfunc" input to the function, specifying the % name of a covariance function. A number of different covariance function are % implemented, and it is not difficult to add new ones. See "help covFunctions" % for the details % % The function can conveniently be used with the "minimize" function to train % a Gaussian process, eg: % % [hyper, fX, i] = minimize(hyper, 'binaryEPGP', length, 'covSEiso', % 'logistic', x, y); % % Copyright (c) 2004, 2005, 2006, 2007 Carl Edward Rasmussen, 2010-07-21. if nargin<4 || nargin>5 disp('Usage: [nlZ, dnlZ ] = binaryEPGP(hyper, covfunc, x, y);') disp(' or: [p, mu, s2, nlZ] = binaryEPGP(hyper, covfunc, x, y, xstar);') return end % Note, this function is just a wrapper provided for backward compatibility, % the functionality is now provided by the more general gp function. lik = @likErf; inf = @infEP; mean = @meanZero; varargout = cell(nargout, 1); % allocate the right number of output arguments hyp.cov = hyper; if nargin==4 [varargout{:}] = gp(hyp,inf,mean,covfunc,lik,x,y); if nargout>1, varargout{2} = varargout{2}.cov; end else [ymu,ys2,fmu,fs2] = gp(hyp,inf,mean,covfunc,lik,x,y,xstar); if nargout>0, varargout{1} = (1+ymu)/2; end if nargout>1, varargout{2} = fmu; end if nargout>2, varargout{3} = fs2; end if nargout>3, varargout{4} = gp(hyp,inf,mean,covfunc,lik,x,y); end % nlZ end gpml-matlab-v3.0-2010-07-23/util/binaryLaplaceGP.m000644 000765 000024 00000007030 11422240326 021175 0ustar00carlstaff000000 000000 function varargout = binaryLaplaceGP(hyper, covfunc, lik, x, y, xstar) % binaryLaplaceGP - Laplace's approximation for binary Gaussian process % classification. Two modes are possible: training or testing: if no test % cases are supplied, then the approximate negative log marginal likelihood % and its partial derivatives wrt the hyperparameters is computed; this mode is % used to fit the hyperparameters. If test cases are given, then the test set % predictive probabilities are returned. The program is flexible in allowing % several different likelihood functions and a multitude of covariance % functions. % % usage: [nlZ, dnlZ ] = binaryLaplaceGP(hyper, covfunc, lik, x, y); % or: [p, mu, s2, nlZ] = binaryLaplaceGP(hyper, covfunc, lik, x, y, xstar); % % where: % % hyper is a (column) vector of hyperparameters % covfunc is the name of the covariance function (see below) % lik is the name of the likelihood function (see below) % x is a n by D matrix of training inputs % y is a (column) vector (of size n) of binary +1/-1 targets % xstar is a nn by D matrix of test inputs % nlZ is the returned value of the negative log marginal likelihood % dnlZ is a (column) vector of partial derivatives of the negative % log marginal likelihood wrt each log hyperparameter % p is a (column) vector (of length nn) of predictive probabilities % mu is a (column) vector (of length nn) of predictive latent means % s2 is a (column) vector (of length nn) of predictive latent variances % % The length of the vector of log hyperparameters depends on the covariance % function, as specified by the "covfunc" input to the function, specifying the % name of a covariance function. A number of different covariance function are % implemented, and it is not difficult to add new ones. See "help covFunctions" % for the details. % % The shape of the likelihood function is given by the "lik" input to the % function, specifying the name of the likelihood function. The two implemented % likelihood functions are: % % logistic the logistic function: 1/(1+exp(-x)) % cumGauss the cumulative Gaussian (error function) % % The function can conveniently be used with the "minimize" function to train % a Gaussian process, eg: % % [hyper, fX, i] = minimize(hyper, 'binaryLaplaceGP', length, 'covSEiso', % 'logistic', x, y); % % Copyright (c) 2004, 2005, 2006, 2007 Carl Edward Rasmussen, 2010-07-21. if nargin<5 || nargin>6 disp('Usage: [nlZ, dnlZ ] = binaryLaplaceGP(hyper, covfunc, lik, x, y);') disp(' or: [p, mu, s2, nlZ] = binaryLaplaceGP(hyper, covfunc, lik, x, y, xstar);') return end % Note, this function is just a wrapper provided for backward compatibility, % the functionality is now provided by the more general gp function. if strcmp(lik,'logistic') lik = @likLogistic; elseif strcmpi(lik,'cumgauss') lik = @likErf; else error('Allowable likelihoods: logistic and cumGauss.'); end inf = @infLaplace; mean = @meanZero; varargout = cell(nargout, 1); % allocate the right number of output arguments hyp.cov = hyper; if nargin==5 [varargout{:}] = gp(hyp,inf,mean,covfunc,lik,x,y); if nargout>1, varargout{2} = varargout{2}.cov; end else [ymu,ys2,fmu,fs2] = gp(hyp,inf,mean,covfunc,lik,x,y,xstar); if nargout>0, varargout{1} = (1+ymu)/2; end if nargout>1, varargout{2} = fmu; end if nargout>2, varargout{3} = fs2; end if nargout>3, varargout{4} = gp(hyp,inf,mean,covfunc,lik,x,y); end % nlZ end gpml-matlab-v3.0-2010-07-23/util/solve_chol.c000644 000765 000024 00000002355 11422240326 020330 0ustar00carlstaff000000 000000 /* solve_chol - solve a linear system A*X = B using the cholesky factorization of A (where A is square, symmetric and positive definite. Copyright (c) 2004 Carl Edward Rasmussen. 2004-10-19. */ #include "mex.h" #include #include extern int dpotrs_(char *, int *, int *, double *, int *, double *, int *, int *); void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) { double *C; int n, m, q; if (nrhs != 2 || nlhs > 1) /* check the input */ mexErrMsgTxt("Usage: X = solve_chol(R, B)"); n = mxGetN(prhs[0]); if (n != mxGetM(prhs[0])) mexErrMsgTxt("Error: First argument matrix must be square"); if (n != mxGetM(prhs[1])) mexErrMsgTxt("Error: First and second argument matrices must have same number of rows"); m = mxGetN(prhs[1]); plhs[0] = mxCreateDoubleMatrix(n, m, mxREAL); /* allocate space for output */ C = mxGetPr(plhs[0]); if (n==0) return; /* if argument was empty matrix, do no more */ memcpy( C, mxGetPr(prhs[1]), n*m*sizeof(double) );/* copy argument matrix */ dpotrs_("U", &n, &m, mxGetPr(prhs[0]), &n, C, &n, &q); /* solve system */ if (q > 0) mexErrMsgTxt("Error: illegal input to solve_chol"); } gpml-matlab-v3.0-2010-07-23/util/solve_chol.m000644 000765 000024 00000001742 11422240326 020341 0ustar00carlstaff000000 000000 % solve_chol - solve linear equations from the Cholesky factorization. % Solve A*X = B for X, where A is square, symmetric, positive definite. The % input to the function is R the Cholesky decomposition of A and the matrix B. % Example: X = solve_chol(chol(A),B); % % NOTE: The program code is written in the C language for efficiency and is % contained in the file solve_chol.c, and should be compiled using matlabs mex % facility. However, this file also contains a (less efficient) matlab % implementation, supplied only as a help to people unfamiliar with mex. If % the C code has been properly compiled and is avaiable, it automatically % takes precendence over the matlab code in this file. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch 2010-01-09. function x = solve_chol(A, B) if nargin ~= 2 || nargout > 1 error('Wrong number of arguments.'); end if size(A,1) ~= size(A,2) || size(A,1) ~= size(B,1) error('Wrong sizes of matrix arguments.'); end x = A\(A'\B); gpml-matlab-v3.0-2010-07-23/util/lbfgsb/000755 000765 000024 00000000000 13330425172 017265 5ustar00carlstaff000000 000000 gpml-matlab-v3.0-2010-07-23/util/minimize_lbfgsb_gradfun.m000644 000765 000024 00000004526 11422240326 023055 0ustar00carlstaff000000 000000 function G = minimize_lbfgsb_gradfun(X,varargin) % extract input arguments varargin = varargin{1}; strctX = varargin{2}; f = varargin{1}; % global variables serve as communication interface between calls global minimize_lbfgsb_iteration_number global minimize_lbfgsb_objective global minimize_lbfgsb_gradient global minimize_lbfgsb_X if norm(unwrap(X)-unwrap(minimize_lbfgsb_X))>1e-10 [y,G] = feval(f,rewrap(strctX,X),varargin{3:end}); else y = minimize_lbfgsb_objective(minimize_lbfgsb_iteration_number); G = minimize_lbfgsb_gradient; end % memorise gradient and position minimize_lbfgsb_gradient = G; minimize_lbfgsb_X = X; G = unwrap(G); % Extract the numerical values from "s" into the column vector "v". The % variable "s" can be of any type, including struct and cell array. % Non-numerical elements are ignored. See also the reverse rewrap.m. function v = unwrap(s) v = []; if isnumeric(s) v = s(:); % numeric values are recast to column vector elseif isstruct(s) v = unwrap(struct2cell(orderfields(s)));% alphabetize, conv to cell, recurse elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially v = [v; unwrap(s{i})]; end end % other types are ignored % Map the numerical elements in the vector "v" onto the variables "s" which can % be of any type. The number of numerical elements must match; on exit "v" % should be empty. Non-numerical entries are just copied. See also unwrap.m. function [s v] = rewrap(s, v) if isnumeric(s) if numel(v) < numel(s) error('The vector for conversion contains too few elements') end s = reshape(v(1:numel(s)), size(s)); % numeric values are reshaped v = v(numel(s)+1:end); % remaining arguments passed on elseif isstruct(s) [s p] = orderfields(s); p(p) = 1:numel(p); % alphabetize, store ordering [t v] = rewrap(struct2cell(s), v); % convert to cell, recurse s = orderfields(cell2struct(t,fieldnames(s),1),p); % conv to struct, reorder elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially [s{i} v] = rewrap(s{i}, v); end end % other types are not processed gpml-matlab-v3.0-2010-07-23/util/minimize_lbfgsb.m000644 000765 000024 00000010574 11422240326 021347 0ustar00carlstaff000000 000000 function [X, fX, i] = minimize_lbfgsb(X, f, length, varargin) % Minimize a differentiable multivariate function using quasi Newton. % % Usage: [X, fX, i] = minimize_lbfgsb(X, f, length, P1, P2, P3, ... ) % % X initial guess; may be of any type, including struct and cell array % f the name or pointer to the function to be minimized. The function % f must return two arguments, the value of the function, and it's % partial derivatives wrt the elements of X. The partial derivative % must have the same type as X. % length length of the run; if it is positive, it gives the maximum number of % line searches, if negative its absolute gives the maximum allowed % number of function evaluations. Optionally, length can have a second % component, which will indicate the reduction in function value to be % expected in the first line-search (defaults to 1.0). % P1, P2 ... parameters are passed to the function f. % % X the returned solution % fX vector of function values indicating progress made % i number of iterations (line searches or function evaluations, % depending on the sign of "length") used at termination. % % The function returns when either its length is up, or if no further progress % can be made (ie, we are at a (local) minimum, or so close that due to % numerical problems, we cannot get any closer). NOTE: If the function % terminates within a few iterations, it could be an indication that the % function values and derivatives are not consistent (ie, there may be a bug in % the implementation of your "f" function). % % Copyright (C) 2010 by Hannes Nickisch, 2010-02-05 % global variables serve as communication interface between calls global minimize_lbfgsb_iteration_number global minimize_lbfgsb_objective global minimize_lbfgsb_gradient global minimize_lbfgsb_X % init global variables minimize_lbfgsb_iteration_number = 0; minimize_lbfgsb_objective = Inf; minimize_lbfgsb_gradient = 0*unwrap(X); minimize_lbfgsb_X = X; X0 = X; lb = -Inf*ones(size(unwrap(X0))); ub = Inf*ones(size(unwrap(X0))); maxiter = abs(length); % max number of iterations % no callback routine used so far % m is the number of saved vectors used to estimate the Hessian % factr is the precision 1e-12 X = lbfgsb( unwrap(X0), lb, ub, 'minimize_lbfgsb_objfun', ... 'minimize_lbfgsb_gradfun', ... {f,X0,varargin{:}}, [], ... 'maxiter',maxiter, 'm',4, 'factr',1e-12, 'pgtol',1e-5); i = minimize_lbfgsb_iteration_number; fX = minimize_lbfgsb_objective; X = rewrap(X0,X); % clear global variables clear minimize_lbfgsb_iteration_number clear minimize_lbfgsb_objective clear minimize_lbfgsb_gradient clear minimize_lbfgsb_X % Extract the numerical values from "s" into the column vector "v". The % variable "s" can be of any type, including struct and cell array. % Non-numerical elements are ignored. See also the reverse rewrap.m. function v = unwrap(s) v = []; if isnumeric(s) v = s(:); % numeric values are recast to column vector elseif isstruct(s) v = unwrap(struct2cell(orderfields(s)));% alphabetize, conv to cell, recurse elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially v = [v; unwrap(s{i})]; end end % other types are ignored % Map the numerical elements in the vector "v" onto the variables "s" which can % be of any type. The number of numerical elements must match; on exit "v" % should be empty. Non-numerical entries are just copied. See also unwrap.m. function [s v] = rewrap(s, v) if isnumeric(s) if numel(v) < numel(s) error('The vector for conversion contains too few elements') end s = reshape(v(1:numel(s)), size(s)); % numeric values are reshaped v = v(numel(s)+1:end); % remaining arguments passed on elseif isstruct(s) [s p] = orderfields(s); p(p) = 1:numel(p); % alphabetize, store ordering [t v] = rewrap(struct2cell(s), v); % convert to cell, recurse s = orderfields(cell2struct(t,fieldnames(s),1),p); % conv to struct, reorder elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially [s{i} v] = rewrap(s{i}, v); end end % other types are not processed gpml-matlab-v3.0-2010-07-23/util/gpr.m000644 000765 000024 00000004353 11422240326 016775 0ustar00carlstaff000000 000000 function varargout = gpr(hyper, cov, x, y, xs) % gpr - Gaussian process regression, with a named covariance function. Two % modes are possible: training and prediction: if no test data are given, the % function returns minus the log likelihood and its partial derivatives with % respect to the hyperparameters; this mode is used to fit the hyperparameters. % If test data are given, then (marginal) Gaussian predictions are computed, % whose mean and variance are returned. Note that in cases where the covariance % function has noise contributions, the variance returned in s2 is for noisy % test targets; if you want the variance of the noise-free latent function, you % must substract the noise variance. % % usage: [nlZ dnlZ] = gpr(hyp, cov, x, y) % or: [mu s2] = gpr(hyp, cov, x, y, xs) % % where: % % hyp is a (column) vector of log hyperparameters % cov is the covariance function % x is a n by D matrix of training inputs % y is a (column) vector (of size n) of targets % xs is a ns by D matrix of test inputs % nlZ is the returned value of the negative log marginal likelihood % dnlZ is a (column) vector of partial derivatives of the negative % log marginal likelihood wrt each log hyperparameter % mu is a (column) vector (of size nn) of prediced means % s2 is a (column) vector (of size nn) of predicted variances % % For more help on covariance functions, see "help covFunctions". % % Copyright (c) 2010 Carl Edward Rasmussen and Hannes Nickisch 2010-06-18. err = 'we need cov = {''covSum'', {''covNoise'', .., covP} }; to map to gp.m'; if ~strcmp(cov{1},'covSum'), error(err), end if ~strcmp(cov{2}{1},'covNoise'), error(err), end cov{2} = cov{2}(2:end); hyp.lik = hyper(1); hyp.cov = hyper(2:end); % Note, this function is just a wrapper provided for backward compatibility, % the functionality is now provided by the more general gp function. lik = @likGauss; inf = @infExact; mean = @meanZero; varargout = cell(nargout, 1); % allocate the right number of output arguments if nargin==4 [varargout{:}] = gp(hyp,inf,mean,cov,lik,x,y); if nargout>1, varargout{2} = [varargout{2}.lik; varargout{2}.cov]; end else [varargout{:}] = gp(hyp,inf,mean,cov,lik,x,y,xs); end gpml-matlab-v3.0-2010-07-23/util/sq_dist.c000644 000765 000024 00000003613 11422240326 017637 0ustar00carlstaff000000 000000 /* sq_dist - a mex function to compute a matrix of all pairwise squared distances between two sets of vectors, stored in the columns of the two matrices that are arguments to the function. The length of the vectors must agree. If only a single argument is given, the missing argument is taken to be identical to the first. If an optional third matrix argument Q is given, it must be of the same size as the output, but in this case a vector of the traces of the product of Q and the coordinatewise squared distances is returned. Copyright (c) 2003, 2004 Carl Edward Rasmussen. 2003-04-22. */ #include "mex.h" #include void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) { double *a, *b, *C, *Q, z, t; int D, n, m, i, j, k; if (nrhs < 1 || nrhs > 3 || nlhs > 1) mexErrMsgTxt("Usage: C = sq_dist(a,b)\n or: C = sq_dist(a)\n or: c = sq_dist(a,b,Q)\nwhere the b matrix may be empty."); a = mxGetPr(prhs[0]); m = mxGetN(prhs[0]); D = mxGetM(prhs[0]); if (nrhs == 1 || mxIsEmpty(prhs[1])) { b = a; n = m; } else { b = mxGetPr(prhs[1]); n = mxGetN(prhs[1]); if (D != mxGetM(prhs[1])) mexErrMsgTxt("Error: column lengths must agree"); } if (nrhs < 3) { plhs[0] = mxCreateDoubleMatrix(m, n, mxREAL); C = mxGetPr(plhs[0]); for (i=0; i3 || nargout>1 error('Wrong number of arguments.'); end if nargin==1 || isempty(b) % input arguments are taken to be b = a; % identical if b is missing or empty end [D, n] = size(a); [d, m] = size(b); if d ~= D error('Error: column lengths must agree.'); end if nargin<3 C = zeros(n,m); for d = 1:D C = C + (repmat(b(d,:), n, 1) - repmat(a(d,:)', 1, m)).^2; end % C = repmat(sum(a.*a)',1,m)+repmat(sum(b.*b),n,1)-2*a'*b could be used to % replace the 3 lines above; it would be faster, but numerically less stable. else if all([n,m]==size(Q)) C = zeros(D,1); for d = 1:D C(d) = sum(sum((repmat(b(d,:), n, 1) - repmat(a(d,:)', 1, m)).^2.*Q)); end else error('Third argument has wrong size.'); end end gpml-matlab-v3.0-2010-07-23/util/make.m000644 000765 000024 00000005624 11422240326 017124 0ustar00carlstaff000000 000000 % For compilation under Linux/Octave you need to install the package % octavex.x-headers. % Under Windows/Octave, compilation works out of the box. % % There is a big variety of platforms running Matlab around i.e. % 'SUN4','SOL2','HP700','SGI','SGI64','IBM_RS','ALPHA', % 'LNX86','GLNX86','GLNXA64', % 'MAC', 'MACI' % the following script is only tested for the most common ones (Mac,Linux,Win) % but should work with minor modifications on your architecture. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch 2010-07-18. OCTAVE = exist('OCTAVE_VERSION') ~= 0; % check if we run Matlab or Octave %% 0) change directory me = mfilename; % what is my filename mydir = which(me); mydir = mydir(1:end-2-numel(me)); % where am I located cd(mydir) if OCTAVE %% 1) compile sq_dist mkoctfile --mex sq_dist.c delete sq_dist.o %% 2) compile solve_chol mkoctfile --mex solve_chol.c delete solve_chol.o else % MATLAB %% 1) compile sq_dist mex -O sq_dist.c %% 2) compile solve_chol if ispc % Windows libext = 'lib'; % remove the trailing underscore after the Lapack function "dpotrs" % we do that in a clumsy way since there is no sed/awk under Windows fid = fopen('solve_chol.c','r'); fid_win = fopen('solve_chol2XXXX.c','w'); while 1 tline = fgetl(fid); if ~ischar(tline), break, end id = strfind(tline,'dpotrs'); if numel(id) tline = [tline(1:id-1),'dpotrs',tline(id+7:end)]; end fprintf(fid_win,[tline,'\n']); end fclose(fid_win); fclose(fid); if numel(strfind(computer,'64')) % 64 bit ospath = 'extern/lib/win64/microsoft'; else % 32 bit ospath = 'extern/lib/win32/lcc'; end % compile comp_cmd = 'mex -O solve_chol2XXXX.c -output solve_chol'; eval([comp_cmd,' ''',matlabroot,'/',ospath,'/libmwlapack.',libext,'''']) delete solve_chol2XXXX* return end comp_cmd = 'mex -O solve_chol.c'; if isunix if ismac % Mac libext = 'dylib'; if numel(strfind(lower(computer),'maci')) % Intel ospath = 'bin/maci'; else ospath = 'bin/mac'; end else libext = 'so'; if numel(strfind(lower(computer),'sol')) % Solaris ospath = 'bin/sol64'; else % Linux if numel(strfind(computer,'64')) ospath = 'bin/glnxa64'; % 32 bit else % 64 bit ospath = 'bin/glnx86'; end end end end eval([comp_cmd,' ',matlabroot,'/',ospath,'/libmwlapack.',libext]) end gpml-matlab-v3.0-2010-07-23/util/unwrap.m000644 000765 000024 00000001213 11422240326 017511 0ustar00carlstaff000000 000000 % Extract the numerical values from "s" into the column vector "v". The % variable "s" can be of any type, including struct and cell array. % Non-numerical elements are ignored. See also the reverse rewrap.m. function v = unwrap(s) v = []; if isnumeric(s) v = s(:); % numeric values are recast to column vector elseif isstruct(s) v = unwrap(struct2cell(orderfields(s))); % alphabetize, conv to cell, recurse elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially v = [v; unwrap(s{i})]; end end % other types are ignored gpml-matlab-v3.0-2010-07-23/util/brentmin.m000644 000765 000024 00000007574 11422240326 020033 0ustar00carlstaff000000 000000 %% BRENTMIN: Brent's minimization method in one dimension function [xmin,fmin,funccount,varargout] = ... brentmin(xlow,xupp,Nitmax,tol,f,nout,varargin) % code taken from % § 10.2 Parabolic Interpolation and Brent's Method in One Dimension % Press, Teukolsky, Vetterling & Flannery % Numerical Recipes in C, Cambridge University Press, 2002 % % [xmin,fmin,funccout,varargout] = BRENTMIN(xlow,xupp,Nit,tol,f,nout,varargin) % Given a function f, and given a search interval this routine isolates % the minimum of fractional precision of about tol using Brent's method. % % INPUT % ----- % xlow,xupp: search interval such that xlow<=xmin<=xupp % Nitmax: maximum number of function evaluations made by the routine % tol: fractional precision % f: [y,varargout{:}] = f(x,varargin{:}) is the function % nout: no. of outputs of f (in varargout) in addition to the y value % % OUTPUT % ------ % fmin: minimal function value % xmin: corresponding abscissa-value % funccount: number of function evaluations made % varargout: additional outputs of f at optimum % % Copyright (c) by Hannes Nickisch 2010-01-10. if nargin<6, nout = 0; end varargout = cell(nout,1); % tolerance is no smaller than machine's floating point precision tol = max(tol,eps); % Evaluate endpoints fa = f(xlow,varargin{:}); fb = f(xupp,varargin{:}); funccount = 2; % number of function evaluations % Compute the start point seps = sqrt(eps); c = 0.5*(3.0 - sqrt(5.0));% golden ratio a = xlow; b = xupp; v = a + c*(b-a); w = v; xf = v; d = 0.0; e = 0.0; x = xf; [fx,varargout{:}] = f(x,varargin{:}); funccount = funccount + 1; fv = fx; fw = fx; xm = 0.5*(a+b); tol1 = seps*abs(xf) + tol/3.0; tol2 = 2.0*tol1; % Main loop while ( abs(xf-xm) > (tol2 - 0.5*(b-a)) ) gs = 1; % Is a parabolic fit possible if abs(e) > tol1 % Yes, so fit parabola gs = 0; r = (xf-w)*(fx-fv); q = (xf-v)*(fx-fw); p = (xf-v)*q-(xf-w)*r; q = 2.0*(q-r); if q > 0.0, p = -p; end q = abs(q); r = e; e = d; % Is the parabola acceptable if ( (abs(p)q*(a-xf)) && (p= xm, e = a-xf; else e = b-xf; end d = c*e; end % The function must not be evaluated too close to xf si = sign(d) + (d == 0); x = xf + si * max( abs(d), tol1 ); [fu,varargout{:}] = f(x,varargin{:}); funccount = funccount + 1; % Update a, b, v, w, x, xm, tol1, tol2 if fu <= fx if x >= xf, a = xf; else b = xf; end v = w; fv = fw; w = xf; fw = fx; xf = x; fx = fu; else % fu > fx if x < xf, a = x; else b = x; end if ( (fu <= fw) || (w == xf) ) v = w; fv = fw; w = x; fw = fu; elseif ( (fu <= fv) || (v == xf) || (v == w) ) v = x; fv = fu; end end xm = 0.5*(a+b); tol1 = seps*abs(xf) + tol/3.0; tol2 = 2.0*tol1; if funccount >= Nitmax % typically we should not get here % warning(sprintf(['Maximum number of iterations (%d) exceeded:', ... % 'precision is not guaranteed'],Nitmax)) % fprintf('[%1.3f,%1.3f,%1.3f]\n',xlow,xf,xupp) break end end % while % check that endpoints are less than the minimum found if ( (fa < fx) && (fa <= fb) ) xf = xlow; fx = fa; elseif fb < fx xf = xupp; fx = fb; end fmin = fx; xmin = xf; gpml-matlab-v3.0-2010-07-23/util/minimize.m000644 000765 000024 00000025676 11422240326 020041 0ustar00carlstaff000000 000000 function [X, fX, i] = minimize(X, f, length, varargin) % Minimize a differentiable multivariate function using conjugate gradients. % % Usage: [X, fX, i] = minimize(X, f, length, P1, P2, P3, ... ) % % X initial guess; may be of any type, including struct and cell array % f the name or pointer to the function to be minimized. The function % f must return two arguments, the value of the function, and it's % partial derivatives wrt the elements of X. The partial derivative % must have the same type as X. % length length of the run; if it is positive, it gives the maximum number of % line searches, if negative its absolute gives the maximum allowed % number of function evaluations. Optionally, length can have a second % component, which will indicate the reduction in function value to be % expected in the first line-search (defaults to 1.0). % P1, P2, ... parameters are passed to the function f. % % X the returned solution % fX vector of function values indicating progress made % i number of iterations (line searches or function evaluations, % depending on the sign of "length") used at termination. % % The function returns when either its length is up, or if no further progress % can be made (ie, we are at a (local) minimum, or so close that due to % numerical problems, we cannot get any closer). NOTE: If the function % terminates within a few iterations, it could be an indication that the % function values and derivatives are not consistent (ie, there may be a bug in % the implementation of your "f" function). % % The Polack-Ribiere flavour of conjugate gradients is used to compute search % directions, and a line search using quadratic and cubic polynomial % approximations and the Wolfe-Powell stopping criteria is used together with % the slope ratio method for guessing initial step sizes. Additionally a bunch % of checks are made to make sure that exploration is taking place and that % extrapolation will not be unboundedly large. % % See also: checkgrad % % Copyright (C) 2001 - 2010 by Carl Edward Rasmussen, 2010-01-03 INT = 0.1; % don't reevaluate within 0.1 of the limit of the current bracket EXT = 3.0; % extrapolate maximum 3 times the current step-size MAX = 20; % max 20 function evaluations per line search RATIO = 10; % maximum allowed slope ratio SIG = 0.1; RHO = SIG/2; % SIG and RHO are the constants controlling the Wolfe- % Powell conditions. SIG is the maximum allowed absolute ratio between % previous and new slopes (derivatives in the search direction), thus setting % SIG to low (positive) values forces higher precision in the line-searches. % RHO is the minimum allowed fraction of the expected (from the slope at the % initial point in the linesearch). Constants must satisfy 0 < RHO < SIG < 1. % Tuning of SIG (depending on the nature of the function to be optimized) may % speed up the minimization; it is probably not worth playing much with RHO. % The code falls naturally into 3 parts, after the initial line search is % started in the direction of steepest descent. 1) we first enter a while loop % which uses point 1 (p1) and (p2) to compute an extrapolation (p3), until we % have extrapolated far enough (Wolfe-Powell conditions). 2) if necessary, we % enter the second loop which takes p2, p3 and p4 chooses the subinterval % containing a (local) minimum, and interpolates it, unil an acceptable point % is found (Wolfe-Powell conditions). Note, that points are always maintained % in order p0 <= p1 <= p2 < p3 < p4. 3) compute a new search direction using % conjugate gradients (Polack-Ribiere flavour), or revert to steepest if there % was a problem in the previous line-search. Return the best value so far, if % two consecutive line-searches fail, or whenever we run out of function % evaluations or line-searches. During extrapolation, the "f" function may fail % either with an error or returning Nan or Inf, and minimize should handle this % gracefully. if max(size(length)) == 2, red=length(2); length=length(1); else red=1; end if length>0, S='Linesearch'; else S='Function evaluation'; end i = 0; % zero the run length counter ls_failed = 0; % no previous line search has failed [f0 df0] = feval(f, X, varargin{:}); % get function value and gradient Z = X; X = unwrap(X); df0 = unwrap(df0); fprintf('%s %6i; Value %4.6e\r', S, i, f0); if exist('fflush','builtin') fflush(stdout); end fX = f0; i = i + (length<0); % count epochs?! s = -df0; d0 = -s'*s; % initial search direction (steepest) and slope x3 = red/(1-d0); % initial step is red/(|s|+1) while i < abs(length) % while not finished i = i + (length>0); % count iterations?! X0 = X; F0 = f0; dF0 = df0; % make a copy of current values if length>0, M = MAX; else M = min(MAX, -length-i); end while 1 % keep extrapolating as long as necessary x2 = 0; f2 = f0; d2 = d0; f3 = f0; df3 = df0; success = 0; while ~success && M > 0 try M = M - 1; i = i + (length<0); % count epochs?! [f3 df3] = feval(f, rewrap(Z,X+x3*s), varargin{:}); df3 = unwrap(df3); if isnan(f3) || isinf(f3) || any(isnan(df3)+isinf(df3)), error(''), end success = 1; catch % catch any error which occured in f x3 = (x2+x3)/2; % bisect and try again end end if f3 < F0, X0 = X+x3*s; F0 = f3; dF0 = df3; end % keep best values d3 = df3'*s; % new slope if d3 > SIG*d0 || f3 > f0+x3*RHO*d0 || M == 0 % are we done extrapolating? break end x1 = x2; f1 = f2; d1 = d2; % move point 2 to point 1 x2 = x3; f2 = f3; d2 = d3; % move point 3 to point 2 A = 6*(f1-f2)+3*(d2+d1)*(x2-x1); % make cubic extrapolation B = 3*(f2-f1)-(2*d1+d2)*(x2-x1); x3 = x1-d1*(x2-x1)^2/(B+sqrt(B*B-A*d1*(x2-x1))); % num. error possible, ok! if ~isreal(x3) || isnan(x3) || isinf(x3) || x3 < 0 % num prob | wrong sign? x3 = x2*EXT; % extrapolate maximum amount elseif x3 > x2*EXT % new point beyond extrapolation limit? x3 = x2*EXT; % extrapolate maximum amount elseif x3 < x2+INT*(x2-x1) % new point too close to previous point? x3 = x2+INT*(x2-x1); end end % end extrapolation while (abs(d3) > -SIG*d0 || f3 > f0+x3*RHO*d0) && M > 0 % keep interpolating if d3 > 0 || f3 > f0+x3*RHO*d0 % choose subinterval x4 = x3; f4 = f3; d4 = d3; % move point 3 to point 4 else x2 = x3; f2 = f3; d2 = d3; % move point 3 to point 2 end if f4 > f0 x3 = x2-(0.5*d2*(x4-x2)^2)/(f4-f2-d2*(x4-x2)); % quadratic interpolation else A = 6*(f2-f4)/(x4-x2)+3*(d4+d2); % cubic interpolation B = 3*(f4-f2)-(2*d2+d4)*(x4-x2); x3 = x2+(sqrt(B*B-A*d2*(x4-x2)^2)-B)/A; % num. error possible, ok! end if isnan(x3) || isinf(x3) x3 = (x2+x4)/2; % if we had a numerical problem then bisect end x3 = max(min(x3, x4-INT*(x4-x2)),x2+INT*(x4-x2)); % don't accept too close [f3 df3] = feval(f, rewrap(Z,X+x3*s), varargin{:}); df3 = unwrap(df3); if f3 < F0, X0 = X+x3*s; F0 = f3; dF0 = df3; end % keep best values M = M - 1; i = i + (length<0); % count epochs?! d3 = df3'*s; % new slope end % end interpolation if abs(d3) < -SIG*d0 && f3 < f0+x3*RHO*d0 % if line search succeeded X = X+x3*s; f0 = f3; fX = [fX' f0]'; % update variables fprintf('%s %6i; Value %4.6e\r', S, i, f0); if exist('fflush','builtin') fflush(stdout); end s = (df3'*df3-df0'*df3)/(df0'*df0)*s - df3; % Polack-Ribiere CG direction df0 = df3; % swap derivatives d3 = d0; d0 = df0'*s; if d0 > 0 % new slope must be negative s = -df0; d0 = -s'*s; % otherwise use steepest direction end x3 = x3 * min(RATIO, d3/(d0-realmin)); % slope ratio but max RATIO ls_failed = 0; % this line search did not fail else X = X0; f0 = F0; df0 = dF0; % restore best point so far if ls_failed || i > abs(length) % line search failed twice in a row break; % or we ran out of time, so we give up end s = -df0; d0 = -s'*s; % try steepest x3 = 1/(1-d0); ls_failed = 1; % this line search failed end end X = rewrap(Z,X); fprintf('\n'); if exist('fflush','builtin') fflush(stdout); end % Extract the numerical values from "s" into the column vector "v". The % variable "s" can be of any type, including struct and cell array. % Non-numerical elements are ignored. See also the reverse rewrap.m. function v = unwrap(s) v = []; if isnumeric(s) v = s(:); % numeric values are recast to column vector elseif isstruct(s) v = unwrap(struct2cell(orderfields(s))); % alphabetize, conv to cell, recurse elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially v = [v; unwrap(s{i})]; end end % other types are ignored % Map the numerical elements in the vector "v" onto the variables "s" which can % be of any type. The number of numerical elements must match; on exit "v" % should be empty. Non-numerical entries are just copied. See also unwrap.m. function [s v] = rewrap(s, v) if isnumeric(s) if numel(v) < numel(s) error('The vector for conversion contains too few elements') end s = reshape(v(1:numel(s)), size(s)); % numeric values are reshaped v = v(numel(s)+1:end); % remaining arguments passed on elseif isstruct(s) [s p] = orderfields(s); p(p) = 1:numel(p); % alphabetize, store ordering [t v] = rewrap(struct2cell(s), v); % convert to cell, recurse s = orderfields(cell2struct(t,fieldnames(s),1),p); % conv to struct, reorder elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially [s{i} v] = rewrap(s{i}, v); end end % other types are not processed gpml-matlab-v3.0-2010-07-23/util/minimize_lbfgsb_objfun.m000644 000765 000024 00000005207 11422240326 022707 0ustar00carlstaff000000 000000 function y = minimize_lbfgsb_objfun(X,varargin) % extract input arguments varargin = varargin{1}; strctX = varargin{2}; f = varargin{1}; % global variables serve as communication interface between calls global minimize_lbfgsb_iteration_number global minimize_lbfgsb_objective global minimize_lbfgsb_gradient global minimize_lbfgsb_X recompute = 0; if minimize_lbfgsb_iteration_number==0 recompute = 1; else if norm(unwrap(X)-unwrap(minimize_lbfgsb_X))>1e-10 recompute = 1; end end if recompute [y,G] = feval(f,rewrap(strctX,X),varargin{3:end}); else y = minimize_lbfgsb_objective(minimize_lbfgsb_iteration_number); G = minimize_lbfgsb_gradient; end % increase global counter, memorise objective function, gradient and position minimize_lbfgsb_iteration_number = minimize_lbfgsb_iteration_number+1; minimize_lbfgsb_objective(minimize_lbfgsb_iteration_number,1) = y; minimize_lbfgsb_gradient = G; minimize_lbfgsb_X = X; % Extract the numerical values from "s" into the column vector "v". The % variable "s" can be of any type, including struct and cell array. % Non-numerical elements are ignored. See also the reverse rewrap.m. function v = unwrap(s) v = []; if isnumeric(s) v = s(:); % numeric values are recast to column vector elseif isstruct(s) v = unwrap(struct2cell(orderfields(s)));% alphabetize, conv to cell, recurse elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially v = [v; unwrap(s{i})]; end end % other types are ignored % Map the numerical elements in the vector "v" onto the variables "s" which can % be of any type. The number of numerical elements must match; on exit "v" % should be empty. Non-numerical entries are just copied. See also unwrap.m. function [s v] = rewrap(s, v) if isnumeric(s) if numel(v) < numel(s) error('The vector for conversion contains too few elements') end s = reshape(v(1:numel(s)), size(s)); % numeric values are reshaped v = v(numel(s)+1:end); % remaining arguments passed on elseif isstruct(s) [s p] = orderfields(s); p(p) = 1:numel(p); % alphabetize, store ordering [t v] = rewrap(struct2cell(s), v); % convert to cell, recurse s = orderfields(cell2struct(t,fieldnames(s),1),p); % conv to struct, reorder elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially [s{i} v] = rewrap(s{i}, v); end end % other types are not processed gpml-matlab-v3.0-2010-07-23/util/rewrap.m000644 000765 000024 00000001766 11422240326 017512 0ustar00carlstaff000000 000000 % Map the numerical elements in the vector "v" onto the variables "s" which can % be of any type. The number of numerical elements must match; on exit "v" % should be empty. Non-numerical entries are just copied. See also unwrap.m. function [s v] = rewrap(s, v) if isnumeric(s) if numel(v) < numel(s) error('The vector for conversion contains too few elements') end s = reshape(v(1:numel(s)), size(s)); % numeric values are reshaped v = v(numel(s)+1:end); % remaining arguments passed on elseif isstruct(s) [s p] = orderfields(s); p(p) = 1:numel(p); % alphabetize, store ordering [t v] = rewrap(struct2cell(s), v); % convert to cell, recurse s = orderfields(cell2struct(t,fieldnames(s),1),p); % conv to struct, reorder elseif iscell(s) for i = 1:numel(s) % cell array elements are handled sequentially [s{i} v] = rewrap(s{i}, v); end end % other types are not processed gpml-matlab-v3.0-2010-07-23/util/binaryGP.m000644 000765 000024 00000011547 11422240326 017723 0ustar00carlstaff000000 000000 function varargout = binaryGP(hyper, approx, covfunc, lik, x, y, xstar) % Approximate binary Gaussian Process classification. Two modes are possible: % training or testing: if no test cases are supplied, then the approximate % negative log marginal likelihood and its partial derivatives wrt the % hyperparameters is computed; this mode is used to fit the hyperparameters. If % test cases are given, then the test set predictive probabilities are % returned. Exact inference is intractible, the function uses a specified % approximation method (see approximations.m), flexible covariance functions % (see covFunctions.m) and likelihood functions (see likelihoods.m). % % usage: [nlZ, dnlZ ] = binaryGP(hyper, approx, covfunc, lik, x, y); % or: [p,mu,s2,nlZ] = binaryGP(hyper, approx, covfunc, lik, x, y, xstar); % % where: % % hyper is a column vector of hyperparameters % approx is a function specifying an approximation method for inference % covfunc is the name of the covariance function (see below) % lik is the name of the likelihood function % x is a n by D matrix of training inputs % y is a (column) vector (of size n) of binary +1/-1 targets % xstar is a nn by D matrix of test inputs % nlZ is the returned value of the negative log marginal likelihood % dnlZ is a (column) vector of partial derivatives of the negative % log marginal likelihood wrt each hyperparameter % p is a (column) vector (of length nn) of predictive probabilities % mu is a (column) vector (of length nn) of predictive latent means % s2 is a (column) vector (of length nn) of predictive latent variances % % The length of the vector of hyperparameters depends on the covariance % function, as specified by the "covfunc" input to the function, specifying the % name of a covariance function. A number of different covariance function are % implemented, and it is not difficult to add new ones. See covFunctions.m for % the details. % % The "lik" input argument specifies the name of the likelihood function (see % likelihoods.m). % % The "approx" input argument to the function specifies an approximation method % (see approximations.m). An approximation method returns a representation of % the approximate Gaussian posterior. Usually, the approximate posterior admits % the form N(m=K*alpha, V=inv(inv(K)+W)), where alpha is a vector and W is % diagonal. The approximation method returns: % % alpha is a (sparse or full column vector) containing inv(K)*m, where K % is the prior covariance matrix and m the approx posterior mean % sW is a (sparse or full column) vector containing diagonal of sqrt(W) % the approximate posterior covariance matrix is inv(inv(K)+W) % L is a (sparse or full) matrix, L = chol(sW*K*sW+eye(n)) % % In cases where the approximate posterior variance does not admit the form % V=inv(inv(K)+W) with diagonal W, L contains instead -inv(K+inv(W)), and sW % is unused. % % The alpha parameter may be sparse. In that case sW and L can either be sparse % or full (retaining only the non-zero rows and columns, as indicated by the % sparsity structure of alpha). The L paramter is allowed to be empty, in % which case it will be computed. % % The function can conveniently be used with the "minimize" function to train % a Gaussian Process, eg: % % [hyper, fX, i] = minimize(hyper, 'binaryGP', length, 'approxEP', 'covSEiso', 'logistic', x, y); % % where "length" gives the length of the run: if it is positive, it gives the % maximum number of line searches, if negative its absolute gives the maximum % allowed number of function evaluations. % % Copyright (c) 2007 Carl Edward Rasmussen and Hannes Nickisch, 2010-07-21. if nargin<6 || nargin>7 disp('Usage: [nlZ, dnlZ ] = binaryGP(hyper,approx,covfunc,lik,x,y);') disp(' or: [p,mu,s2,nlZ] = binaryGP(hyper,approx,covfunc,lik,x,y,xstar);') return end % Note, this function is just a wrapper provided for backward compatibility, % the functionality is now provided by the more general gp function. if strcmp(lik,'logistic') lik = @likLogistic; elseif strcmpi(lik,'cumgauss') lik = @likErf; else error('Allowable likelihoods: logistic and cumGauss.'); end if strcmp(approx,'approxLA') inf = @infLaplace; elseif strcmp(approx,'approxEP') inf = @infEP; else error('Allowable approximations: approxLA and approxEP.'); end mean = @meanZero; varargout = cell(nargout, 1); % allocate the right number of output arguments hyp.cov = hyper; if nargin==6 [varargout{:}] = gp(hyp,inf,mean,covfunc,lik,x,y); if nargout>1, varargout{2} = varargout{2}.cov; end else [ymu,ys2,fmu,fs2] = gp(hyp,inf,mean,covfunc,lik,x,y,xstar); if nargout>0, varargout{1} = (1+ymu)/2; end if nargout>1, varargout{2} = fmu; end if nargout>2, varargout{3} = fs2; end if nargout>3, varargout{4} = gp(hyp,inf,mean,covfunc,lik,x,y); end % nlZ endgpml-matlab-v3.0-2010-07-23/util/elsympol.m000644 000765 000024 00000001273 11422240326 020047 0ustar00carlstaff000000 000000 % Evaluate the order R elementary symmetric polynomial Newton's identity aka % the Newton–Girard formulae: http://en.wikipedia.org/wiki/Newton's_identities % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-01-10. function E = elsympol(Z,R) % evaluate 'power sums' of the individual terms in Z sz = size(Z); P = zeros([sz(1:2),R]); for r=1:R, P(:,:,r) = sum(Z.^r,3); end E = zeros([sz(1:2),R+1]); % E(:,:,r+1) yields polynomial r E(:,:,1) = ones(sz(1:2)); if R==0, return, end % init recursion E(:,:,2) = P(:,:,1); if R==1, return, end % init recursion for r=2:R for i=1:r E(:,:,r+1) = E(:,:,r+1) + P(:,:,i).*E(:,:,r+1-i)*(-1)^(i-1)/r; end end gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabscalar.cpp000644 000765 000024 00000001730 11422240326 022414 0ustar00carlstaff000000 000000 #include "matlabscalar.h" #include "matlabexception.h" // Function definitions. // ----------------------------------------------------------------- double& getMatlabScalar (const mxArray* ptr) { if (!mxIsDouble(ptr)) throw MatlabException("Matlab array must be of type double"); if (mxGetNumberOfElements(ptr) != 1) throw MatlabException("The Matlab array must be a scalar"); return *mxGetPr(ptr); } double& createMatlabScalar (mxArray*& ptr) { ptr = mxCreateDoubleScalar(0); return *mxGetPr(ptr); } // Function definitions for class MatlabScalar. // --------------------------------------------------------------- MatlabScalar::MatlabScalar (const mxArray* ptr) : x(getMatlabScalar(ptr)) { } MatlabScalar::MatlabScalar (mxArray*& ptr, double value) : x(createMatlabScalar(ptr)) { x = value; } MatlabScalar::MatlabScalar (MatlabScalar& source) : x(source.x) { } MatlabScalar& MatlabScalar::operator= (double value) { x = value; return *this; } gpml-matlab-v3.0-2010-07-23/util/lbfgsb/program.cpp000644 000765 000024 00000007675 11422240326 021453 0ustar00carlstaff000000 000000 #include "program.h" #include // Function definitions. // ----------------------------------------------------------------- // Copy a C-style string (a null-terminated character array) to a // non-C-style string (a simple character array). The length of the // destination character array is given by "ndest". If the source is // shorter than the destination, the destination is padded with blank // characters. void copyCStrToCharArray (const char* source, char* dest, int ndest) { // Get the length of the source C string. int nsource = strlen(source); // Only perform the copy if the source can fit into the destination. if (nsource < ndest) { // Copy the string. strcpy(dest,source); // Fill in the rest of the string with blanks. for (int i = nsource; i < ndest; i++) dest[i] = ' '; } } // Return true if the two strings are the same. The second input // argument must be a C-style string (a null-terminated character // array), but the first input argument need not be one. We only // compare the two strings up to the length of "cstr". bool strIsEqualToCStr (const char* str, const char* cstr) { // Get the length of the C string. int n = strlen(cstr); return !strncmp(str,cstr,n); } // Function definitions for class Program. // ----------------------------------------------------------------- Program::Program (int n, double* x, double* lb, double* ub, int* btype, int m, int maxiter, double factr, double pgtol) { this->n = n; this->x = x; this->lb = lb; this->ub = ub; this->btype = btype; this->m = m; this->maxiter = maxiter; this->factr = factr; this->pgtol = pgtol; this->iprint = defaultprintlevel; this->owner = false; initStructures(); } Program::Program (int n, int m, int maxiter, double factr, double pgtol) { this->n = n; this->x = new double[n]; this->lb = new double[n]; this->ub = new double[n]; this->btype = new int[n]; this->m = m; this->maxiter = maxiter; this->factr = factr; this->pgtol = pgtol; this->iprint = defaultprintlevel; this->owner = true; initStructures(); } Program::~Program() { if (owner) { delete[] x; delete[] lb; delete[] ub; delete[] btype; } delete[] g; delete[] wa; delete[] iwa; } SolverExitStatus Program::runSolver() { SolverExitStatus status = success; // The return value. // Initialize the objective function and gradient to zero. f = 0; for (int i = 0; i < n; i++) g[i] = 0; // This initial call sets up the structures for L-BFGS. callLBFGS("START"); // Repeat until we've reached the maximum number of iterations. int t = 0; while (true) { // Do something according to the "task" from the previous call to // L-BFGS. if (strIsEqualToCStr(task,"FG")) { // Evaluate the objective function and the gradient of the // objective at the current point. f = computeObjective(n,x); computeGradient(n,x,g); } else if (strIsEqualToCStr(task,"NEW_X")) { // Go to the next iteration and call the iterative callback // routine. t++; iterCallback(t,x,f); // If we've reached the maximum number of iterations, terminate // the optimization. if (t == maxiter) { callLBFGS("STOP"); break; } } else if (strIsEqualToCStr(task,"CONV")) break; else if (strIsEqualToCStr(task,"ABNO")) { status = abnormalTermination; break; } else if (strIsEqualToCStr(task,"ERROR")) { status = errorOnInput; break; } // Call L-BFGS again. callLBFGS(); } return status; } void Program::callLBFGS (const char* cmd) { if (cmd) copyCStrToCharArray(cmd,task,60); setulb_(&n,&m,x,lb,ub,btype,&f,g,&factr,&pgtol,wa,iwa,task,&iprint, csave,lsave,isave,dsave); } void Program::initStructures() { f = 0; g = new double[n]; wa = new double[(2*m + 4)*n + 12*m*(m + 1)]; iwa = new int[3*n]; } gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabstring.cpp000644 000765 000024 00000002007 11422240326 022453 0ustar00carlstaff000000 000000 #include "matlabstring.h" #include "matlabexception.h" // Function definitions. // ----------------------------------------------------------------- char* copystring (const char* source) { int n = strlen(source); // The length of the string. char* dest = new char[n+1]; // The return value. strcpy(dest,source); return dest; } // Function definitions for class MatlabString. // ----------------------------------------------------------------- MatlabString::MatlabString (const mxArray* ptr) { s = 0; // Check to make sure the Matlab array is a string. if (!mxIsChar(ptr)) throw MatlabException("Matlab array must be a string (of type CHAR)"); // Get the string passed as a Matlab array. s = mxArrayToString(ptr); if (s == 0) throw MatlabException("Unable to obtain string from Matlab array"); } MatlabString::MatlabString (const MatlabString& source) { s = 0; // Copy the source string. s = copystring(source.s); } MatlabString::~MatlabString() { if (s) mxFree(s); } gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabexception.h000644 000765 000024 00000001545 11422240326 022616 0ustar00carlstaff000000 000000 #ifndef INCLUDE_MATLABEXCEPTION #define INCLUDE_MATLABEXCEPTION #include // Class MatlabException // ----------------------------------------------------------------- // This class just makes it easier for me to throw exceptions. Its // functionality really has nothing to do with MATLAB. class MatlabException : public std::exception { public: MatlabException (const char* message) throw(); ~MatlabException() throw() { }; // The copy constructor makes a shallow copy. MatlabException (const MatlabException& source) throw(); // The copy assignment operator makes a shallow copy as well. MatlabException& operator= (const MatlabException& source); // Return the message string. virtual const char* what () const throw() { return message; }; private: const char* message; // The error message. }; #endif gpml-matlab-v3.0-2010-07-23/util/lbfgsb/Makefile000644 000765 000024 00000003161 11422240326 020722 0ustar00carlstaff000000 000000 # path to your Matlab installation # you can find it by the commands 'locate matlab' or 'find / -name "matlab"' MATLAB_HOME=/agbs/share/sw/matlab #MATLAB_HOME=/usr/local/share/matlabR2008b # for 'make mex' you need to provide the variable MEX and have mex set up # for 'make nomex' you need to provide the variables MEX_SUFFIX and MATLAB_LIB # 32 bit Linux MEX = $(MATLAB_HOME)/bin/mex MEX_SUFFIX = mexglx MATLAB_LIB = -L$(MATLAB_HOME)/bin/glnx86 -lmex # 64 bit Linux #MEX = $(MATLAB_HOME)/bin/mex #MEX_SUFFIX = mexa64 #MATLAB_LIB = -L$(MATLAB_HOME)/bin/glnxa64 -lmex # Mac OS X #MEX = /Applications/MATLAB7/bin/mex #MEX_SUFFIX = mexmac #MATLAB_LIB = -L$(MATLAB_HOME)/bin/maci -lmex # Do not edit below ############################################################ CXX = g++ F77 = f77 CFLAGS = -O3 -ffast-math -fomit-frame-pointer -fPIC -Werror -pthread -Wall -ansi FFLAGS = -O3 -fPIC -fexceptions FORTRAN_LIB=g2c MATLAB_INCLUDE=-I$(MATLAB_HOME)/extern/include TARGET = lbfgsb OBJS = solver.o matlabexception.o matlabscalar.o matlabstring.o \ matlabmatrix.o arrayofmatrices.o program.o matlabprogram.o \ lbfgsb.o %.o: %.cpp $(CXX) $(CFLAGS) $(MATLAB_INCLUDE) -o $@ -c $^ %.o: %.f $(F77) $(FFLAGS) -o $@ -c $^ mex: $(TARGET) clean nomex: $(TARGET)_nomex clean $(TARGET): $(OBJS) $(MEX) -cxx CXX=$(CXX) CC=$(CXX) FC=$(FCC) LD=$(CXX) -l$(FORTRAN_LIB) -lm \ -O -output $@ $^ $(TARGET)_nomex: $(OBJS) $(CXX) $^ -shared -o $(TARGET).$(MEX_SUFFIX) $(MATLAB_LIB) -l$(FORTRAN_LIB) -lm clean: rm -f *.o cp $(TARGET).$(MEX_SUFFIX) ../$(TARGET).mex mv $(TARGET).$(MEX_SUFFIX) .. gpml-matlab-v3.0-2010-07-23/util/lbfgsb/solver.f000644 000765 000024 00000423410 11422240326 020746 0ustar00carlstaff000000 000000 c================ L-BFGS-B (version 2.4) ========================== subroutine setulb(n, m, x, l, u, nbd, f, g, factr, pgtol, wa, iwa, + task, iprint, csave, lsave, isave, dsave) character*60 task, csave logical lsave(4) integer n, m, iprint, + nbd(n), iwa(3*n), isave(44) double precision f, factr, pgtol, x(n), l(n), u(n), g(n), + wa(2*m*n+4*n+11*m*m+8*m), dsave(29) c ************ c c Subroutine setulb c c This subroutine partitions the working arrays wa and iwa, and c then uses the limited memory BFGS method to solve the bound c constrained optimization problem by calling mainlb. c (The direct method will be used in the subspace minimization.) c c n is an integer variable. c On entry n is the dimension of the problem. c On exit n is unchanged. c c m is an integer variable. c On entry m is the maximum number of variable metric corrections c used to define the limited memory matrix. c On exit m is unchanged. c c x is a double precision array of dimension n. c On entry x is an approximation to the solution. c On exit x is the current approximation. c c l is a double precision array of dimension n. c On entry l is the lower bound on x. c On exit l is unchanged. c c u is a double precision array of dimension n. c On entry u is the upper bound on x. c On exit u is unchanged. c c nbd is an integer array of dimension n. c On entry nbd represents the type of bounds imposed on the c variables, and must be specified as follows: c nbd(i)=0 if x(i) is unbounded, c 1 if x(i) has only a lower bound, c 2 if x(i) has both lower and upper bounds, and c 3 if x(i) has only an upper bound. c On exit nbd is unchanged. c c f is a double precision variable. c On first entry f is unspecified. c On final exit f is the value of the function at x. c c g is a double precision array of dimension n. c On first entry g is unspecified. c On final exit g is the value of the gradient at x. c c factr is a double precision variable. c On entry factr >= 0 is specified by the user. The iteration c will stop when c c (f^k - f^{k+1})/max{|f^k|,|f^{k+1}|,1} <= factr*epsmch c c where epsmch is the machine precision, which is automatically c generated by the code. Typical values for factr: 1.d+12 for c low accuracy; 1.d+7 for moderate accuracy; 1.d+1 for extremely c high accuracy. c On exit factr is unchanged. c c pgtol is a double precision variable. c On entry pgtol >= 0 is specified by the user. The iteration c will stop when c c max{|proj g_i | i = 1, ..., n} <= pgtol c c where pg_i is the ith component of the projected gradient. c On exit pgtol is unchanged. c c wa is a double precision working array of length c (2mmax + 4)nmax + 11mmax^2 + 8mmax. c c iwa is an integer working array of length 3nmax. c c task is a working string of characters of length 60 indicating c the current job when entering and quitting this subroutine. c c iprint is an integer variable that must be set by the user. c It controls the frequency and type of output generated: c iprint<0 no output is generated; c iprint=0 print only one line at the last iteration; c 0100 print details of every iteration including x and g; c When iprint > 0, the file iterate.dat will be created to c summarize the iteration. c c csave is a working string of characters of length 60. c c lsave is a logical working array of dimension 4. c On exit with 'task' = NEW_X, the following information is c available: c If lsave(1) = .true. then the initial X has been replaced by c its projection in the feasible set; c If lsave(2) = .true. then the problem is constrained; c If lsave(3) = .true. then each variable has upper and lower c bounds; c c isave is an integer working array of dimension 44. c On exit with 'task' = NEW_X, the following information is c available: c isave(22) = the total number of intervals explored in the c search of Cauchy points; c isave(26) = the total number of skipped BFGS updates before c the current iteration; c isave(30) = the number of current iteration; c isave(31) = the total number of BFGS updates prior the current c iteration; c isave(33) = the number of intervals explored in the search of c Cauchy point in the current iteration; c isave(34) = the total number of function and gradient c evaluations; c isave(36) = the number of function value or gradient c evaluations in the current iteration; c if isave(37) = 0 then the subspace argmin is within the box; c if isave(37) = 1 then the subspace argmin is beyond the box; c isave(38) = the number of free variables in the current c iteration; c isave(39) = the number of active constraints in the current c iteration; c n + 1 - isave(40) = the number of variables leaving the set of c active constraints in the current iteration; c isave(41) = the number of variables entering the set of active c constraints in the current iteration. c c dsave is a double precision working array of dimension 29. c On exit with 'task' = NEW_X, the following information is c available: c dsave(1) = current 'theta' in the BFGS matrix; c dsave(2) = f(x) in the previous iteration; c dsave(3) = factr*epsmch; c dsave(4) = 2-norm of the line search direction vector; c dsave(5) = the machine precision epsmch generated by the code; c dsave(7) = the accumulated time spent on searching for c Cauchy points; c dsave(8) = the accumulated time spent on c subspace minimization; c dsave(9) = the accumulated time spent on line search; c dsave(11) = the slope of the line search function at c the current point of line search; c dsave(12) = the maximum relative step length imposed in c line search; c dsave(13) = the infinity norm of the projected gradient; c dsave(14) = the relative step length in the line search; c dsave(15) = the slope of the line search function at c the starting point of the line search; c dsave(16) = the square of the 2-norm of the line search c direction vector. c c Subprograms called: c c L-BFGS-B Library ... mainlb. c c c References: c c [1] R. H. Byrd, P. Lu, J. Nocedal and C. Zhu, ``A limited c memory algorithm for bound constrained optimization'', c SIAM J. Scientific Computing 16 (1995), no. 5, pp. 1190--1208. c c [2] C. Zhu, R.H. Byrd, P. Lu, J. Nocedal, ``L-BFGS-B: a c limited memory FORTRAN code for solving bound constrained c optimization problems'', Tech. Report, NAM-11, EECS Department, c Northwestern University, 1994. c c (Postscript files of these papers are available via anonymous c ftp to ece.nwu.edu in the directory pub/lbfgs/lbfgs_bcm.) c c * * * c c NEOS, November 1994. (Latest revision April 1997.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer l1,l2,l3,lws,lr,lz,lt,ld,lwa,lwy,lsy,lss,lwt,lwn,lsnd if (task .eq. 'START') then isave(1) = m*n isave(2) = m**2 isave(3) = 4*m**2 isave(4) = 1 isave(5) = isave(4) + isave(1) isave(6) = isave(5) + isave(1) isave(7) = isave(6) + isave(2) isave(8) = isave(7) + isave(2) isave(9) = isave(8) isave(10) = isave(9) + isave(2) isave(11) = isave(10) + isave(3) isave(12) = isave(11) + isave(3) isave(13) = isave(12) + n isave(14) = isave(13) + n isave(15) = isave(14) + n isave(16) = isave(15) + n endif l1 = isave(1) l2 = isave(2) l3 = isave(3) lws = isave(4) lwy = isave(5) lsy = isave(6) lss = isave(7) lwt = isave(9) lwn = isave(10) lsnd = isave(11) lz = isave(12) lr = isave(13) ld = isave(14) lt = isave(15) lwa = isave(16) call mainlb(n,m,x,l,u,nbd,f,g,factr,pgtol, + wa(lws),wa(lwy),wa(lsy),wa(lss),wa(lwt), + wa(lwn),wa(lsnd),wa(lz),wa(lr),wa(ld),wa(lt), + wa(lwa),iwa(1),iwa(n+1),iwa(2*n+1),task,iprint, + csave,lsave,isave(22),dsave) return end c======================= The end of setulb ============================= subroutine mainlb(n, m, x, l, u, nbd, f, g, factr, pgtol, ws, wy, + sy, ss, wt, wn, snd, z, r, d, t, wa, + index, iwhere, indx2, task, iprint, + csave, lsave, isave, dsave) character*60 task, csave logical lsave(4) integer n, m, iprint, nbd(n), index(n), + iwhere(n), indx2(n), isave(23) double precision f, factr, pgtol, + x(n), l(n), u(n), g(n), z(n), r(n), d(n), t(n), + wa(8*m), ws(n, m), wy(n, m), sy(m, m), ss(m, m), + wt(m, m), wn(2*m, 2*m), snd(2*m, 2*m), dsave(29) c ************ c c Subroutine mainlb c c This subroutine solves bound constrained optimization problems by c using the compact formula of the limited memory BFGS updates. c c n is an integer variable. c On entry n is the number of variables. c On exit n is unchanged. c c m is an integer variable. c On entry m is the maximum number of variable metric c corrections allowed in the limited memory matrix. c On exit m is unchanged. c c x is a double precision array of dimension n. c On entry x is an approximation to the solution. c On exit x is the current approximation. c c l is a double precision array of dimension n. c On entry l is the lower bound of x. c On exit l is unchanged. c c u is a double precision array of dimension n. c On entry u is the upper bound of x. c On exit u is unchanged. c c nbd is an integer array of dimension n. c On entry nbd represents the type of bounds imposed on the c variables, and must be specified as follows: c nbd(i)=0 if x(i) is unbounded, c 1 if x(i) has only a lower bound, c 2 if x(i) has both lower and upper bounds, c 3 if x(i) has only an upper bound. c On exit nbd is unchanged. c c f is a double precision variable. c On first entry f is unspecified. c On final exit f is the value of the function at x. c c g is a double precision array of dimension n. c On first entry g is unspecified. c On final exit g is the value of the gradient at x. c c factr is a double precision variable. c On entry factr >= 0 is specified by the user. The iteration c will stop when c c (f^k - f^{k+1})/max{|f^k|,|f^{k+1}|,1} <= factr*epsmch c c where epsmch is the machine precision, which is automatically c generated by the code. c On exit factr is unchanged. c c pgtol is a double precision variable. c On entry pgtol >= 0 is specified by the user. The iteration c will stop when c c max{|proj g_i | i = 1, ..., n} <= pgtol c c where pg_i is the ith component of the projected gradient. c On exit pgtol is unchanged. c c ws, wy, sy, and wt are double precision working arrays used to c store the following information defining the limited memory c BFGS matrix: c ws, of dimension n x m, stores S, the matrix of s-vectors; c wy, of dimension n x m, stores Y, the matrix of y-vectors; c sy, of dimension m x m, stores S'Y; c ss, of dimension m x m, stores S'S; c wt, of dimension m x m, stores the Cholesky factorization c of (theta*S'S+LD^(-1)L'); see eq. c (2.26) in [3]. c c wn is a double precision working array of dimension 2m x 2m c used to store the LEL^T factorization of the indefinite matrix c K = [-D -Y'ZZ'Y/theta L_a'-R_z' ] c [L_a -R_z theta*S'AA'S ] c c where E = [-I 0] c [ 0 I] c c snd is a double precision working array of dimension 2m x 2m c used to store the lower triangular part of c N = [Y' ZZ'Y L_a'+R_z'] c [L_a +R_z S'AA'S ] c c z(n),r(n),d(n),t(n),wa(8*m) are double precision working arrays. c z is used at different times to store the Cauchy point and c the Newton point. c c c index is an integer working array of dimension n. c In subroutine freev, index is used to store the free and fixed c variables at the Generalized Cauchy Point (GCP). c c iwhere is an integer working array of dimension n used to record c the status of the vector x for GCP computation. c iwhere(i)=0 or -3 if x(i) is free and has bounds, c 1 if x(i) is fixed at l(i), and l(i) .ne. u(i) c 2 if x(i) is fixed at u(i), and u(i) .ne. l(i) c 3 if x(i) is always fixed, i.e., u(i)=x(i)=l(i) c -1 if x(i) is always free, i.e., no bounds on it. c c indx2 is an integer working array of dimension n. c Within subroutine cauchy, indx2 corresponds to the array iorder. c In subroutine freev, a list of variables entering and leaving c the free set is stored in indx2, and it is passed on to c subroutine formk with this information. c c task is a working string of characters of length 60 indicating c the current job when entering and leaving this subroutine. c c iprint is an INTEGER variable that must be set by the user. c It controls the frequency and type of output generated: c iprint<0 no output is generated; c iprint=0 print only one line at the last iteration; c 0100 print details of every iteration including x and g; c When iprint > 0, the file iterate.dat will be created to c summarize the iteration. c c csave is a working string of characters of length 60. c c lsave is a logical working array of dimension 4. c c isave is an integer working array of dimension 23. c c dsave is a double precision working array of dimension 29. c c c Subprograms called c c L-BFGS-B Library ... cauchy, subsm, lnsrlb, formk, c c errclb, prn1lb, prn2lb, prn3lb, active, projgr, c c freev, cmprlb, matupd, formt. c c Minpack2 Library ... timer, dpmeps. c c Linpack Library ... dcopy, ddot. c c c References: c c [1] R. H. Byrd, P. Lu, J. Nocedal and C. Zhu, ``A limited c memory algorithm for bound constrained optimization'', c SIAM J. Scientific Computing 16 (1995), no. 5, pp. 1190--1208. c c [2] C. Zhu, R.H. Byrd, P. Lu, J. Nocedal, ``L-BFGS-B: FORTRAN c Subroutines for Large Scale Bound Constrained Optimization'' c Tech. Report, NAM-11, EECS Department, Northwestern University, c 1994. c c [3] R. Byrd, J. Nocedal and R. Schnabel "Representations of c Quasi-Newton Matrices and their use in Limited Memory Methods'', c Mathematical Programming 63 (1994), no. 4, pp. 129-156. c c (Postscript files of these papers are available via anonymous c ftp to ece.nwu.edu in the directory pub/lbfgs/lbfgs_bcm.) c c * * * c c NEOS, November 1994. (Latest revision April 1997.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ logical prjctd,cnstnd,boxed,updatd,wrk character*3 word integer i,k,nintol,itfile,iback,nskip, + head,col,iter,itail,iupdat, + nint,nfgv,info,ifun, + iword,nfree,nact,ileave,nenter double precision theta,fold,ddot,dr,rr,tol,dpmeps, + xstep,sbgnrm,ddum,dnorm,dtd,epsmch, + cpu1,cpu2,cachyt,sbtime,lnscht,time1,time2, + gd,gdold,stp,stpmx,time double precision one,zero parameter (one=1.0d0,zero=0.0d0) if (task .eq. 'START') then call timer(time1) c Generate the current machine precision. epsmch = dpmeps() fold = 0.0d0 dnorm = 0.0d0 cpu1 = 0.0d0 gd = 0.0d0 sbgnrm = 0.0d0 stp = 0.0d0 stpmx = 0.0d0 gdold = 0.0d0 dtd = 0.0d0 c Initialize counters and scalars when task='START'. c for the limited memory BFGS matrices: col = 0 head = 1 theta = one iupdat = 0 updatd = .false. iback = 0 itail = 0 ifun = 0 iword = 0 nact = 0 ileave = 0 nenter = 0 c for operation counts: iter = 0 nfgv = 0 nint = 0 nintol = 0 nskip = 0 nfree = n c for stopping tolerance: tol = factr*epsmch c for measuring running time: cachyt = 0 sbtime = 0 lnscht = 0 c 'word' records the status of subspace solutions. word = '---' c 'info' records the termination information. info = 0 itfile = 0 if (iprint .ge. 1) then c open a summary file 'iterate.dat' open (8, file = 'iterate.dat', status = 'unknown') itfile = 8 endif c Check the input arguments for errors. call errclb(n,m,factr,l,u,nbd,task,info,k) if (task(1:5) .eq. 'ERROR') then call prn3lb(n,x,f,task,iprint,info,itfile, + iter,nfgv,nintol,nskip,nact,sbgnrm, + zero,nint,word,iback,stp,xstep,k, + cachyt,sbtime,lnscht) return endif call prn1lb(n,m,l,u,x,iprint,itfile,epsmch) c Initialize iwhere & project x onto the feasible set. call active(n,l,u,nbd,x,iwhere,iprint,prjctd,cnstnd,boxed) c The end of the initialization. else c restore local variables. prjctd = lsave(1) cnstnd = lsave(2) boxed = lsave(3) updatd = lsave(4) nintol = isave(1) itfile = isave(3) iback = isave(4) nskip = isave(5) head = isave(6) col = isave(7) itail = isave(8) iter = isave(9) iupdat = isave(10) nint = isave(12) nfgv = isave(13) info = isave(14) ifun = isave(15) iword = isave(16) nfree = isave(17) nact = isave(18) ileave = isave(19) nenter = isave(20) theta = dsave(1) fold = dsave(2) tol = dsave(3) dnorm = dsave(4) epsmch = dsave(5) cpu1 = dsave(6) cachyt = dsave(7) sbtime = dsave(8) lnscht = dsave(9) time1 = dsave(10) gd = dsave(11) stpmx = dsave(12) sbgnrm = dsave(13) stp = dsave(14) gdold = dsave(15) dtd = dsave(16) c After returning from the driver go to the point where execution c is to resume. if (task(1:5) .eq. 'FG_LN') goto 666 if (task(1:5) .eq. 'NEW_X') goto 777 if (task(1:5) .eq. 'FG_ST') goto 111 if (task(1:4) .eq. 'STOP') then if (task(7:9) .eq. 'CPU') then c restore the previous iterate. call dcopy(n,t,1,x,1) call dcopy(n,r,1,g,1) f = fold endif goto 999 endif endif c Compute f0 and g0. task = 'FG_START' c return to the driver to calculate f and g; reenter at 111. goto 1000 111 continue nfgv = 1 c Compute the infinity norm of the (-) projected gradient. call projgr(n,l,u,nbd,x,g,sbgnrm) if (iprint .ge. 1) then write (6,1002) iter,f,sbgnrm write (itfile,1003) iter,nfgv,sbgnrm,f endif if (sbgnrm .le. pgtol) then c terminate the algorithm. task = 'CONVERGENCE: NORM OF PROJECTED GRADIENT <= PGTOL' goto 999 endif c ----------------- the beginning of the loop -------------------------- 222 continue if (iprint .ge. 99) write (6,1001) iter + 1 iword = -1 c if (.not. cnstnd .and. col .gt. 0) then c skip the search for GCP. call dcopy(n,x,1,z,1) wrk = updatd nint = 0 goto 333 endif cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc c c Compute the Generalized Cauchy Point (GCP). c cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc call timer(cpu1) call cauchy(n,x,l,u,nbd,g,indx2,iwhere,t,d,z, + m,wy,ws,sy,wt,theta,col,head, + wa(1),wa(2*m+1),wa(4*m+1),wa(6*m+1),nint, + iprint,sbgnrm,info,epsmch) if (info .ne. 0) then c singular triangular system detected; refresh the lbfgs memory. if(iprint .ge. 1) write (6, 1005) info = 0 col = 0 head = 1 theta = one iupdat = 0 updatd = .false. call timer(cpu2) cachyt = cachyt + cpu2 - cpu1 goto 222 endif call timer(cpu2) cachyt = cachyt + cpu2 - cpu1 nintol = nintol + nint c Count the entering and leaving variables for iter > 0; c find the index set of free and active variables at the GCP. call freev(n,nfree,index,nenter,ileave,indx2, + iwhere,wrk,updatd,cnstnd,iprint,iter) nact = n - nfree 333 continue c If there are no free variables or B=theta*I, then c skip the subspace minimization. if (nfree .eq. 0 .or. col .eq. 0) goto 555 cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc c c Subspace minimization. c cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc call timer(cpu1) c Form the LEL^T factorization of the indefinite c matrix K = [-D -Y'ZZ'Y/theta L_a'-R_z' ] c [L_a -R_z theta*S'AA'S ] c where E = [-I 0] c [ 0 I] if (wrk) call formk(n,nfree,index,nenter,ileave,indx2,iupdat, + updatd,wn,snd,m,ws,wy,sy,theta,col,head,info) if (info .ne. 0) then c nonpositive definiteness in Cholesky factorization; c refresh the lbfgs memory and restart the iteration. if(iprint .ge. 1) write (6, 1006) info = 0 col = 0 head = 1 theta = one iupdat = 0 updatd = .false. call timer(cpu2) sbtime = sbtime + cpu2 - cpu1 goto 222 endif c compute r=-Z'B(xcp-xk)-Z'g (using wa(2m+1)=W'(xcp-x) c from 'cauchy'). call cmprlb(n,m,x,g,ws,wy,sy,wt,z,r,wa,index, + theta,col,head,nfree,cnstnd,info) if (info .ne. 0) goto 444 c call the direct method. call subsm(n,m,nfree,index,l,u,nbd,z,r,ws,wy,theta, + col,head,iword,wa,wn,iprint,info) 444 continue if (info .ne. 0) then c singular triangular system detected; c refresh the lbfgs memory and restart the iteration. if(iprint .ge. 1) write (6, 1005) info = 0 col = 0 head = 1 theta = one iupdat = 0 updatd = .false. call timer(cpu2) sbtime = sbtime + cpu2 - cpu1 goto 222 endif call timer(cpu2) sbtime = sbtime + cpu2 - cpu1 555 continue cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc c c Line search and optimality tests. c cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc c Generate the search direction d:=z-x. do 40 i = 1, n d(i) = z(i) - x(i) 40 continue call timer(cpu1) 666 continue call lnsrlb(n,l,u,nbd,x,f,fold,gd,gdold,g,d,r,t,z,stp,dnorm, + dtd,xstep,stpmx,iter,ifun,iback,nfgv,info,task, + boxed,cnstnd,csave,isave(22),dsave(17)) if (info .ne. 0 .or. iback .ge. 20) then c restore the previous iterate. call dcopy(n,t,1,x,1) call dcopy(n,r,1,g,1) f = fold if (col .eq. 0) then c abnormal termination. if (info .eq. 0) then info = -9 c restore the actual number of f and g evaluations etc. nfgv = nfgv - 1 ifun = ifun - 1 iback = iback - 1 endif task = 'ABNORMAL_TERMINATION_IN_LNSRCH' iter = iter + 1 goto 999 else c refresh the lbfgs memory and restart the iteration. if(iprint .ge. 1) write (6, 1008) if (info .eq. 0) nfgv = nfgv - 1 info = 0 col = 0 head = 1 theta = one iupdat = 0 updatd = .false. task = 'RESTART_FROM_LNSRCH' call timer(cpu2) lnscht = lnscht + cpu2 - cpu1 goto 222 endif else if (task(1:5) .eq. 'FG_LN') then c return to the driver for calculating f and g; reenter at 666. goto 1000 else c calculate and print out the quantities related to the new X. call timer(cpu2) lnscht = lnscht + cpu2 - cpu1 iter = iter + 1 c Compute the infinity norm of the projected (-)gradient. call projgr(n,l,u,nbd,x,g,sbgnrm) c Print iteration information. call prn2lb(n,x,f,g,iprint,itfile,iter,nfgv,nact, + sbgnrm,nint,word,iword,iback,stp,xstep) goto 1000 endif 777 continue c Test for termination. if (sbgnrm .le. pgtol) then c terminate the algorithm. task = 'CONVERGENCE: NORM OF PROJECTED GRADIENT <= PGTOL' goto 999 endif ddum = max(abs(fold), abs(f), one) if ((fold - f) .le. tol*ddum) then c terminate the algorithm. task = 'CONVERGENCE: REL_REDUCTION_OF_F <= FACTR*EPSMCH' if (iback .ge. 10) info = -5 c i.e., to issue a warning if iback>10 in the line search. goto 999 endif c Compute d=newx-oldx, r=newg-oldg, rr=y'y and dr=y's. do 42 i = 1, n r(i) = g(i) - r(i) 42 continue rr = ddot(n,r,1,r,1) if (stp .eq. one) then dr = gd - gdold ddum = -gdold else dr = (gd - gdold)*stp call dscal(n,stp,d,1) ddum = -gdold*stp endif if (dr .le. epsmch*ddum) then c skip the L-BFGS update. nskip = nskip + 1 updatd = .false. if (iprint .ge. 1) write (6,1004) dr, ddum goto 888 endif cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc c c Update the L-BFGS matrix. c cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc updatd = .true. iupdat = iupdat + 1 c Update matrices WS and WY and form the middle matrix in B. call matupd(n,m,ws,wy,sy,ss,d,r,itail, + iupdat,col,head,theta,rr,dr,stp,dtd) c Form the upper half of the pds T = theta*SS + L*D^(-1)*L'; c Store T in the upper triangular of the array wt; c Cholesky factorize T to J*J' with c J' stored in the upper triangular of wt. call formt(m,wt,sy,ss,col,theta,info) if (info .ne. 0) then c nonpositive definiteness in Cholesky factorization; c refresh the lbfgs memory and restart the iteration. if(iprint .ge. 1) write (6, 1007) info = 0 col = 0 head = 1 theta = one iupdat = 0 updatd = .false. goto 222 endif c Now the inverse of the middle matrix in B is c [ D^(1/2) O ] [ -D^(1/2) D^(-1/2)*L' ] c [ -L*D^(-1/2) J ] [ 0 J' ] 888 continue c -------------------- the end of the loop ----------------------------- goto 222 999 continue call timer(time2) time = time2 - time1 call prn3lb(n,x,f,task,iprint,info,itfile, + iter,nfgv,nintol,nskip,nact,sbgnrm, + time,nint,word,iback,stp,xstep,k, + cachyt,sbtime,lnscht) 1000 continue c Save local variables. lsave(1) = prjctd lsave(2) = cnstnd lsave(3) = boxed lsave(4) = updatd isave(1) = nintol isave(3) = itfile isave(4) = iback isave(5) = nskip isave(6) = head isave(7) = col isave(8) = itail isave(9) = iter isave(10) = iupdat isave(12) = nint isave(13) = nfgv isave(14) = info isave(15) = ifun isave(16) = iword isave(17) = nfree isave(18) = nact isave(19) = ileave isave(20) = nenter dsave(1) = theta dsave(2) = fold dsave(3) = tol dsave(4) = dnorm dsave(5) = epsmch dsave(6) = cpu1 dsave(7) = cachyt dsave(8) = sbtime dsave(9) = lnscht dsave(10) = time1 dsave(11) = gd dsave(12) = stpmx dsave(13) = sbgnrm dsave(14) = stp dsave(15) = gdold dsave(16) = dtd 1001 format (//,'ITERATION ',i5) 1002 format + (/,'At iterate',i5,4x,'f= ',1p,d12.5,4x,'|proj g|= ',1p,d12.5) 1003 format (2(1x,i4),5x,'-',5x,'-',3x,'-',5x,'-',5x,'-',8x,'-',3x, + 1p,2(1x,d10.3)) 1004 format (' ys=',1p,e10.3,' -gs=',1p,e10.3,' BFGS update SKIPPED') 1005 format (/, +' Singular triangular system detected;',/, +' refresh the lbfgs memory and restart the iteration.') 1006 format (/, +' Nonpositive definiteness in Cholesky factorization in formk;',/, +' refresh the lbfgs memory and restart the iteration.') 1007 format (/, +' Nonpositive definiteness in Cholesky factorization in formt;',/, +' refresh the lbfgs memory and restart the iteration.') 1008 format (/, +' Bad direction in the line search;',/, +' refresh the lbfgs memory and restart the iteration.') return end c======================= The end of mainlb ============================= subroutine active(n, l, u, nbd, x, iwhere, iprint, + prjctd, cnstnd, boxed) logical prjctd, cnstnd, boxed integer n, iprint, nbd(n), iwhere(n) double precision x(n), l(n), u(n) c ************ c c Subroutine active c c This subroutine initializes iwhere and projects the initial x to c the feasible set if necessary. c c iwhere is an integer array of dimension n. c On entry iwhere is unspecified. c On exit iwhere(i)=-1 if x(i) has no bounds c 3 if l(i)=u(i) c 0 otherwise. c In cauchy, iwhere is given finer gradations. c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer nbdd,i double precision zero parameter (zero=0.0d0) c Initialize nbdd, prjctd, cnstnd and boxed. nbdd = 0 prjctd = .false. cnstnd = .false. boxed = .true. c Project the initial x to the easible set if necessary. do 10 i = 1, n if (nbd(i) .gt. 0) then if (nbd(i) .le. 2 .and. x(i) .le. l(i)) then if (x(i) .lt. l(i)) then prjctd = .true. x(i) = l(i) endif nbdd = nbdd + 1 else if (nbd(i) .ge. 2 .and. x(i) .ge. u(i)) then if (x(i) .gt. u(i)) then prjctd = .true. x(i) = u(i) endif nbdd = nbdd + 1 endif endif 10 continue c Initialize iwhere and assign values to cnstnd and boxed. do 20 i = 1, n if (nbd(i) .ne. 2) boxed = .false. if (nbd(i) .eq. 0) then c this variable is always free iwhere(i) = -1 c otherwise set x(i)=mid(x(i), u(i), l(i)). else cnstnd = .true. if (nbd(i) .eq. 2 .and. u(i) - l(i) .le. zero) then c this variable is always fixed iwhere(i) = 3 else iwhere(i) = 0 endif endif 20 continue if (iprint .ge. 0) then if (prjctd) write (6,*) + 'The initial X is infeasible. Restart with its projection.' if (.not. cnstnd) + write (6,*) 'This problem is unconstrained.' endif if (iprint .gt. 0) write (6,1001) nbdd 1001 format (/,'At X0 ',i9,' variables are exactly at the bounds') return end c======================= The end of active ============================= subroutine bmv(m, sy, wt, col, v, p, info) integer m, col, info double precision sy(m, m), wt(m, m), v(2*col), p(2*col) c ************ c c Subroutine bmv c c This subroutine computes the product of the 2m x 2m middle matrix c in the compact L-BFGS formula of B and a 2m vector v; c it returns the product in p. c c m is an integer variable. c On entry m is the maximum number of variable metric corrections c used to define the limited memory matrix. c On exit m is unchanged. c c sy is a double precision array of dimension m x m. c On entry sy specifies the matrix S'Y. c On exit sy is unchanged. c c wt is a double precision array of dimension m x m. c On entry wt specifies the upper triangular matrix J' which is c the Cholesky factor of (thetaS'S+LD^(-1)L'). c On exit wt is unchanged. c c col is an integer variable. c On entry col specifies the number of s-vectors (or y-vectors) c stored in the compact L-BFGS formula. c On exit col is unchanged. c c v is a double precision array of dimension 2col. c On entry v specifies vector v. c On exit v is unchanged. c c p is a double precision array of dimension 2col. c On entry p is unspecified. c On exit p is the product Mv. c c info is an integer variable. c On entry info is unspecified. c On exit info = 0 for normal return, c = nonzero for abnormal return when the system c to be solved by dtrsl is singular. c c Subprograms called: c c Linpack ... dtrsl. c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer i,k,i2 double precision sum if (col .eq. 0) return c PART I: solve [ D^(1/2) O ] [ p1 ] = [ v1 ] c [ -L*D^(-1/2) J ] [ p2 ] [ v2 ]. c solve Jp2=v2+LD^(-1)v1. p(col + 1) = v(col + 1) do 20 i = 2, col i2 = col + i sum = 0.0d0 do 10 k = 1, i - 1 sum = sum + sy(i,k)*v(k)/sy(k,k) 10 continue p(i2) = v(i2) + sum 20 continue c Solve the triangular system call dtrsl(wt,m,col,p(col+1),11,info) if (info .ne. 0) return c solve D^(1/2)p1=v1. cc do 30 i = 1, col cc p(i) = v(i)/sqrt(sy(i,i)) cc30 continue c PART II: solve [ -D^(1/2) D^(-1/2)*L' ] [ p1 ] = [ p1 ] c [ 0 J' ] [ p2 ] [ p2 ]. c solve J^Tp2=p2. call dtrsl(wt,m,col,p(col+1),01,info) if (info .ne. 0) return c compute p1=-D^(-1/2)(p1-D^(-1/2)L'p2) c =-D^(-1/2)p1+D^(-1)L'p2. do 40 i = 1, col cc p(i) = -p(i)/sqrt(sy(i,i)) combined with do 30 loop cc into the next line p(i) = -v(i)/sy(i,i) 40 continue do 60 i = 1, col sum = 0.d0 do 50 k = i + 1, col sum = sum + sy(k,i)*p(col+k)/sy(i,i) 50 continue p(i) = p(i) + sum 60 continue return end c======================== The end of bmv =============================== subroutine cauchy(n, x, l, u, nbd, g, iorder, iwhere, t, d, xcp, + m, wy, ws, sy, wt, theta, col, head, p, c, wbp, + v, nint, iprint, sbgnrm, info, epsmch) integer n, m, head, col, nint, iprint, info, + nbd(n), iorder(n), iwhere(n) double precision theta, epsmch, + x(n), l(n), u(n), g(n), t(n), d(n), xcp(n), + wy(n, col), ws(n, col), sy(m, m), + wt(m, m), p(2*m), c(2*m), wbp(2*m), v(2*m) c ************ c c Subroutine cauchy c c For given x, l, u, g (with sbgnrm > 0), and a limited memory c BFGS matrix B defined in terms of matrices WY, WS, WT, and c scalars head, col, and theta, this subroutine computes the c generalized Cauchy point (GCP), defined as the first local c minimizer of the quadratic c c Q(x + s) = g's + 1/2 s'Bs c c along the projected gradient direction P(x-tg,l,u). c The routine returns the GCP in xcp. c c n is an integer variable. c On entry n is the dimension of the problem. c On exit n is unchanged. c c x is a double precision array of dimension n. c On entry x is the starting point for the GCP computation. c On exit x is unchanged. c c l is a double precision array of dimension n. c On entry l is the lower bound of x. c On exit l is unchanged. c c u is a double precision array of dimension n. c On entry u is the upper bound of x. c On exit u is unchanged. c c nbd is an integer array of dimension n. c On entry nbd represents the type of bounds imposed on the c variables, and must be specified as follows: c nbd(i)=0 if x(i) is unbounded, c 1 if x(i) has only a lower bound, c 2 if x(i) has both lower and upper bounds, and c 3 if x(i) has only an upper bound. c On exit nbd is unchanged. c c g is a double precision array of dimension n. c On entry g is the gradient of f(x). g must be a nonzero vector. c On exit g is unchanged. c c iorder is an integer working array of dimension n. c iorder will be used to store the breakpoints in the piecewise c linear path and free variables encountered. On exit, c iorder(1),...,iorder(nleft) are indices of breakpoints c which have not been encountered; c iorder(nleft+1),...,iorder(nbreak) are indices of c encountered breakpoints; and c iorder(nfree),...,iorder(n) are indices of variables which c have no bound constraits along the search direction. c c iwhere is an integer array of dimension n. c On entry iwhere indicates only the permanently fixed (iwhere=3) c or free (iwhere= -1) components of x. c On exit iwhere records the status of the current x variables. c iwhere(i)=-3 if x(i) is free and has bounds, but is not moved c 0 if x(i) is free and has bounds, and is moved c 1 if x(i) is fixed at l(i), and l(i) .ne. u(i) c 2 if x(i) is fixed at u(i), and u(i) .ne. l(i) c 3 if x(i) is always fixed, i.e., u(i)=x(i)=l(i) c -1 if x(i) is always free, i.e., it has no bounds. c c t is a double precision working array of dimension n. c t will be used to store the break points. c c d is a double precision array of dimension n used to store c the Cauchy direction P(x-tg)-x. c c xcp is a double precision array of dimension n used to return the c GCP on exit. c c m is an integer variable. c On entry m is the maximum number of variable metric corrections c used to define the limited memory matrix. c On exit m is unchanged. c c ws, wy, sy, and wt are double precision arrays. c On entry they store information that defines the c limited memory BFGS matrix: c ws(n,m) stores S, a set of s-vectors; c wy(n,m) stores Y, a set of y-vectors; c sy(m,m) stores S'Y; c wt(m,m) stores the c Cholesky factorization of (theta*S'S+LD^(-1)L'). c On exit these arrays are unchanged. c c theta is a double precision variable. c On entry theta is the scaling factor specifying B_0 = theta I. c On exit theta is unchanged. c c col is an integer variable. c On entry col is the actual number of variable metric c corrections stored so far. c On exit col is unchanged. c c head is an integer variable. c On entry head is the location of the first s-vector c (or y-vector) in S (or Y). c On exit col is unchanged. c c p is a double precision working array of dimension 2m. c p will be used to store the vector p = W^(T)d. c c c is a double precision working array of dimension 2m. c c will be used to store the vector c = W^(T)(xcp-x). c c wbp is a double precision working array of dimension 2m. c wbp will be used to store the row of W corresponding c to a breakpoint. c c v is a double precision working array of dimension 2m. c c nint is an integer variable. c On exit nint records the number of quadratic segments explored c in searching for the GCP. c c iprint is an INTEGER variable that must be set by the user. c It controls the frequency and type of output generated: c iprint<0 no output is generated; c iprint=0 print only one line at the last iteration; c 0100 print details of every iteration including x and g; c When iprint > 0, the file iterate.dat will be created to c summarize the iteration. c c sbgnrm is a double precision variable. c On entry sbgnrm is the norm of the projected gradient at x. c On exit sbgnrm is unchanged. c c info is an integer variable. c On entry info is 0. c On exit info = 0 for normal return, c = nonzero for abnormal return when the the system c used in routine bmv is singular. c c Subprograms called: c c L-BFGS-B Library ... hpsolb, bmv. c c Linpack ... dscal dcopy, daxpy. c c c References: c c [1] R. H. Byrd, P. Lu, J. Nocedal and C. Zhu, ``A limited c memory algorithm for bound constrained optimization'', c SIAM J. Scientific Computing 16 (1995), no. 5, pp. 1190--1208. c c [2] C. Zhu, R.H. Byrd, P. Lu, J. Nocedal, ``L-BFGS-B: FORTRAN c Subroutines for Large Scale Bound Constrained Optimization'' c Tech. Report, NAM-11, EECS Department, Northwestern University, c 1994. c c (Postscript files of these papers are available via anonymous c ftp to ece.nwu.edu in the directory pub/lbfgs/lbfgs_bcm.) c c * * * c c NEOS, November 1994. (Latest revision April 1997.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ logical xlower,xupper,bnded integer i,j,col2,nfree,nbreak,pointr, + ibp,nleft,ibkmin,iter double precision f1,f2,dt,dtm,tsum,dibp,zibp,dibp2,bkmin, + tu,tl,wmc,wmp,wmw,ddot,tj,tj0,neggi,sbgnrm, + f2_org double precision one,zero parameter (one=1.0d0,zero=0.0d0) c Check the status of the variables, reset iwhere(i) if necessary; c compute the Cauchy direction d and the breakpoints t; initialize c the derivative f1 and the vector p = W'd (for theta = 1). if (sbgnrm .le. zero) then if (iprint .ge. 0) write (6,*) 'Subgnorm = 0. GCP = X.' call dcopy(n,x,1,xcp,1) return endif bnded = .true. nfree = n + 1 nbreak = 0 ibkmin = 0 bkmin = zero col2 = 2*col f1 = zero if (iprint .ge. 99) write (6,3010) c We set p to zero and build it up as we determine d. do 20 i = 1, col2 p(i) = zero 20 continue c In the following loop we determine for each variable its bound c status and its breakpoint, and update p accordingly. c Smallest breakpoint is identified. do 50 i = 1, n neggi = -g(i) if (iwhere(i) .ne. 3 .and. iwhere(i) .ne. -1) then c if x(i) is not a constant and has bounds, c compute the difference between x(i) and its bounds. if (nbd(i) .le. 2) tl = x(i) - l(i) if (nbd(i) .ge. 2) tu = u(i) - x(i) c If a variable is close enough to a bound c we treat it as at bound. xlower = nbd(i) .le. 2 .and. tl .le. zero xupper = nbd(i) .ge. 2 .and. tu .le. zero c reset iwhere(i). iwhere(i) = 0 if (xlower) then if (neggi .le. zero) iwhere(i) = 1 else if (xupper) then if (neggi .ge. zero) iwhere(i) = 2 else if (abs(neggi) .le. zero) iwhere(i) = -3 endif endif pointr = head if (iwhere(i) .ne. 0 .and. iwhere(i) .ne. -1) then d(i) = zero else d(i) = neggi f1 = f1 - neggi*neggi c calculate p := p - W'e_i* (g_i). do 40 j = 1, col p(j) = p(j) + wy(i,pointr)* neggi p(col + j) = p(col + j) + ws(i,pointr)*neggi pointr = mod(pointr,m) + 1 40 continue if (nbd(i) .le. 2 .and. nbd(i) .ne. 0 + .and. neggi .lt. zero) then c x(i) + d(i) is bounded; compute t(i). nbreak = nbreak + 1 iorder(nbreak) = i t(nbreak) = tl/(-neggi) if (nbreak .eq. 1 .or. t(nbreak) .lt. bkmin) then bkmin = t(nbreak) ibkmin = nbreak endif else if (nbd(i) .ge. 2 .and. neggi .gt. zero) then c x(i) + d(i) is bounded; compute t(i). nbreak = nbreak + 1 iorder(nbreak) = i t(nbreak) = tu/neggi if (nbreak .eq. 1 .or. t(nbreak) .lt. bkmin) then bkmin = t(nbreak) ibkmin = nbreak endif else c x(i) + d(i) is not bounded. nfree = nfree - 1 iorder(nfree) = i if (abs(neggi) .gt. zero) bnded = .false. endif endif 50 continue c The indices of the nonzero components of d are now stored c in iorder(1),...,iorder(nbreak) and iorder(nfree),...,iorder(n). c The smallest of the nbreak breakpoints is in t(ibkmin)=bkmin. if (theta .ne. one) then c complete the initialization of p for theta not= one. call dscal(col,theta,p(col+1),1) endif c Initialize GCP xcp = x. call dcopy(n,x,1,xcp,1) if (nbreak .eq. 0 .and. nfree .eq. n + 1) then c is a zero vector, return with the initial xcp as GCP. if (iprint .gt. 100) write (6,1010) (xcp(i), i = 1, n) return endif c Initialize c = W'(xcp - x) = 0. do 60 j = 1, col2 c(j) = zero 60 continue c Initialize derivative f2. f2 = -theta*f1 f2_org = f2 if (col .gt. 0) then call bmv(m,sy,wt,col,p,v,info) if (info .ne. 0) return f2 = f2 - ddot(col2,v,1,p,1) endif dtm = -f1/f2 tsum = zero nint = 1 if (iprint .ge. 99) + write (6,*) 'There are ',nbreak,' breakpoints ' c If there are no breakpoints, locate the GCP and return. if (nbreak .eq. 0) goto 888 nleft = nbreak iter = 1 tj = zero c------------------- the beginning of the loop ------------------------- 777 continue c Find the next smallest breakpoint; c compute dt = t(nleft) - t(nleft + 1). tj0 = tj if (iter .eq. 1) then c Since we already have the smallest breakpoint we need not do c heapsort yet. Often only one breakpoint is used and the c cost of heapsort is avoided. tj = bkmin ibp = iorder(ibkmin) else if (iter .eq. 2) then c Replace the already used smallest breakpoint with the c breakpoint numbered nbreak > nlast, before heapsort call. if (ibkmin .ne. nbreak) then t(ibkmin) = t(nbreak) iorder(ibkmin) = iorder(nbreak) endif c Update heap structure of breakpoints c (if iter=2, initialize heap). endif call hpsolb(nleft,t,iorder,iter-2) tj = t(nleft) ibp = iorder(nleft) endif dt = tj - tj0 if (dt .ne. zero .and. iprint .ge. 100) then write (6,4011) nint,f1,f2 write (6,5010) dt write (6,6010) dtm endif c If a minimizer is within this interval, c locate the GCP and return. if (dtm .lt. dt) goto 888 c Otherwise fix one variable and c reset the corresponding component of d to zero. tsum = tsum + dt nleft = nleft - 1 iter = iter + 1 dibp = d(ibp) d(ibp) = zero if (dibp .gt. zero) then zibp = u(ibp) - x(ibp) xcp(ibp) = u(ibp) iwhere(ibp) = 2 else zibp = l(ibp) - x(ibp) xcp(ibp) = l(ibp) iwhere(ibp) = 1 endif if (iprint .ge. 100) write (6,*) 'Variable ',ibp,' is fixed.' if (nleft .eq. 0 .and. nbreak .eq. n) then c all n variables are fixed, c return with xcp as GCP. dtm = dt goto 999 endif c Update the derivative information. nint = nint + 1 dibp2 = dibp**2 c Update f1 and f2. c temporarily set f1 and f2 for col=0. f1 = f1 + dt*f2 + dibp2 - theta*dibp*zibp f2 = f2 - theta*dibp2 if (col .gt. 0) then c update c = c + dt*p. call daxpy(col2,dt,p,1,c,1) c choose wbp, c the row of W corresponding to the breakpoint encountered. pointr = head do 70 j = 1,col wbp(j) = wy(ibp,pointr) wbp(col + j) = theta*ws(ibp,pointr) pointr = mod(pointr,m) + 1 70 continue c compute (wbp)Mc, (wbp)Mp, and (wbp)M(wbp)'. call bmv(m,sy,wt,col,wbp,v,info) if (info .ne. 0) return wmc = ddot(col2,c,1,v,1) wmp = ddot(col2,p,1,v,1) wmw = ddot(col2,wbp,1,v,1) c update p = p - dibp*wbp. call daxpy(col2,-dibp,wbp,1,p,1) c complete updating f1 and f2 while col > 0. f1 = f1 + dibp*wmc f2 = f2 + 2.0d0*dibp*wmp - dibp2*wmw endif f2 = max(epsmch*f2_org,f2) if (nleft .gt. 0) then dtm = -f1/f2 goto 777 c to repeat the loop for unsearched intervals. else if(bnded) then f1 = zero f2 = zero dtm = zero else dtm = -f1/f2 endif c------------------- the end of the loop ------------------------------- 888 continue if (iprint .ge. 99) then write (6,*) write (6,*) 'GCP found in this segment' write (6,4010) nint,f1,f2 write (6,6010) dtm endif if (dtm .le. zero) dtm = zero tsum = tsum + dtm c Move free variables (i.e., the ones w/o breakpoints) and c the variables whose breakpoints haven't been reached. call daxpy(n,tsum,d,1,xcp,1) 999 continue c Update c = c + dtm*p = W'(x^c - x) c which will be used in computing r = Z'(B(x^c - x) + g). if (col .gt. 0) call daxpy(col2,dtm,p,1,c,1) if (iprint .gt. 100) write (6,1010) (xcp(i),i = 1,n) if (iprint .ge. 99) write (6,2010) 1010 format ('Cauchy X = ',/,(4x,1p,6(1x,d11.4))) 2010 format (/,'---------------- exit CAUCHY----------------------',/) 3010 format (/,'---------------- CAUCHY entered-------------------') 4010 format ('Piece ',i3,' --f1, f2 at start point ',1p,2(1x,d11.4)) 4011 format (/,'Piece ',i3,' --f1, f2 at start point ', + 1p,2(1x,d11.4)) 5010 format ('Distance to the next break point = ',1p,d11.4) 6010 format ('Distance to the stationary point = ',1p,d11.4) return end c====================== The end of cauchy ============================== subroutine cmprlb(n, m, x, g, ws, wy, sy, wt, z, r, wa, index, + theta, col, head, nfree, cnstnd, info) logical cnstnd integer n, m, col, head, nfree, info, index(n) double precision theta, + x(n), g(n), z(n), r(n), wa(4*m), + ws(n, m), wy(n, m), sy(m, m), wt(m, m) c ************ c c Subroutine cmprlb c c This subroutine computes r=-Z'B(xcp-xk)-Z'g by using c wa(2m+1)=W'(xcp-x) from subroutine cauchy. c c Subprograms called: c c L-BFGS-B Library ... bmv. c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer i,j,k,pointr double precision a1,a2 if (.not. cnstnd .and. col .gt. 0) then do 26 i = 1, n r(i) = -g(i) 26 continue else do 30 i = 1, nfree k = index(i) r(i) = -theta*(z(k) - x(k)) - g(k) 30 continue call bmv(m,sy,wt,col,wa(2*m+1),wa(1),info) if (info .ne. 0) then info = -8 return endif pointr = head do 34 j = 1, col a1 = wa(j) a2 = theta*wa(col + j) do 32 i = 1, nfree k = index(i) r(i) = r(i) + wy(k,pointr)*a1 + ws(k,pointr)*a2 32 continue pointr = mod(pointr,m) + 1 34 continue endif return end c======================= The end of cmprlb ============================= subroutine errclb(n, m, factr, l, u, nbd, task, info, k) character*60 task integer n, m, info, k, nbd(n) double precision factr, l(n), u(n) c ************ c c Subroutine errclb c c This subroutine checks the validity of the input data. c c c * * * c c NEOS, November 1994. (Latest revision April 1997.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer i double precision zero parameter (zero=0.0d0) c Check the input arguments for errors. if (n .le. 0) task = 'ERROR: N .LE. 0' if (m .le. 0) task = 'ERROR: M .LE. 0' if (factr .lt. zero) task = 'ERROR: FACTR .LT. 0' c Check the validity of the arrays nbd(i), u(i), and l(i). do 10 i = 1, n if (nbd(i) .lt. 0 .or. nbd(i) .gt. 3) then c return task = 'ERROR: INVALID NBD' info = -6 k = i endif if (nbd(i) .eq. 2) then if (l(i) .gt. u(i)) then c return task = 'ERROR: NO FEASIBLE SOLUTION' info = -7 k = i endif endif 10 continue return end c======================= The end of errclb ============================= subroutine formk(n, nsub, ind, nenter, ileave, indx2, iupdat, + updatd, wn, wn1, m, ws, wy, sy, theta, col, + head, info) integer n, nsub, m, col, head, nenter, ileave, iupdat, + info, ind(n), indx2(n) double precision theta, wn(2*m, 2*m), wn1(2*m, 2*m), + ws(n, m), wy(n, m), sy(m, m) logical updatd c ************ c c Subroutine formk c c This subroutine forms the LEL^T factorization of the indefinite c c matrix K = [-D -Y'ZZ'Y/theta L_a'-R_z' ] c [L_a -R_z theta*S'AA'S ] c where E = [-I 0] c [ 0 I] c The matrix K can be shown to be equal to the matrix M^[-1]N c occurring in section 5.1 of [1], as well as to the matrix c Mbar^[-1] Nbar in section 5.3. c c n is an integer variable. c On entry n is the dimension of the problem. c On exit n is unchanged. c c nsub is an integer variable c On entry nsub is the number of subspace variables in free set. c On exit nsub is not changed. c c ind is an integer array of dimension nsub. c On entry ind specifies the indices of subspace variables. c On exit ind is unchanged. c c nenter is an integer variable. c On entry nenter is the number of variables entering the c free set. c On exit nenter is unchanged. c c ileave is an integer variable. c On entry indx2(ileave),...,indx2(n) are the variables leaving c the free set. c On exit ileave is unchanged. c c indx2 is an integer array of dimension n. c On entry indx2(1),...,indx2(nenter) are the variables entering c the free set, while indx2(ileave),...,indx2(n) are the c variables leaving the free set. c On exit indx2 is unchanged. c c iupdat is an integer variable. c On entry iupdat is the total number of BFGS updates made so far. c On exit iupdat is unchanged. c c updatd is a logical variable. c On entry 'updatd' is true if the L-BFGS matrix is updatd. c On exit 'updatd' is unchanged. c c wn is a double precision array of dimension 2m x 2m. c On entry wn is unspecified. c On exit the upper triangle of wn stores the LEL^T factorization c of the 2*col x 2*col indefinite matrix c [-D -Y'ZZ'Y/theta L_a'-R_z' ] c [L_a -R_z theta*S'AA'S ] c c wn1 is a double precision array of dimension 2m x 2m. c On entry wn1 stores the lower triangular part of c [Y' ZZ'Y L_a'+R_z'] c [L_a+R_z S'AA'S ] c in the previous iteration. c On exit wn1 stores the corresponding updated matrices. c The purpose of wn1 is just to store these inner products c so they can be easily updated and inserted into wn. c c m is an integer variable. c On entry m is the maximum number of variable metric corrections c used to define the limited memory matrix. c On exit m is unchanged. c c ws, wy, sy, and wtyy are double precision arrays; c theta is a double precision variable; c col is an integer variable; c head is an integer variable. c On entry they store the information defining the c limited memory BFGS matrix: c ws(n,m) stores S, a set of s-vectors; c wy(n,m) stores Y, a set of y-vectors; c sy(m,m) stores S'Y; c wtyy(m,m) stores the Cholesky factorization c of (theta*S'S+LD^(-1)L') c theta is the scaling factor specifying B_0 = theta I; c col is the number of variable metric corrections stored; c head is the location of the 1st s- (or y-) vector in S (or Y). c On exit they are unchanged. c c info is an integer variable. c On entry info is unspecified. c On exit info = 0 for normal return; c = -1 when the 1st Cholesky factorization failed; c = -2 when the 2st Cholesky factorization failed. c c Subprograms called: c c Linpack ... dcopy, dpofa, dtrsl. c c c References: c [1] R. H. Byrd, P. Lu, J. Nocedal and C. Zhu, ``A limited c memory algorithm for bound constrained optimization'', c SIAM J. Scientific Computing 16 (1995), no. 5, pp. 1190--1208. c c [2] C. Zhu, R.H. Byrd, P. Lu, J. Nocedal, ``L-BFGS-B: a c limited memory FORTRAN code for solving bound constrained c optimization problems'', Tech. Report, NAM-11, EECS Department, c Northwestern University, 1994. c c (Postscript files of these papers are available via anonymous c ftp to ece.nwu.edu in the directory pub/lbfgs/lbfgs_bcm.) c c * * * c c NEOS, November 1994. (Latest revision April 1997.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer m2,ipntr,jpntr,iy,is,jy,js,is1,js1,k1,i,k, + col2,pbegin,pend,dbegin,dend,upcl double precision ddot,temp1,temp2,temp3,temp4 double precision zero parameter (zero=0.0d0) c Form the lower triangular part of c WN1 = [Y' ZZ'Y L_a'+R_z'] c [L_a+R_z S'AA'S ] c where L_a is the strictly lower triangular part of S'AA'Y c R_z is the upper triangular part of S'ZZ'Y. if (updatd) then if (iupdat .gt. m) then c shift old part of WN1. do 10 jy = 1, m - 1 js = m + jy call dcopy(m-jy,wn1(jy+1,jy+1),1,wn1(jy,jy),1) call dcopy(m-jy,wn1(js+1,js+1),1,wn1(js,js),1) call dcopy(m-1,wn1(m+2,jy+1),1,wn1(m+1,jy),1) 10 continue endif c put new rows in blocks (1,1), (2,1) and (2,2). pbegin = 1 pend = nsub dbegin = nsub + 1 dend = n iy = col is = m + col ipntr = head + col - 1 if (ipntr .gt. m) ipntr = ipntr - m jpntr = head do 20 jy = 1, col js = m + jy temp1 = zero temp2 = zero temp3 = zero c compute element jy of row 'col' of Y'ZZ'Y do 15 k = pbegin, pend k1 = ind(k) temp1 = temp1 + wy(k1,ipntr)*wy(k1,jpntr) 15 continue c compute elements jy of row 'col' of L_a and S'AA'S do 16 k = dbegin, dend k1 = ind(k) temp2 = temp2 + ws(k1,ipntr)*ws(k1,jpntr) temp3 = temp3 + ws(k1,ipntr)*wy(k1,jpntr) 16 continue wn1(iy,jy) = temp1 wn1(is,js) = temp2 wn1(is,jy) = temp3 jpntr = mod(jpntr,m) + 1 20 continue c put new column in block (2,1). jy = col jpntr = head + col - 1 if (jpntr .gt. m) jpntr = jpntr - m ipntr = head do 30 i = 1, col is = m + i temp3 = zero c compute element i of column 'col' of R_z do 25 k = pbegin, pend k1 = ind(k) temp3 = temp3 + ws(k1,ipntr)*wy(k1,jpntr) 25 continue ipntr = mod(ipntr,m) + 1 wn1(is,jy) = temp3 30 continue upcl = col - 1 else upcl = col endif c modify the old parts in blocks (1,1) and (2,2) due to changes c in the set of free variables. ipntr = head do 45 iy = 1, upcl is = m + iy jpntr = head do 40 jy = 1, iy js = m + jy temp1 = zero temp2 = zero temp3 = zero temp4 = zero do 35 k = 1, nenter k1 = indx2(k) temp1 = temp1 + wy(k1,ipntr)*wy(k1,jpntr) temp2 = temp2 + ws(k1,ipntr)*ws(k1,jpntr) 35 continue do 36 k = ileave, n k1 = indx2(k) temp3 = temp3 + wy(k1,ipntr)*wy(k1,jpntr) temp4 = temp4 + ws(k1,ipntr)*ws(k1,jpntr) 36 continue wn1(iy,jy) = wn1(iy,jy) + temp1 - temp3 wn1(is,js) = wn1(is,js) - temp2 + temp4 jpntr = mod(jpntr,m) + 1 40 continue ipntr = mod(ipntr,m) + 1 45 continue c modify the old parts in block (2,1). ipntr = head do 60 is = m + 1, m + upcl jpntr = head do 55 jy = 1, upcl temp1 = zero temp3 = zero do 50 k = 1, nenter k1 = indx2(k) temp1 = temp1 + ws(k1,ipntr)*wy(k1,jpntr) 50 continue do 51 k = ileave, n k1 = indx2(k) temp3 = temp3 + ws(k1,ipntr)*wy(k1,jpntr) 51 continue if (is .le. jy + m) then wn1(is,jy) = wn1(is,jy) + temp1 - temp3 else wn1(is,jy) = wn1(is,jy) - temp1 + temp3 endif jpntr = mod(jpntr,m) + 1 55 continue ipntr = mod(ipntr,m) + 1 60 continue c Form the upper triangle of WN = [D+Y' ZZ'Y/theta -L_a'+R_z' ] c [-L_a +R_z S'AA'S*theta] m2 = 2*m do 70 iy = 1, col is = col + iy is1 = m + iy do 65 jy = 1, iy js = col + jy js1 = m + jy wn(jy,iy) = wn1(iy,jy)/theta wn(js,is) = wn1(is1,js1)*theta 65 continue do 66 jy = 1, iy - 1 wn(jy,is) = -wn1(is1,jy) 66 continue do 67 jy = iy, col wn(jy,is) = wn1(is1,jy) 67 continue wn(iy,iy) = wn(iy,iy) + sy(iy,iy) 70 continue c Form the upper triangle of c WN= [ LL' L^-1(-L_a'+R_z')] c [(-L_a +R_z)L'^-1 S'AA'S*theta ] c first Cholesky factor (1,1) block of wn to get LL' c with L' stored in the upper triangle of wn. call dpofa(wn,m2,col,info) if (info .ne. 0) then info = -1 return endif c then form L^-1(-L_a'+R_z') in the (1,2) block. col2 = 2*col do 71 js = col+1 ,col2 call dtrsl(wn,m2,col,wn(1,js),11,info) 71 continue c Form S'AA'S*theta + (L^-1(-L_a'+R_z'))'L^-1(-L_a'+R_z') in the c upper triangle of (2,2) block of wn. do 72 is = col+1, col2 do 74 js = is, col2 wn(is,js) = wn(is,js) + ddot(col,wn(1,is),1,wn(1,js),1) 74 continue 72 continue c Cholesky factorization of (2,2) block of wn. call dpofa(wn(col+1,col+1),m2,col,info) if (info .ne. 0) then info = -2 return endif return end c======================= The end of formk ============================== subroutine formt(m, wt, sy, ss, col, theta, info) integer m, col, info double precision theta, wt(m, m), sy(m, m), ss(m, m) c ************ c c Subroutine formt c c This subroutine forms the upper half of the pos. def. and symm. c T = theta*SS + L*D^(-1)*L', stores T in the upper triangle c of the array wt, and performs the Cholesky factorization of T c to produce J*J', with J' stored in the upper triangle of wt. c c Subprograms called: c c Linpack ... dpofa. c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer i,j,k,k1 double precision ddum double precision zero parameter (zero=0.0d0) c Form the upper half of T = theta*SS + L*D^(-1)*L', c store T in the upper triangle of the array wt. do 52 j = 1, col wt(1,j) = theta*ss(1,j) 52 continue do 55 i = 2, col do 54 j = i, col k1 = min(i,j) - 1 ddum = zero do 53 k = 1, k1 ddum = ddum + sy(i,k)*sy(j,k)/sy(k,k) 53 continue wt(i,j) = ddum + theta*ss(i,j) 54 continue 55 continue c Cholesky factorize T to J*J' with c J' stored in the upper triangle of wt. call dpofa(wt,m,col,info) if (info .ne. 0) then info = -3 endif return end c======================= The end of formt ============================== subroutine freev(n, nfree, index, nenter, ileave, indx2, + iwhere, wrk, updatd, cnstnd, iprint, iter) integer n, nfree, nenter, ileave, iprint, iter, + index(n), indx2(n), iwhere(n) logical wrk, updatd, cnstnd c ************ c c Subroutine freev c c This subroutine counts the entering and leaving variables when c iter > 0, and finds the index set of free and active variables c at the GCP. c c cnstnd is a logical variable indicating whether bounds are present c c index is an integer array of dimension n c for i=1,...,nfree, index(i) are the indices of free variables c for i=nfree+1,...,n, index(i) are the indices of bound variables c On entry after the first iteration, index gives c the free variables at the previous iteration. c On exit it gives the free variables based on the determination c in cauchy using the array iwhere. c c indx2 is an integer array of dimension n c On entry indx2 is unspecified. c On exit with iter>0, indx2 indicates which variables c have changed status since the previous iteration. c For i= 1,...,nenter, indx2(i) have changed from bound to free. c For i= ileave+1,...,n, indx2(i) have changed from free to bound. c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer iact,i,k nenter = 0 ileave = n + 1 if (iter .gt. 0 .and. cnstnd) then c count the entering and leaving variables. do 20 i = 1, nfree k = index(i) if (iwhere(k) .gt. 0) then ileave = ileave - 1 indx2(ileave) = k if (iprint .ge. 100) write (6,*) + 'Variable ',k,' leaves the set of free variables' endif 20 continue do 22 i = 1 + nfree, n k = index(i) if (iwhere(k) .le. 0) then nenter = nenter + 1 indx2(nenter) = k if (iprint .ge. 100) write (6,*) + 'Variable ',k,' enters the set of free variables' endif 22 continue if (iprint .ge. 99) write (6,*) + n+1-ileave,' variables leave; ',nenter,' variables enter' endif wrk = (ileave .lt. n+1) .or. (nenter .gt. 0) .or. updatd c Find the index set of free and active variables at the GCP. nfree = 0 iact = n + 1 do 24 i = 1, n if (iwhere(i) .le. 0) then nfree = nfree + 1 index(nfree) = i else iact = iact - 1 index(iact) = i endif 24 continue if (iprint .ge. 99) write (6,*) + nfree,' variables are free at GCP ',iter + 1 return end c======================= The end of freev ============================== subroutine hpsolb(n, t, iorder, iheap) integer iheap, n, iorder(n) double precision t(n) c ************ c c Subroutine hpsolb c c This subroutine sorts out the least element of t, and puts the c remaining elements of t in a heap. c c n is an integer variable. c On entry n is the dimension of the arrays t and iorder. c On exit n is unchanged. c c t is a double precision array of dimension n. c On entry t stores the elements to be sorted, c On exit t(n) stores the least elements of t, and t(1) to t(n-1) c stores the remaining elements in the form of a heap. c c iorder is an integer array of dimension n. c On entry iorder(i) is the index of t(i). c On exit iorder(i) is still the index of t(i), but iorder may be c permuted in accordance with t. c c iheap is an integer variable specifying the task. c On entry iheap should be set as follows: c iheap .eq. 0 if t(1) to t(n) is not in the form of a heap, c iheap .ne. 0 if otherwise. c On exit iheap is unchanged. c c c References: c Algorithm 232 of CACM (J. W. J. Williams): HEAPSORT. c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c ************ integer i,j,k,indxin,indxou double precision ddum,out if (iheap .eq. 0) then c Rearrange the elements t(1) to t(n) to form a heap. do 20 k = 2, n ddum = t(k) indxin = iorder(k) c Add ddum to the heap. i = k 10 continue if (i.gt.1) then j = i/2 if (ddum .lt. t(j)) then t(i) = t(j) iorder(i) = iorder(j) i = j goto 10 endif endif t(i) = ddum iorder(i) = indxin 20 continue endif c Assign to 'out' the value of t(1), the least member of the heap, c and rearrange the remaining members to form a heap as c elements 1 to n-1 of t. if (n .gt. 1) then i = 1 out = t(1) indxou = iorder(1) ddum = t(n) indxin = iorder(n) c Restore the heap 30 continue j = i+i if (j .le. n-1) then if (t(j+1) .lt. t(j)) j = j+1 if (t(j) .lt. ddum ) then t(i) = t(j) iorder(i) = iorder(j) i = j goto 30 endif endif t(i) = ddum iorder(i) = indxin c Put the least member in t(n). t(n) = out iorder(n) = indxou endif return end c====================== The end of hpsolb ============================== subroutine lnsrlb(n, l, u, nbd, x, f, fold, gd, gdold, g, d, r, t, + z, stp, dnorm, dtd, xstep, stpmx, iter, ifun, + iback, nfgv, info, task, boxed, cnstnd, csave, + isave, dsave) character*60 task, csave logical boxed, cnstnd integer n, iter, ifun, iback, nfgv, info, + nbd(n), isave(2) double precision f, fold, gd, gdold, stp, dnorm, dtd, xstep, + stpmx, x(n), l(n), u(n), g(n), d(n), r(n), t(n), + z(n), dsave(13) c ********** c c Subroutine lnsrlb c c This subroutine calls subroutine dcsrch from the Minpack2 library c to perform the line search. Subroutine dscrch is safeguarded so c that all trial points lie within the feasible region. c c Subprograms called: c c Minpack2 Library ... dcsrch. c c Linpack ... dtrsl, ddot. c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ********** integer i double precision ddot,a1,a2 double precision one,zero,big parameter (one=1.0d0,zero=0.0d0,big=1.0d+10) double precision ftol,gtol,xtol parameter (ftol=1.0d-3,gtol=0.9d0,xtol=0.1d0) if (task(1:5) .eq. 'FG_LN') goto 556 dtd = ddot(n,d,1,d,1) dnorm = sqrt(dtd) c Determine the maximum step length. stpmx = big if (cnstnd) then if (iter .eq. 0) then stpmx = one else do 43 i = 1, n a1 = d(i) if (nbd(i) .ne. 0) then if (a1 .lt. zero .and. nbd(i) .le. 2) then a2 = l(i) - x(i) if (a2 .ge. zero) then stpmx = zero else if (a1*stpmx .lt. a2) then stpmx = a2/a1 endif else if (a1 .gt. zero .and. nbd(i) .ge. 2) then a2 = u(i) - x(i) if (a2 .le. zero) then stpmx = zero else if (a1*stpmx .gt. a2) then stpmx = a2/a1 endif endif endif 43 continue endif endif if (iter .eq. 0 .and. .not. boxed) then stp = min(one/dnorm, stpmx) else stp = one endif call dcopy(n,x,1,t,1) call dcopy(n,g,1,r,1) fold = f ifun = 0 iback = 0 csave = 'START' 556 continue gd = ddot(n,g,1,d,1) if (ifun .eq. 0) then gdold=gd if (gd .ge. zero) then c the directional derivative >=0. c Line search is impossible. info = -4 return endif endif call dcsrch(f,gd,stp,ftol,gtol,xtol,zero,stpmx,csave,isave,dsave) xstep = stp*dnorm if (csave(1:4) .ne. 'CONV' .and. csave(1:4) .ne. 'WARN') then task = 'FG_LNSRCH' ifun = ifun + 1 nfgv = nfgv + 1 iback = ifun - 1 if (stp .eq. one) then call dcopy(n,z,1,x,1) else do 41 i = 1, n x(i) = stp*d(i) + t(i) 41 continue endif else task = 'NEW_X' endif return end c======================= The end of lnsrlb ============================= subroutine matupd(n, m, ws, wy, sy, ss, d, r, itail, + iupdat, col, head, theta, rr, dr, stp, dtd) integer n, m, itail, iupdat, col, head double precision theta, rr, dr, stp, dtd, d(n), r(n), + ws(n, m), wy(n, m), sy(m, m), ss(m, m) c ************ c c Subroutine matupd c c This subroutine updates matrices WS and WY, and forms the c middle matrix in B. c c Subprograms called: c c Linpack ... dcopy, ddot. c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer j,pointr double precision ddot double precision one parameter (one=1.0d0) c Set pointers for matrices WS and WY. if (iupdat .le. m) then col = iupdat itail = mod(head+iupdat-2,m) + 1 else itail = mod(itail,m) + 1 head = mod(head,m) + 1 endif c Update matrices WS and WY. call dcopy(n,d,1,ws(1,itail),1) call dcopy(n,r,1,wy(1,itail),1) c Set theta=yy/ys. theta = rr/dr c Form the middle matrix in B. c update the upper triangle of SS, c and the lower triangle of SY: if (iupdat .gt. m) then c move old information do 50 j = 1, col - 1 call dcopy(j,ss(2,j+1),1,ss(1,j),1) call dcopy(col-j,sy(j+1,j+1),1,sy(j,j),1) 50 continue endif c add new information: the last row of SY c and the last column of SS: pointr = head do 51 j = 1, col - 1 sy(col,j) = ddot(n,d,1,wy(1,pointr),1) ss(j,col) = ddot(n,ws(1,pointr),1,d,1) pointr = mod(pointr,m) + 1 51 continue if (stp .eq. one) then ss(col,col) = dtd else ss(col,col) = stp*stp*dtd endif sy(col,col) = dr return end c======================= The end of matupd ============================= subroutine prn1lb(n, m, l, u, x, iprint, itfile, epsmch) integer n, m, iprint, itfile double precision epsmch, x(n), l(n), u(n) c ************ c c Subroutine prn1lb c c This subroutine prints the input data, initial point, upper and c lower bounds of each variable, machine precision, as well as c the headings of the output. c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer i if (iprint .ge. 0) then write (6,7001) epsmch write (6,*) 'N = ',n,' M = ',m if (iprint .ge. 1) then write (itfile,2001) epsmch write (itfile,*)'N = ',n,' M = ',m write (itfile,9001) if (iprint .gt. 100) then write (6,1004) 'L =',(l(i),i = 1,n) write (6,1004) 'X0 =',(x(i),i = 1,n) write (6,1004) 'U =',(u(i),i = 1,n) endif endif endif 1004 format (/,a4, 1p, 6(1x,d11.4),/,(4x,1p,6(1x,d11.4))) 2001 format ('RUNNING THE L-BFGS-B CODE',/,/, + 'it = iteration number',/, + 'nf = number of function evaluations',/, + 'nint = number of segments explored during the Cauchy search',/, + 'nact = number of active bounds at the generalized Cauchy point' + ,/, + 'sub = manner in which the subspace minimization terminated:' + ,/,' con = converged, bnd = a bound was reached',/, + 'itls = number of iterations performed in the line search',/, + 'stepl = step length used',/, + 'tstep = norm of the displacement (total step)',/, + 'projg = norm of the projected gradient',/, + 'f = function value',/,/, + ' * * *',/,/, + 'Machine precision =',1p,d10.3) 7001 format ('RUNNING THE L-BFGS-B CODE',/,/, + ' * * *',/,/, + 'Machine precision =',1p,d10.3) 9001 format (/,3x,'it',3x,'nf',2x,'nint',2x,'nact',2x,'sub',2x,'itls', + 2x,'stepl',4x,'tstep',5x,'projg',8x,'f') return end c======================= The end of prn1lb ============================= subroutine prn2lb(n, x, f, g, iprint, itfile, iter, nfgv, nact, + sbgnrm, nint, word, iword, iback, stp, xstep) character*3 word integer n, iprint, itfile, iter, nfgv, nact, nint, + iword, iback double precision f, sbgnrm, stp, xstep, x(n), g(n) c ************ c c Subroutine prn2lb c c This subroutine prints out new information after a successful c line search. c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer i,imod c 'word' records the status of subspace solutions. if (iword .eq. 0) then c the subspace minimization converged. word = 'con' else if (iword .eq. 1) then c the subspace minimization stopped at a bound. word = 'bnd' else if (iword .eq. 5) then c the truncated Newton step has been used. word = 'TNT' else word = '---' endif if (iprint .ge. 99) then write (6,*) 'LINE SEARCH',iback,' times; norm of step = ',xstep write (6,2001) iter,f,sbgnrm if (iprint .gt. 100) then write (6,1004) 'X =',(x(i), i = 1, n) write (6,1004) 'G =',(g(i), i = 1, n) endif else if (iprint .gt. 0) then imod = mod(iter,iprint) if (imod .eq. 0) write (6,2001) iter,f,sbgnrm endif if (iprint .ge. 1) write (itfile,3001) + iter,nfgv,nint,nact,word,iback,stp,xstep,sbgnrm,f 1004 format (/,a4, 1p, 6(1x,d11.4),/,(4x,1p,6(1x,d11.4))) 2001 format + (/,'At iterate',i5,4x,'f= ',1p,d12.5,4x,'|proj g|= ',1p,d12.5) 3001 format(2(1x,i4),2(1x,i5),2x,a3,1x,i4,1p,2(2x,d7.1),1p,2(1x,d10.3)) return end c======================= The end of prn2lb ============================= subroutine prn3lb(n, x, f, task, iprint, info, itfile, + iter, nfgv, nintol, nskip, nact, sbgnrm, + time, nint, word, iback, stp, xstep, k, + cachyt, sbtime, lnscht) character*60 task character*3 word integer n, iprint, info, itfile, iter, nfgv, nintol, + nskip, nact, nint, iback, k double precision f, sbgnrm, time, stp, xstep, cachyt, sbtime, + lnscht, x(n) c ************ c c Subroutine prn3lb c c This subroutine prints out information when either a built-in c convergence test is satisfied or when an error message is c generated. c c c * * * c c NEOS, November 1994. (Latest revision April 1997.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer i if (task(1:5) .eq. 'ERROR') goto 999 if (iprint .ge. 0) then write (6,3003) write (6,3004) write(6,3005) n,iter,nfgv,nintol,nskip,nact,sbgnrm,f if (iprint .ge. 100) then write (6,1004) 'X =',(x(i),i = 1,n) endif if (iprint .ge. 1) write (6,*) ' F =',f endif 999 continue if (iprint .ge. 0) then write (6,3009) task if (info .ne. 0) then if (info .eq. -1) write (6,9011) if (info .eq. -2) write (6,9012) if (info .eq. -3) write (6,9013) if (info .eq. -4) write (6,9014) if (info .eq. -5) write (6,9015) if (info .eq. -6) write (6,*)' Input nbd(',k,') is invalid.' if (info .eq. -7) + write (6,*)' l(',k,') > u(',k,'). No feasible solution.' if (info .eq. -8) write (6,9018) if (info .eq. -9) write (6,9019) endif if (iprint .ge. 1) write (6,3007) cachyt,sbtime,lnscht write (6,3008) time if (iprint .ge. 1) then if (info .eq. -4 .or. info .eq. -9) then write (itfile,3002) + iter,nfgv,nint,nact,word,iback,stp,xstep endif write (itfile,3009) task if (info .ne. 0) then if (info .eq. -1) write (itfile,9011) if (info .eq. -2) write (itfile,9012) if (info .eq. -3) write (itfile,9013) if (info .eq. -4) write (itfile,9014) if (info .eq. -5) write (itfile,9015) if (info .eq. -8) write (itfile,9018) if (info .eq. -9) write (itfile,9019) endif write (itfile,3008) time endif endif 1004 format (/,a4, 1p, 6(1x,d11.4),/,(4x,1p,6(1x,d11.4))) 3002 format(2(1x,i4),2(1x,i5),2x,a3,1x,i4,1p,2(2x,d7.1),6x,'-',10x,'-') 3003 format (/, + ' * * *',/,/, + 'Tit = total number of iterations',/, + 'Tnf = total number of function evaluations',/, + 'Tnint = total number of segments explored during', + ' Cauchy searches',/, + 'Skip = number of BFGS updates skipped',/, + 'Nact = number of active bounds at final generalized', + ' Cauchy point',/, + 'Projg = norm of the final projected gradient',/, + 'F = final function value',/,/, + ' * * *') 3004 format (/,3x,'N',3x,'Tit',2x,'Tnf',2x,'Tnint',2x, + 'Skip',2x,'Nact',5x,'Projg',8x,'F') 3005 format (i5,2(1x,i4),(1x,i6),(2x,i4),(1x,i5),1p,2(2x,d10.3)) 3007 format (/,' Cauchy time',1p,e10.3,' seconds.',/ + ' Subspace minimization time',1p,e10.3,' seconds.',/ + ' Line search time',1p,e10.3,' seconds.') 3008 format (/,' Total User time',1p,e10.3,' seconds.',/) 3009 format (/,a60) 9011 format (/, +' Matrix in 1st Cholesky factorization in formk is not Pos. Def.') 9012 format (/, +' Matrix in 2st Cholesky factorization in formk is not Pos. Def.') 9013 format (/, +' Matrix in the Cholesky factorization in formt is not Pos. Def.') 9014 format (/, +' Derivative >= 0, backtracking line search impossible.',/, +' Previous x, f and g restored.',/, +' Possible causes: 1 error in function or gradient evaluation;',/, +' 2 rounding errors dominate computation.') 9015 format (/, +' Warning: more than 10 function and gradient',/, +' evaluations in the last line search. Termination',/, +' may possibly be caused by a bad search direction.') 9018 format (/,' The triangular system is singular.') 9019 format (/, +' Line search cannot locate an adequate point after 20 function',/ +,' and gradient evaluations. Previous x, f and g restored.',/, +' Possible causes: 1 error in function or gradient evaluation;',/, +' 2 rounding error dominate computation.') return end c======================= The end of prn3lb ============================= subroutine projgr(n, l, u, nbd, x, g, sbgnrm) integer n, nbd(n) double precision sbgnrm, x(n), l(n), u(n), g(n) c ************ c c Subroutine projgr c c This subroutine computes the infinity norm of the projected c gradient. c c c * * * c c NEOS, November 1994. (Latest revision April 1997.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ integer i double precision gi double precision zero parameter (zero=0.0d0) sbgnrm = zero do 15 i = 1, n gi = g(i) if (nbd(i) .ne. 0) then if (gi .lt. zero) then if (nbd(i) .ge. 2) gi = max((x(i)-u(i)),gi) else if (nbd(i) .le. 2) gi = min((x(i)-l(i)),gi) endif endif sbgnrm = max(sbgnrm,abs(gi)) 15 continue return end c======================= The end of projgr ============================= subroutine subsm(n, m, nsub, ind, l, u, nbd, x, d, ws, wy, theta, + col, head, iword, wv, wn, iprint, info) integer n, m, nsub, col, head, iword, iprint, info, + ind(nsub), nbd(n) double precision theta, + l(n), u(n), x(n), d(n), + ws(n, m), wy(n, m), + wv(2*m), wn(2*m, 2*m) c ************ c c Subroutine subsm c c Given xcp, l, u, r, an index set that specifies c the active set at xcp, and an l-BFGS matrix B c (in terms of WY, WS, SY, WT, head, col, and theta), c this subroutine computes an approximate solution c of the subspace problem c c (P) min Q(x) = r'(x-xcp) + 1/2 (x-xcp)' B (x-xcp) c c subject to l<=x<=u c x_i=xcp_i for all i in A(xcp) c c along the subspace unconstrained Newton direction c c d = -(Z'BZ)^(-1) r. c c The formula for the Newton direction, given the L-BFGS matrix c and the Sherman-Morrison formula, is c c d = (1/theta)r + (1/theta*2) Z'WK^(-1)W'Z r. c c where c K = [-D -Y'ZZ'Y/theta L_a'-R_z' ] c [L_a -R_z theta*S'AA'S ] c c Note that this procedure for computing d differs c from that described in [1]. One can show that the matrix K is c equal to the matrix M^[-1]N in that paper. c c n is an integer variable. c On entry n is the dimension of the problem. c On exit n is unchanged. c c m is an integer variable. c On entry m is the maximum number of variable metric corrections c used to define the limited memory matrix. c On exit m is unchanged. c c nsub is an integer variable. c On entry nsub is the number of free variables. c On exit nsub is unchanged. c c ind is an integer array of dimension nsub. c On entry ind specifies the coordinate indices of free variables. c On exit ind is unchanged. c c l is a double precision array of dimension n. c On entry l is the lower bound of x. c On exit l is unchanged. c c u is a double precision array of dimension n. c On entry u is the upper bound of x. c On exit u is unchanged. c c nbd is a integer array of dimension n. c On entry nbd represents the type of bounds imposed on the c variables, and must be specified as follows: c nbd(i)=0 if x(i) is unbounded, c 1 if x(i) has only a lower bound, c 2 if x(i) has both lower and upper bounds, and c 3 if x(i) has only an upper bound. c On exit nbd is unchanged. c c x is a double precision array of dimension n. c On entry x specifies the Cauchy point xcp. c On exit x(i) is the minimizer of Q over the subspace of c free variables. c c d is a double precision array of dimension n. c On entry d is the reduced gradient of Q at xcp. c On exit d is the Newton direction of Q. c c ws and wy are double precision arrays; c theta is a double precision variable; c col is an integer variable; c head is an integer variable. c On entry they store the information defining the c limited memory BFGS matrix: c ws(n,m) stores S, a set of s-vectors; c wy(n,m) stores Y, a set of y-vectors; c theta is the scaling factor specifying B_0 = theta I; c col is the number of variable metric corrections stored; c head is the location of the 1st s- (or y-) vector in S (or Y). c On exit they are unchanged. c c iword is an integer variable. c On entry iword is unspecified. c On exit iword specifies the status of the subspace solution. c iword = 0 if the solution is in the box, c 1 if some bound is encountered. c c wv is a double precision working array of dimension 2m. c c wn is a double precision array of dimension 2m x 2m. c On entry the upper triangle of wn stores the LEL^T factorization c of the indefinite matrix c c K = [-D -Y'ZZ'Y/theta L_a'-R_z' ] c [L_a -R_z theta*S'AA'S ] c where E = [-I 0] c [ 0 I] c On exit wn is unchanged. c c iprint is an INTEGER variable that must be set by the user. c It controls the frequency and type of output generated: c iprint<0 no output is generated; c iprint=0 print only one line at the last iteration; c 0100 print details of every iteration including x and g; c When iprint > 0, the file iterate.dat will be created to c summarize the iteration. c c info is an integer variable. c On entry info is unspecified. c On exit info = 0 for normal return, c = nonzero for abnormal return c when the matrix K is ill-conditioned. c c Subprograms called: c c Linpack dtrsl. c c c References: c c [1] R. H. Byrd, P. Lu, J. Nocedal and C. Zhu, ``A limited c memory algorithm for bound constrained optimization'', c SIAM J. Scientific Computing 16 (1995), no. 5, pp. 1190--1208. c c c c * * * c c NEOS, November 1994. (Latest revision June 1996.) c Optimization Technology Center. c Argonne National Laboratory and Northwestern University. c Written by c Ciyou Zhu c in collaboration with R.H. Byrd, P. Lu-Chen and J. Nocedal. c c c ************ logical temp1_updated integer pointr,m2,col2,ibd,jy,js,i,j,k double precision alpha,dk,temp1,temp2 double precision one,zero parameter (one=1.0d0,zero=0.0d0) if (nsub .le. 0) return if (iprint .ge. 99) write (6,1001) c Compute wv = W'Zd. pointr = head do 20 i = 1, col temp1 = zero temp2 = zero do 10 j = 1, nsub k = ind(j) temp1 = temp1 + wy(k,pointr)*d(j) temp2 = temp2 + ws(k,pointr)*d(j) 10 continue wv(i) = temp1 wv(col + i) = theta*temp2 pointr = mod(pointr,m) + 1 20 continue c Compute wv:=K^(-1)wv. m2 = 2*m col2 = 2*col call dtrsl(wn,m2,col2,wv,11,info) if (info .ne. 0) return do 25 i = 1, col wv(i) = -wv(i) 25 continue call dtrsl(wn,m2,col2,wv,01,info) if (info .ne. 0) return c Compute d = (1/theta)d + (1/theta**2)Z'W wv. pointr = head do 40 jy = 1, col js = col + jy do 30 i = 1, nsub k = ind(i) d(i) = d(i) + wy(k,pointr)*wv(jy)/theta + + ws(k,pointr)*wv(js) 30 continue pointr = mod(pointr,m) + 1 40 continue do 50 i = 1, nsub d(i) = d(i)/theta 50 continue c Backtrack to the feasible region. alpha = one temp1 = alpha do 60 i = 1, nsub k = ind(i) dk = d(i) if (nbd(k) .ne. 0) then temp1_updated = .false. if (dk .lt. zero .and. nbd(k) .le. 2) then temp2 = l(k) - x(k) if (temp2 .ge. zero) then temp1 = zero temp1_updated = .true. else if (dk*alpha .lt. temp2) then temp1 = temp2/dk temp1_updated = .true. endif else if (dk .gt. zero .and. nbd(k) .ge. 2) then temp2 = u(k) - x(k) if (temp2 .le. zero) then temp1 = zero temp1_updated = .true. else if (dk*alpha .gt. temp2) then temp1 = temp2/dk temp1_updated = .true. endif endif cc logical variable temp1_updated added to eliminate unexpected cc trigger of the if statement due to possible difference between cc hardware precision and double precision. if (temp1_updated .and. temp1 .lt. alpha) then alpha = temp1 ibd = i endif endif 60 continue if (alpha .lt. one) then dk = d(ibd) k = ind(ibd) if (dk .gt. zero) then x(k) = u(k) d(ibd) = zero else if (dk .lt. zero) then x(k) = l(k) d(ibd) = zero endif endif do 70 i = 1, nsub k = ind(i) x(k) = x(k) + alpha*d(i) 70 continue if (iprint .ge. 99) then if (alpha .lt. one) then write (6,1002) alpha else write (6,*) 'SM solution inside the box' end if if (iprint .gt.100) write (6,1003) (x(i),i=1,n) endif if (alpha .lt. one) then iword = 1 else iword = 0 endif if (iprint .ge. 99) write (6,1004) 1001 format (/,'----------------SUBSM entered-----------------',/) 1002 format ( 'ALPHA = ',f7.5,' backtrack to the BOX') 1003 format ('Subspace solution X = ',/,(4x,1p,6(1x,d11.4))) 1004 format (/,'----------------exit SUBSM --------------------',/) return end c====================== The end of subsm =============================== subroutine dcsrch(f,g,stp,ftol,gtol,xtol,stpmin,stpmax, + task,isave,dsave) character*(*) task integer isave(2) double precision f,g,stp,ftol,gtol,xtol,stpmin,stpmax double precision dsave(13) c ********** c c Subroutine dcsrch c c This subroutine finds a step that satisfies a sufficient c decrease condition and a curvature condition. c c Each call of the subroutine updates an interval with c endpoints stx and sty. The interval is initially chosen c so that it contains a minimizer of the modified function c c psi(stp) = f(stp) - f(0) - ftol*stp*f'(0). c c If psi(stp) <= 0 and f'(stp) >= 0 for some step, then the c interval is chosen so that it contains a minimizer of f. c c The algorithm is designed to find a step that satisfies c the sufficient decrease condition c c f(stp) <= f(0) + ftol*stp*f'(0), c c and the curvature condition c c abs(f'(stp)) <= gtol*abs(f'(0)). c c If ftol is less than gtol and if, for example, the function c is bounded below, then there is always a step which satisfies c both conditions. c c If no step can be found that satisfies both conditions, then c the algorithm stops with a warning. In this case stp only c satisfies the sufficient decrease condition. c c A typical invocation of dcsrch has the following outline: c c task = 'START' c 10 continue c call dcsrch( ... ) c if (task .eq. 'FG') then c Evaluate the function and the gradient at stp c goto 10 c end if c c NOTE: The user must no alter work arrays between calls. c c The subroutine statement is c c subroutine dcsrch(f,g,stp,ftol,gtol,xtol,stpmin,stpmax, c task,isave,dsave) c where c c f is a double precision variable. c On initial entry f is the value of the function at 0. c On subsequent entries f is the value of the c function at stp. c On exit f is the value of the function at stp. c c g is a double precision variable. c On initial entry g is the derivative of the function at 0. c On subsequent entries g is the derivative of the c function at stp. c On exit g is the derivative of the function at stp. c c stp is a double precision variable. c On entry stp is the current estimate of a satisfactory c step. On initial entry, a positive initial estimate c must be provided. c On exit stp is the current estimate of a satisfactory step c if task = 'FG'. If task = 'CONV' then stp satisfies c the sufficient decrease and curvature condition. c c ftol is a double precision variable. c On entry ftol specifies a nonnegative tolerance for the c sufficient decrease condition. c On exit ftol is unchanged. c c gtol is a double precision variable. c On entry gtol specifies a nonnegative tolerance for the c curvature condition. c On exit gtol is unchanged. c c xtol is a double precision variable. c On entry xtol specifies a nonnegative relative tolerance c for an acceptable step. The subroutine exits with a c warning if the relative difference between sty and stx c is less than xtol. c On exit xtol is unchanged. c c stpmin is a double precision variable. c On entry stpmin is a nonnegative lower bound for the step. c On exit stpmin is unchanged. c c stpmax is a double precision variable. c On entry stpmax is a nonnegative upper bound for the step. c On exit stpmax is unchanged. c c task is a character variable of length at least 60. c On initial entry task must be set to 'START'. c On exit task indicates the required action: c c If task(1:2) = 'FG' then evaluate the function and c derivative at stp and call dcsrch again. c c If task(1:4) = 'CONV' then the search is successful. c c If task(1:4) = 'WARN' then the subroutine is not able c to satisfy the convergence conditions. The exit value of c stp contains the best point found during the search. c c If task(1:5) = 'ERROR' then there is an error in the c input arguments. c c On exit with convergence, a warning or an error, the c variable task contains additional information. c c isave is an integer work array of dimension 2. c c dsave is a double precision work array of dimension 13. c c Subprograms called c c MINPACK-2 ... dcstep c c MINPACK-1 Project. June 1983. c Argonne National Laboratory. c Jorge J. More' and David J. Thuente. c c MINPACK-2 Project. October 1993. c Argonne National Laboratory and University of Minnesota. c Brett M. Averick, Richard G. Carter, and Jorge J. More'. c c ********** double precision zero,p5,p66 parameter(zero=0.0d0,p5=0.5d0,p66=0.66d0) double precision xtrapl,xtrapu parameter(xtrapl=1.1d0,xtrapu=4.0d0) logical brackt integer stage double precision finit,ftest,fm,fx,fxm,fy,fym,ginit,gtest, + gm,gx,gxm,gy,gym,stx,sty,stmin,stmax,width,width1 c Initialization block. if (task(1:5) .eq. 'START') then c Check the input arguments for errors. if (stp .lt. stpmin) task = 'ERROR: STP .LT. STPMIN' if (stp .gt. stpmax) task = 'ERROR: STP .GT. STPMAX' if (g .ge. zero) task = 'ERROR: INITIAL G .GE. ZERO' if (ftol .lt. zero) task = 'ERROR: FTOL .LT. ZERO' if (gtol .lt. zero) task = 'ERROR: GTOL .LT. ZERO' if (xtol .lt. zero) task = 'ERROR: XTOL .LT. ZERO' if (stpmin .lt. zero) task = 'ERROR: STPMIN .LT. ZERO' if (stpmax .lt. stpmin) task = 'ERROR: STPMAX .LT. STPMIN' c Exit if there are errors on input. if (task(1:5) .eq. 'ERROR') return c Initialize local variables. brackt = .false. stage = 1 finit = f ginit = g gtest = ftol*ginit width = stpmax - stpmin width1 = width/p5 c The variables stx, fx, gx contain the values of the step, c function, and derivative at the best step. c The variables sty, fy, gy contain the value of the step, c function, and derivative at sty. c The variables stp, f, g contain the values of the step, c function, and derivative at stp. stx = zero fx = finit gx = ginit sty = zero fy = finit gy = ginit stmin = zero stmax = stp + xtrapu*stp task = 'FG' goto 1000 else c Restore local variables. if (isave(1) .eq. 1) then brackt = .true. else brackt = .false. endif stage = isave(2) ginit = dsave(1) gtest = dsave(2) gx = dsave(3) gy = dsave(4) finit = dsave(5) fx = dsave(6) fy = dsave(7) stx = dsave(8) sty = dsave(9) stmin = dsave(10) stmax = dsave(11) width = dsave(12) width1 = dsave(13) endif c If psi(stp) <= 0 and f'(stp) >= 0 for some step, then the c algorithm enters the second stage. ftest = finit + stp*gtest if (stage .eq. 1 .and. f .le. ftest .and. g .ge. zero) + stage = 2 c Test for warnings. if (brackt .and. (stp .le. stmin .or. stp .ge. stmax)) + task = 'WARNING: ROUNDING ERRORS PREVENT PROGRESS' if (brackt .and. stmax - stmin .le. xtol*stmax) + task = 'WARNING: XTOL TEST SATISFIED' if (stp .eq. stpmax .and. f .le. ftest .and. g .le. gtest) + task = 'WARNING: STP = STPMAX' if (stp .eq. stpmin .and. (f .gt. ftest .or. g .ge. gtest)) + task = 'WARNING: STP = STPMIN' cc New warning statement added to eliminate the unexpected case cc of stp=stx due to possible difference between hardware precision cc and double precision. if (stp .eq. stx) + task = 'WARNING: STP = STX' c Test for convergence. if (f .le. ftest .and. abs(g) .le. gtol*(-ginit)) + task = 'CONVERGENCE' c Test for termination. if (task(1:4) .eq. 'WARN' .or. task(1:4) .eq. 'CONV') goto 1000 c A modified function is used to predict the step during the c first stage if a lower function value has been obtained but c the decrease is not sufficient. if (stage .eq. 1 .and. f .le. fx .and. f .gt. ftest) then c Define the modified function and derivative values. fm = f - stp*gtest fxm = fx - stx*gtest fym = fy - sty*gtest gm = g - gtest gxm = gx - gtest gym = gy - gtest c Call dcstep to update stx, sty, and to compute the new step. call dcstep(stx,fxm,gxm,sty,fym,gym,stp,fm,gm, + brackt,stmin,stmax) c Reset the function and derivative values for f. fx = fxm + stx*gtest fy = fym + sty*gtest gx = gxm + gtest gy = gym + gtest else c Call dcstep to update stx, sty, and to compute the new step. call dcstep(stx,fx,gx,sty,fy,gy,stp,f,g, + brackt,stmin,stmax) endif c Decide if a bisection step is needed. if (brackt) then if (abs(sty-stx) .ge. p66*width1) stp = stx + p5*(sty - stx) width1 = width width = abs(sty-stx) endif c Set the minimum and maximum steps allowed for stp. if (brackt) then stmin = min(stx,sty) stmax = max(stx,sty) else stmin = stp + xtrapl*(stp - stx) stmax = stp + xtrapu*(stp - stx) endif c Force the step to be within the bounds stpmax and stpmin. stp = max(stp,stpmin) stp = min(stp,stpmax) c If further progress is not possible, let stp be the best c point obtained during the search. if (brackt .and. (stp .le. stmin .or. stp .ge. stmax) + .or. (brackt .and. stmax-stmin .le. xtol*stmax)) stp = stx c Obtain another function and derivative. task = 'FG' 1000 continue c Save local variables. if (brackt) then isave(1) = 1 else isave(1) = 0 endif isave(2) = stage dsave(1) = ginit dsave(2) = gtest dsave(3) = gx dsave(4) = gy dsave(5) = finit dsave(6) = fx dsave(7) = fy dsave(8) = stx dsave(9) = sty dsave(10) = stmin dsave(11) = stmax dsave(12) = width dsave(13) = width1 end c====================== The end of dcsrch ============================== subroutine dcstep(stx,fx,dx,sty,fy,dy,stp,fp,dp,brackt, + stpmin,stpmax) logical brackt double precision stx,fx,dx,sty,fy,dy,stp,fp,dp,stpmin,stpmax c ********** c c Subroutine dcstep c c This subroutine computes a safeguarded step for a search c procedure and updates an interval that contains a step that c satisfies a sufficient decrease and a curvature condition. c c The parameter stx contains the step with the least function c value. If brackt is set to .true. then a minimizer has c been bracketed in an interval with endpoints stx and sty. c The parameter stp contains the current step. c The subroutine assumes that if brackt is set to .true. then c c min(stx,sty) < stp < max(stx,sty), c c and that the derivative at stx is negative in the direction c of the step. c c The subroutine statement is c c subroutine dcstep(stx,fx,dx,sty,fy,dy,stp,fp,dp,brackt, c stpmin,stpmax) c c where c c stx is a double precision variable. c On entry stx is the best step obtained so far and is an c endpoint of the interval that contains the minimizer. c On exit stx is the updated best step. c c fx is a double precision variable. c On entry fx is the function at stx. c On exit fx is the function at stx. c c dx is a double precision variable. c On entry dx is the derivative of the function at c stx. The derivative must be negative in the direction of c the step, that is, dx and stp - stx must have opposite c signs. c On exit dx is the derivative of the function at stx. c c sty is a double precision variable. c On entry sty is the second endpoint of the interval that c contains the minimizer. c On exit sty is the updated endpoint of the interval that c contains the minimizer. c c fy is a double precision variable. c On entry fy is the function at sty. c On exit fy is the function at sty. c c dy is a double precision variable. c On entry dy is the derivative of the function at sty. c On exit dy is the derivative of the function at the exit sty. c c stp is a double precision variable. c On entry stp is the current step. If brackt is set to .true. c then on input stp must be between stx and sty. c On exit stp is a new trial step. c c fp is a double precision variable. c On entry fp is the function at stp c On exit fp is unchanged. c c dp is a double precision variable. c On entry dp is the the derivative of the function at stp. c On exit dp is unchanged. c c brackt is an logical variable. c On entry brackt specifies if a minimizer has been bracketed. c Initially brackt must be set to .false. c On exit brackt specifies if a minimizer has been bracketed. c When a minimizer is bracketed brackt is set to .true. c c stpmin is a double precision variable. c On entry stpmin is a lower bound for the step. c On exit stpmin is unchanged. c c stpmax is a double precision variable. c On entry stpmax is an upper bound for the step. c On exit stpmax is unchanged. c c MINPACK-1 Project. June 1983 c Argonne National Laboratory. c Jorge J. More' and David J. Thuente. c c MINPACK-2 Project. October 1993. c Argonne National Laboratory and University of Minnesota. c Brett M. Averick and Jorge J. More'. c c ********** double precision zero,p66,two,three parameter(zero=0.0d0,p66=0.66d0,two=2.0d0,three=3.0d0) double precision gamma,p,q,r,s,sgnd,stpc,stpf,stpq,theta sgnd = dp*(dx/abs(dx)) c First case: A higher function value. The minimum is bracketed. c If the cubic step is closer to stx than the quadratic step, the c cubic step is taken, otherwise the average of the cubic and c quadratic steps is taken. if (fp .gt. fx) then theta = three*(fx - fp)/(stp - stx) + dx + dp s = max(abs(theta),abs(dx),abs(dp)) gamma = s*sqrt((theta/s)**2 - (dx/s)*(dp/s)) if (stp .lt. stx) gamma = -gamma p = (gamma - dx) + theta q = ((gamma - dx) + gamma) + dp r = p/q stpc = stx + r*(stp - stx) stpq = stx + ((dx/((fx - fp)/(stp - stx) + dx))/two)* + (stp - stx) if (abs(stpc-stx) .lt. abs(stpq-stx)) then stpf = stpc else stpf = stpc + (stpq - stpc)/two endif brackt = .true. c Second case: A lower function value and derivatives of opposite c sign. The minimum is bracketed. If the cubic step is farther from c stp than the secant step, the cubic step is taken, otherwise the c secant step is taken. else if (sgnd .lt. zero) then theta = three*(fx - fp)/(stp - stx) + dx + dp s = max(abs(theta),abs(dx),abs(dp)) gamma = s*sqrt((theta/s)**2 - (dx/s)*(dp/s)) if (stp .gt. stx) gamma = -gamma p = (gamma - dp) + theta q = ((gamma - dp) + gamma) + dx r = p/q stpc = stp + r*(stx - stp) stpq = stp + (dp/(dp - dx))*(stx - stp) if (abs(stpc-stp) .gt. abs(stpq-stp)) then stpf = stpc else stpf = stpq endif brackt = .true. c Third case: A lower function value, derivatives of the same sign, c and the magnitude of the derivative decreases. else if (abs(dp) .lt. abs(dx)) then c The cubic step is computed only if the cubic tends to infinity c in the direction of the step or if the minimum of the cubic c is beyond stp. Otherwise the cubic step is defined to be the c secant step. theta = three*(fx - fp)/(stp - stx) + dx + dp s = max(abs(theta),abs(dx),abs(dp)) c The case gamma = 0 only arises if the cubic does not tend c to infinity in the direction of the step. gamma = s*sqrt(max(zero,(theta/s)**2-(dx/s)*(dp/s))) if (stp .gt. stx) gamma = -gamma p = (gamma - dp) + theta q = (gamma + (dx - dp)) + gamma r = p/q if (r .lt. zero .and. gamma .ne. zero) then stpc = stp + r*(stx - stp) else if (stp .gt. stx) then stpc = stpmax else stpc = stpmin endif stpq = stp + (dp/(dp - dx))*(stx - stp) if (brackt) then c A minimizer has been bracketed. If the cubic step is c closer to stp than the secant step, the cubic step is c taken, otherwise the secant step is taken. if (abs(stpc-stp) .lt. abs(stpq-stp)) then stpf = stpc else stpf = stpq endif if (stp .gt. stx) then stpf = min(stp+p66*(sty-stp),stpf) else stpf = max(stp+p66*(sty-stp),stpf) endif else c A minimizer has not been bracketed. If the cubic step is c farther from stp than the secant step, the cubic step is c taken, otherwise the secant step is taken. if (abs(stpc-stp) .gt. abs(stpq-stp)) then stpf = stpc else stpf = stpq endif stpf = min(stpmax,stpf) stpf = max(stpmin,stpf) endif c Fourth case: A lower function value, derivatives of the c same sign, and the magnitude of the derivative does not c decrease. If the minimum is not bracketed, the step is either c stpmin or stpmax, otherwise the cubic step is taken. else if (brackt) then theta = three*(fp - fy)/(sty - stp) + dy + dp s = max(abs(theta),abs(dy),abs(dp)) gamma = s*sqrt((theta/s)**2 - (dy/s)*(dp/s)) if (stp .gt. sty) gamma = -gamma p = (gamma - dp) + theta q = ((gamma - dp) + gamma) + dy r = p/q stpc = stp + r*(sty - stp) stpf = stpc else if (stp .gt. stx) then stpf = stpmax else stpf = stpmin endif endif c Update the interval which contains a minimizer. if (fp .gt. fx) then sty = stp fy = fp dy = dp else if (sgnd .lt. zero) then sty = stx fy = fx dy = dx endif stx = stp fx = fp dx = dp endif c Compute the new step. stp = stpf end c====================== The end of dcstep ============================== subroutine timer(ttime) double precision ttime c ********* c c Subroutine timer c c This subroutine is used to determine user time. In a typical c application, the user time for a code segment requires calls c to subroutine timer to determine the initial and final time. c c The subroutine statement is c c subroutine timer(ttime) c c where c c ttime is an output variable which specifies the user time. c c Argonne National Laboratory and University of Minnesota. c MINPACK-2 Project. c c Modified October 1990 by Brett M. Averick. c c ********** real temp real tarray(2) real etime c The first element of the array tarray specifies user time temp = etime(tarray) ttime = dble(tarray(1)) return end c====================== The end of timer =============================== double precision function dnrm2(n,x,incx) integer n,incx double precision x(n) c ********** c c Function dnrm2 c c Given a vector x of length n, this function calculates the c Euclidean norm of x with stride incx. c c The function statement is c c double precision function dnrm2(n,x,incx) c c where c c n is a positive integer input variable. c c x is an input array of length n. c c incx is a positive integer variable that specifies the c stride of the vector. c c Subprograms called c c FORTRAN-supplied ... abs, max, sqrt c c MINPACK-2 Project. February 1991. c Argonne National Laboratory. c Brett M. Averick. c c ********** integer i double precision scale dnrm2 = 0.0d0 scale = 0.0d0 do 10 i = 1, n, incx scale = max(scale, abs(x(i))) 10 continue if (scale .eq. 0.0d0) return do 20 i = 1, n, incx dnrm2 = dnrm2 + (x(i)/scale)**2 20 continue dnrm2 = scale*sqrt(dnrm2) return end c====================== The end of dnrm2 =============================== double precision function dpmeps() c ********** c c Subroutine dpeps c c This subroutine computes the machine precision parameter c dpmeps as the smallest floating point number such that c 1 + dpmeps differs from 1. c c This subroutine is based on the subroutine machar described in c c W. J. Cody, c MACHAR: A subroutine to dynamically determine machine parameters, c ACM Trans. Math. Soft., 14, 1988, pages 303-311. c c The subroutine statement is: c c subroutine dpeps(dpmeps) c c where c c dpmeps is a double precision variable. c On entry dpmeps need not be specified. c On exit dpmeps is the machine precision. c c MINPACK-2 Project. February 1991. c Argonne National Laboratory and University of Minnesota. c Brett M. Averick. c c ******* integer i,ibeta,irnd,it,itemp,negep double precision a,b,beta,betain,betah,temp,tempa,temp1, + zero,one,two data zero,one,two /0.0d0,1.0d0,2.0d0/ c determine ibeta, beta ala malcolm. a = one b = one 10 continue a = a + a temp = a + one temp1 = temp - a if (temp1 - one .eq. zero) go to 10 20 continue b = b + b temp = a + b itemp = int(temp - a) if (itemp .eq. 0) go to 20 ibeta = itemp beta = dble(ibeta) c determine it, irnd. it = 0 b = one 30 continue it = it + 1 b = b * beta temp = b + one temp1 = temp - b if (temp1 - one .eq. zero) go to 30 irnd = 0 betah = beta/two temp = a + betah if (temp - a .ne. zero) irnd = 1 tempa = a + beta temp = tempa + betah if ((irnd .eq. 0) .and. (temp - tempa .ne. zero)) irnd = 2 c determine dpmeps. negep = it + 3 betain = one/beta a = one do 40 i = 1, negep a = a*betain 40 continue 50 continue temp = one + a if (temp - one .ne. zero) go to 60 a = a*beta go to 50 60 continue dpmeps = a if ((ibeta .eq. 2) .or. (irnd .eq. 0)) go to 70 a = (a*(one + a))/two temp = one + a if (temp - one .ne. zero) dpmeps = a 70 return end c====================== The end of dpmeps ============================== subroutine daxpy(n,da,dx,incx,dy,incy) c c constant times a vector plus a vector. c uses unrolled loops for increments equal to one. c jack dongarra, linpack, 3/11/78. c double precision dx(*),dy(*),da integer i,incx,incy,ix,iy,m,mp1,n c if(n.le.0)return if (da .eq. 0.0d0) return if(incx.eq.1.and.incy.eq.1)go to 20 c c code for unequal increments or equal increments c not equal to 1 c ix = 1 iy = 1 if(incx.lt.0)ix = (-n+1)*incx + 1 if(incy.lt.0)iy = (-n+1)*incy + 1 do 10 i = 1,n dy(iy) = dy(iy) + da*dx(ix) ix = ix + incx iy = iy + incy 10 continue return c c code for both increments equal to 1 c c c clean-up loop c 20 m = mod(n,4) if( m .eq. 0 ) go to 40 do 30 i = 1,m dy(i) = dy(i) + da*dx(i) 30 continue if( n .lt. 4 ) return 40 mp1 = m + 1 do 50 i = mp1,n,4 dy(i) = dy(i) + da*dx(i) dy(i + 1) = dy(i + 1) + da*dx(i + 1) dy(i + 2) = dy(i + 2) + da*dx(i + 2) dy(i + 3) = dy(i + 3) + da*dx(i + 3) 50 continue return end c====================== The end of daxpy =============================== subroutine dcopy(n,dx,incx,dy,incy) c c copies a vector, x, to a vector, y. c uses unrolled loops for increments equal to one. c jack dongarra, linpack, 3/11/78. c double precision dx(*),dy(*) integer i,incx,incy,ix,iy,m,mp1,n c if(n.le.0)return if(incx.eq.1.and.incy.eq.1)go to 20 c c code for unequal increments or equal increments c not equal to 1 c ix = 1 iy = 1 if(incx.lt.0)ix = (-n+1)*incx + 1 if(incy.lt.0)iy = (-n+1)*incy + 1 do 10 i = 1,n dy(iy) = dx(ix) ix = ix + incx iy = iy + incy 10 continue return c c code for both increments equal to 1 c c c clean-up loop c 20 m = mod(n,7) if( m .eq. 0 ) go to 40 do 30 i = 1,m dy(i) = dx(i) 30 continue if( n .lt. 7 ) return 40 mp1 = m + 1 do 50 i = mp1,n,7 dy(i) = dx(i) dy(i + 1) = dx(i + 1) dy(i + 2) = dx(i + 2) dy(i + 3) = dx(i + 3) dy(i + 4) = dx(i + 4) dy(i + 5) = dx(i + 5) dy(i + 6) = dx(i + 6) 50 continue return end c====================== The end of dcopy =============================== double precision function ddot(n,dx,incx,dy,incy) c c forms the dot product of two vectors. c uses unrolled loops for increments equal to one. c jack dongarra, linpack, 3/11/78. c double precision dx(*),dy(*),dtemp integer i,incx,incy,ix,iy,m,mp1,n c ddot = 0.0d0 dtemp = 0.0d0 if(n.le.0)return if(incx.eq.1.and.incy.eq.1)go to 20 c c code for unequal increments or equal increments c not equal to 1 c ix = 1 iy = 1 if(incx.lt.0)ix = (-n+1)*incx + 1 if(incy.lt.0)iy = (-n+1)*incy + 1 do 10 i = 1,n dtemp = dtemp + dx(ix)*dy(iy) ix = ix + incx iy = iy + incy 10 continue ddot = dtemp return c c code for both increments equal to 1 c c c clean-up loop c 20 m = mod(n,5) if( m .eq. 0 ) go to 40 do 30 i = 1,m dtemp = dtemp + dx(i)*dy(i) 30 continue if( n .lt. 5 ) go to 60 40 mp1 = m + 1 do 50 i = mp1,n,5 dtemp = dtemp + dx(i)*dy(i) + dx(i + 1)*dy(i + 1) + * dx(i + 2)*dy(i + 2) + dx(i + 3)*dy(i + 3) + dx(i + 4)*dy(i + 4) 50 continue 60 ddot = dtemp return end c====================== The end of ddot ================================ subroutine dpofa(a,lda,n,info) integer lda,n,info double precision a(lda,*) c c dpofa factors a double precision symmetric positive definite c matrix. c c dpofa is usually called by dpoco, but it can be called c directly with a saving in time if rcond is not needed. c (time for dpoco) = (1 + 18/n)*(time for dpofa) . c c on entry c c a double precision(lda, n) c the symmetric matrix to be factored. only the c diagonal and upper triangle are used. c c lda integer c the leading dimension of the array a . c c n integer c the order of the matrix a . c c on return c c a an upper triangular matrix r so that a = trans(r)*r c where trans(r) is the transpose. c the strict lower triangle is unaltered. c if info .ne. 0 , the factorization is not complete. c c info integer c = 0 for normal return. c = k signals an error condition. the leading minor c of order k is not positive definite. c c linpack. this version dated 08/14/78 . c cleve moler, university of new mexico, argonne national lab. c c subroutines and functions c c blas ddot c fortran sqrt c c internal variables c double precision ddot,t double precision s integer j,jm1,k c begin block with ...exits to 40 c c do 30 j = 1, n info = j s = 0.0d0 jm1 = j - 1 if (jm1 .lt. 1) go to 20 do 10 k = 1, jm1 t = a(k,j) - ddot(k-1,a(1,k),1,a(1,j),1) t = t/a(k,k) a(k,j) = t s = s + t*t 10 continue 20 continue s = a(j,j) - s c ......exit if (s .le. 0.0d0) go to 40 a(j,j) = sqrt(s) 30 continue info = 0 40 continue return end c====================== The end of dpofa =============================== subroutine dscal(n,da,dx,incx) c c scales a vector by a constant. c uses unrolled loops for increment equal to one. c jack dongarra, linpack, 3/11/78. c modified 3/93 to return if incx .le. 0. c double precision da,dx(*) integer i,incx,m,mp1,n,nincx c if( n.le.0 .or. incx.le.0 )return if(incx.eq.1)go to 20 c c code for increment not equal to 1 c nincx = n*incx do 10 i = 1,nincx,incx dx(i) = da*dx(i) 10 continue return c c code for increment equal to 1 c c c clean-up loop c 20 m = mod(n,5) if( m .eq. 0 ) go to 40 do 30 i = 1,m dx(i) = da*dx(i) 30 continue if( n .lt. 5 ) return 40 mp1 = m + 1 do 50 i = mp1,n,5 dx(i) = da*dx(i) dx(i + 1) = da*dx(i + 1) dx(i + 2) = da*dx(i + 2) dx(i + 3) = da*dx(i + 3) dx(i + 4) = da*dx(i + 4) 50 continue return end c====================== The end of dscal =============================== subroutine dtrsl(t,ldt,n,b,job,info) integer ldt,n,job,info double precision t(ldt,*),b(*) c c c dtrsl solves systems of the form c c t * x = b c or c trans(t) * x = b c c where t is a triangular matrix of order n. here trans(t) c denotes the transpose of the matrix t. c c on entry c c t double precision(ldt,n) c t contains the matrix of the system. the zero c elements of the matrix are not referenced, and c the corresponding elements of the array can be c used to store other information. c c ldt integer c ldt is the leading dimension of the array t. c c n integer c n is the order of the system. c c b double precision(n). c b contains the right hand side of the system. c c job integer c job specifies what kind of system is to be solved. c if job is c c 00 solve t*x=b, t lower triangular, c 01 solve t*x=b, t upper triangular, c 10 solve trans(t)*x=b, t lower triangular, c 11 solve trans(t)*x=b, t upper triangular. c c on return c c b b contains the solution, if info .eq. 0. c otherwise b is unaltered. c c info integer c info contains zero if the system is nonsingular. c otherwise info contains the index of c the first zero diagonal element of t. c c linpack. this version dated 08/14/78 . c g. w. stewart, university of maryland, argonne national lab. c c subroutines and functions c c blas daxpy,ddot c fortran mod c c internal variables c double precision ddot,temp integer case,j,jj c c begin block permitting ...exits to 150 c c check for zero diagonal elements. c do 10 info = 1, n c ......exit if (t(info,info) .eq. 0.0d0) go to 150 10 continue info = 0 c c determine the task and go to it. c case = 1 if (mod(job,10) .ne. 0) case = 2 if (mod(job,100)/10 .ne. 0) case = case + 2 go to (20,50,80,110), case c c solve t*x=b for t lower triangular c 20 continue b(1) = b(1)/t(1,1) if (n .lt. 2) go to 40 do 30 j = 2, n temp = -b(j-1) call daxpy(n-j+1,temp,t(j,j-1),1,b(j),1) b(j) = b(j)/t(j,j) 30 continue 40 continue go to 140 c c solve t*x=b for t upper triangular. c 50 continue b(n) = b(n)/t(n,n) if (n .lt. 2) go to 70 do 60 jj = 2, n j = n - jj + 1 temp = -b(j+1) call daxpy(j,temp,t(1,j+1),1,b(1),1) b(j) = b(j)/t(j,j) 60 continue 70 continue go to 140 c c solve trans(t)*x=b for t lower triangular. c 80 continue b(n) = b(n)/t(n,n) if (n .lt. 2) go to 100 do 90 jj = 2, n j = n - jj + 1 b(j) = b(j) - ddot(jj-1,t(j+1,j),1,b(j+1),1) b(j) = b(j)/t(j,j) 90 continue 100 continue go to 140 c c solve trans(t)*x=b for t upper triangular. c 110 continue b(1) = b(1)/t(1,1) if (n .lt. 2) go to 130 do 120 j = 2, n b(j) = b(j) - ddot(j-1,t(1,j),1,b(1),1) b(j) = b(j)/t(j,j) 120 continue 130 continue 140 continue 150 continue return end c====================== The end of dtrsl =============================== gpml-matlab-v3.0-2010-07-23/util/lbfgsb/lbfgsb.cpp000644 000765 000024 00000007734 11422240326 021237 0ustar00carlstaff000000 000000 #include "mex.h" #include "matrix.h" #include "matlabexception.h" #include "matlabscalar.h" #include "matlabstring.h" #include "matlabmatrix.h" #include "arrayofmatrices.h" #include "program.h" #include "matlabprogram.h" #include #include extern void _main(); // Constants. // ----------------------------------------------------------------- const int minNumInputArgs = 5; // Function definitions. // ----------------------------------------------------------------- void mexFunction (int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) try { // Check to see if we have the correct number of input and output // arguments. if (nrhs < minNumInputArgs) throw MatlabException("Incorrect number of input arguments"); // Get the starting point for the variables. This is specified in // the first input argument. The variables must be either a single // matrix or a cell array of matrices. int k = 0; // The index of the current input argument. ArrayOfMatrices x0(prhs[k++]); // Create the output, which stores the solution obtained from // running IPOPT. There should be as many output arguments as cell // entries in X. if (nlhs != x0.length()) throw MatlabException("Incorrect number of output arguments"); ArrayOfMatrices x(plhs,x0); // Load the lower and upper bounds on the variables as // ArrayOfMatrices objects. They should have the same structure as // the ArrayOfMatrices object "x". ArrayOfMatrices lb(prhs[k++]); ArrayOfMatrices ub(prhs[k++]); // Check to make sure the bounds make sense. if (lb != x || ub != x) throw MatlabException("Input arguments LB and UB must have the same \ structure as X"); // Get the Matlab callback functions. MatlabString objFunc(prhs[k++]); MatlabString gradFunc(prhs[k++]); // Get the auxiliary data. const mxArray* auxData; const mxArray* ptr = prhs[k++]; if (nrhs > 5) { if (mxIsEmpty(ptr)) auxData = 0; else auxData = ptr; } else auxData = 0; // Get the intermediate callback function. MatlabString* iterFunc; ptr = prhs[k++]; if (nrhs > 6) { if (mxIsEmpty(ptr)) iterFunc = 0; else iterFunc = new MatlabString(ptr); } else iterFunc = 0; // Set the options for the L-BFGS algorithm to their defaults. int maxiter = defaultmaxiter; int m = defaultm; double factr = defaultfactr; double pgtol = defaultpgtol; // Process the remaining input arguments, which set options for // the IPOPT algorithm. while (k < nrhs) { // Get the option label from the Matlab input argument. MatlabString optionLabel(prhs[k++]); if (k < nrhs) { // Get the option value from the Matlab input argument. MatlabScalar optionValue(prhs[k++]); double value = optionValue; if (!strcmp(optionLabel,"maxiter")) maxiter = (int) value; else if (!strcmp(optionLabel,"m")) m = (int) value; else if (!strcmp(optionLabel,"factr")) factr = value / mxGetEps(); else if (!strcmp(optionLabel,"pgtol")) pgtol = value; else { if (iterFunc) delete iterFunc; throw MatlabException("Nonexistent option"); } } } // Create a new instance of the optimization problem. x = x0; MatlabProgram program(x,lb,ub,&objFunc,&gradFunc,iterFunc, (mxArray*) auxData,m,maxiter,factr,pgtol); // Run the L-BFGS-B solver. SolverExitStatus exitStatus = program.runSolver(); if (exitStatus == abnormalTermination) { if (iterFunc) delete iterFunc; throw MatlabException("Solver unable to satisfy convergence \ criteria due to abnormal termination"); } else if (exitStatus == errorOnInput) { if (iterFunc) delete iterFunc; throw MatlabException("Invalid inputs to L-BFGS routine"); } // Free the dynamically allocated memory. if (iterFunc) delete iterFunc; } catch (std::exception& error) { mexErrMsgTxt(error.what()); } gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabstring.h000644 000765 000024 00000002202 11422240326 022115 0ustar00carlstaff000000 000000 #ifndef INCLUDE_MATLABSTRING #define INCLUDE_MATLABSTRING #include "mex.h" #include // Function declarations. // ----------------------------------------------------------------- // Copy a C-style string (i.e. a null-terminated character array). char* copystring (const char* source); // Class MatlabString. // ----------------------------------------------------------------- // This class encapsulates read-only access to a MATLAB character // array. class MatlabString { public: // The constructor accepts as input a pointer to a Matlab array, // which must be a valid string (array of type CHAR). explicit MatlabString (const mxArray* ptr); // The copy constructor makes a full copy of the source string. MatlabString (const MatlabString& source); // The destructor. ~MatlabString(); // Conversion operator for null-terminated string. operator const char* () const { return s; }; protected: char* s; // The null-terminated string. // The copy assignment operator is not proper, thus remains // protected. MatlabString& operator= (const MatlabString& source) { return *this; }; }; #endif gpml-matlab-v3.0-2010-07-23/util/lbfgsb/README.html000644 000765 000024 00000043521 11422240326 021111 0ustar00carlstaff000000 000000 MATLAB interface for L-BFGS-B

A MATLAB interface for L-BFGS-B

by Peter Carbonetto
Dept. of Computer Science
University of British Columbia

L-BFGS-B is a collection of Fortran 77 routines for solving large-scale nonlinear optimization problems with bound constraints on the variables. One of the key features of the nonlinear solver is that knowledge of the Hessian is not required; the solver computes search directions by keeping track of a quadratic model of the objective function with a limited-memory BFGS (Broyden-Fletcher-Goldfarb-Shanno) approximation to the Hessian.1 The algorithm was developed by Ciyou Zhu, Richard Byrd and Jorge Nocedal. For more information, go to the original distribution site for the L-BFGS-B software package.

I've designed an interface to the L-BFGS-B solver so that it can be called like any other function in MATLAB.2 See the text below for more information on installing and calling it in MATLAB. Along the way, I've also developed a C++ class that encapsulates all the "messy" details in executing the L-BFGS-B code. See below for instructions on how to use this class for your own C++ code.

If you have any questions, praise, or comments, or would like to report a bug, do not hesitate to contact the author. I've tested this software in MATLAB version 7.2 on both the Mac OS X and Linux operating systems.

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 2.5 Canada License. Please cite the source as Peter Carbonetto, Department of Computer Science, University of British Columbia.

Installation

These installation instructions assume you have a UNIX-based operating system, such as Mac OS X or Linux. It is surely possible to install this software on Windows, I'm just not sure how. These instructions also assume you have GNU make installed.

Download the source. First, download and unpack the tar archive. It contains some MATLAB files (ending in .m), some C++ source and header files (ending in .h and .cpp), a single Fortran 77 source file solver.f containing the L-BFGS-B routines, and a Makefile. I've included a version of the Fortran routines that is more recent than what is available for download at the distribution site at Northwestern University.

What we will do is create a MEX file, which is basically a file that contains a routine that can be called from MATLAB as if it were a built-it function. To learn about MEX files, I refer you to this document at the MathWorks website.

Install the C++ and Fortran 77 compilers. In order to build the MEX file, you will need both a C++ compiler and a Fortran 77 compiler. Unfortunately, you can't just use any compiler. You have to use the precise one supported by MATLAB. For instance, if you are running Mac OS X 7.2 on a Linux operating system, you will need to install the GNU compiler collection (GCC) version 3.4.4. Even if you already have a compiler installed on Linux, it may be the wrong version, and if you use the wrong version things could go horribly wrong! It is important that you use the correct version of the compiler, otherwise you will encounter linking errors. Refer to this document to find out which compiler is supported by your version of MATLAB.

Configure MATLAB. Next, you need to set up and configure MATLAB to build MEX Files. This is explained quite nicely in this MathWorks support document.

Modify the Makefile. You are almost ready to build the MEX file. But before you do so, you need to edit the Makefile to coincide with your system setup. At the top of the file are several variables you may need to modify. The variable MEX is the executable called to build the MEX file. The variables CXX and F77 must be the names of your C++ and Fortran 77 compilers. MEXSUFFIX is the MEX file extension specific to your platform. (For instance, the extension for Linux is mexglx.) The variable MATLAB_HOME must be the base directory of your MATLAB installation. Finally, CFLAGS and FFLAGS are options passed to the C++ and Fortran compilers, respectively. Often these flags coincide with your MEX options file (see here for more information on the options file). But often the settings in the MEX option file are incorrect, and so the options must be set manually. MATLAB requires some special compilation flags for various reasons, one being that it requires position-independent code. These instructions are vague (I apologize), and this step may require a bit of trial and error until before you get it right.

On my laptop running Mac OS X 10.3.9 with MATLAB 7.2, my settings for the Makefile are

  MEX         = mex
  MEXSUFFIX   = mexmac
  MATLAB_HOME = /Applications/MATLAB72
  CXX         = g++
  F77         = g77 
  CFLAGS      = -O3 -fPIC -fno-common -fexceptions \
                -no-cpp-precomp 
  FFLAGS      = -O3 -x f77-cpp-input -fPIC -fno-common

On my Linux machine, I set the variables in the Makefile like so:

  MEX         = mex
  MEXSUFFIX   = mexglx
  MATLAB_HOME = /cs/local/generic/lib/pkg/matlab-7.2
  CXX         = g++-3.4.5
  F77         = g77-3.4.5
  CFLAGS      = -O3 -fPIC -pthread 
  FFLAGS      = -O3 -fPIC -fexceptions

It may be helpful to look at the GCC documentation in order to understand what these various compiler flags mean.

Build the MEX file. If you are in the directory containing all the source files, typing make in the command prompt will first compile the Fortran and C++ source files into object code (.o files). After that, the make program calls the MEX script, which in turn links all the object files together into a single MEX file. If you didn't get any errors, then you are ready to try out the bound-constrained solver in MATLAB. Note that even if you didn't get any errors, there's still a possibility that you didn't link the MEX file properly, in which case executing the MEX file will cause MATLAB to crash. But there's only one way to find out: the hard way.

A brief tutorial

I've written a short script examplehs038.m which demonstrates how to call lbfgsb for solving a small optimization problem, the Hock & Schittkowski test problem no. 38.3 The optimization problem has four variables. They are bounded from below by -10 and bounded from above by +10. We've set the starting point to (-3, -1, -3, -1). We've written two MATLAB functions for computing the objective function and the gradient of the objective function at a given point. These functions can be found in the files computeObjectiveHS038.m and computeGradientHS038.m. If you run the script, the solver should very quickly progress toward the point (1, 1, 1, 1). The objective is 0 at this point.

The basic MATLAB function call is

  lbfgsb(x0,lb,ub,objfunc,gradfunc)

The first input argument x0 declares the starting point for the solver. The second and third input arguments define the lower and upper bounds on the variables, respectively. If an entry of lb is set to -Inf, then this is equivalent to no lower bound for that entry (the same applies to upper bounds, except that it must be +Inf instead). The next two input arguments must be the names of MATLAB routines (M-files). The first routine must calculate the objective function at the current point. The output must be a scalar representing the objective evaluated at the current point. The second function must compute the gradient of the objective at the current point. The input is the same as objfunc, but it must return as many outputs as there are inputs. For a complete guide on using the MATLAB interface, type help lbfgsb in the MATLAB prompt.

The script exampleldaimages.m is a more complicated example. It uses the L-BFGS-B solver to compute some statistics of a posterior distribution for a text analysis model called latent Dirichlet allication (LDA). The model is used to estimate the topics for a collection of documents. The optimization problem is to compute an approximation to the posterior, since it is not possible to compute the posterior exactly. In this case, the optimization problem corresponds to minimizing the distance between the true posterior distribution and an approximate distribution. For more information, see the paper of Blei, Ng and Jordan (2003).4 In order to run this example, you will need to install the lightspeed toolbox for MATLAB developed by Tom Minka at Microsoft Research.

The script starts by generating a bunch of canonical images that represent topics; if a pixel is white then a word at that position is more likely to appear in that topic. Note that the order of the pixels in the image is unimportant, since LDA is a "bag of words" model. In another figure we show a small sample of the data generated from the topics. Each image represents a document, and the pixels in the image depict the word proportions in that document.5

Next, the script runs the L-BFGS-B solver to find a local minimum to the variational objective. Buried in the M-file mflda.m is the call to the solver: 

  [xi gamma Phi] = lbfgsb({xi gamma Phi},...
      {repmat(lb,W,K)  repmat(lb,K,D)  repmat(lb,K,M)},...
      {repmat(inf,W,K) repmat(inf,K,D) repmat(inf,K,M)},...
      'computeObjectiveMFLDA','computeGradientMFLDA',...
      {nu eta L w lnZconst},'callbackMFLDA',...
      'm',m_lbfgs,'factr',1e-12,'pgtol',...
      tolerance,'maxiter',maxiter);

This example will give us the opportunity to demonstrate some of the more complicated aspects of the our MATLAB interface. The first input to lbfgsb sets the initial iterate of the optimization algorithm. Notice that we've passed a cell array with entries that are matrices. Indeed, it is possible to pass either a matrix or a cell array. Whatever structure is passed, the other input arguments (and outputs) must also abide by this structure. The lower and upper bounds on variables are also cell arrays. All the upper bounds are set to infinity, which means that the variables are only bounded from below.

It is instructive to examine the callback routines. They look like: 

  f = computeObjectiveMFLDA(xi,gamma,Phi,auxdata)

  [dxi, dgamma, dPhi] = ...
      computeGradientMFLDA(xi,gamma,Phi,auxdata)

Notice that each entry to the cell array is its own input argument to the callback routines. The same applies for the output arguments of the gradient callback function.

The sixth input argument is a cell array that contains auxiliary data. This data will be passed to the MATLAB functions that evaluate the objective and gradient. The seventh input argument specifies a callback routine that is called exactly once for every iteration. This callback routine is most useful for examining the progress of the solver. The remaining input arguments are label/value pairs that override some of the default settings of L-BFGS-B.

After converging to a solution, we display topic samples drawn from the variational approximation. A good solution should come close to recovering the true topics from which the data was generated (although they might not be in the same order).

Don't be surprised if it takes many iterations before the optimization algorithm converges on a local minimum. (In repeated trials, I found it as many as two thousand iterations to converge to a stationary point of the objective.) This particular problem demonstrates the limitations of limited-memory quasi-Newton approximations to the Hessian; the low storage requirements can come at the cost of slow convergence to the solution.

The C++ interface

One of the nice byproducts of writing a MATLAB interface for L-BFGS-B is that I've ended up with a neat little C++ class that encapsulates the nuts and bolts of executing the solver. A brief description of the Program class can be found in the header file program.h.

Program is an abstract class, which means that its impossible to instantiate a Program object. This is because the class has member functions that aren't yet implemented. These are called pure virtual functions. In order to use the Program class, one needs to define another class that inherits the Program class and that implements the pure virtual functions. In fact, the class MatlabProgram (in matlabprogram.h) is an example of such a class.

The new child class must declare and implement these two functions:

  virtual double computeObjective (int n, double* x);
  virtual void computeGradient (int n, double* x, double* g);

See the header file for a detailed description of these functions. The only remaining detail is calling the Program constructor. After that, it is just a matter of declaring a new object of type MyProgram then calling the function runSolver.

Contents of the tar archive

solver.f
Fortran 77 source for the L-BFGS-B solver routines.

program.h
program.cpp
Header and source file for the C++ interface.

array.h
arrayofmatrices.h
matlabexception.h
matlabmatrix.h
matlabprogram.h
matlabscalar.h
matlabstring.h
arrayofmatrices.cpp
matlabexception.cpp
matlabmatrix.cpp
matlabprogram.cpp
matlabscalar.cpp
matlabstring.cpp
lbfgsb.cpp
Header and source files for the MATLAB interface.

lbfgsb.m
Usage instructions for the MEX file.

examplehs038.m
MATLAB script for solving the H&S test example no. 38.

computeObjectiveHS038.m
computeGradientHS038.m
genericcallback.m
MATLAB functions used by the script examplehs038.m.

exampleldaimages.m
MATLAB script that generates synthetic documents and topics, then computes a mean field variational approximation to the posterior distribution of the latent Dirichlet allocation model, then displays the result.

mflda.m
This function computes a mean field variational approximation to the posterior distribution of the latent Dirichlet allocation model by minimizing the distance between the variational distribution and the true distribution. It uses the L-BFGS-B to minimize the objective function subject to the bound constraints. The objective function also acts as a lower bound on the logarithm of the denominator that appears from the application of Bayes' rule.4

callbackMFLDA.m
computeObjectiveMFLDA.m
computeGradientMFLDA.m
computelnZ.m
computelnZconst.m
computem.m
dirichletrnd.m
gammarnd.m
createsyntheticdata.m
mfldainit.m
Some functions used by exampleldaimages.m and mflda.m.

Footnotes

1 Ciyou Zhu, Richard H. Byrd, Peihuang Lu and Jorge Nocedal (1997). Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound-constrained optimization. ACM Transactions on Mathematical Software, Vol. 23, No. 4., pp. 550-560.

1 Another MATLAB interface for the L-BFGS-B routines has been developed by Liam Stewart at the University of Toronto.

3 Willi Hock and Klaus Schittkowski (1981). Test Examples for Nonlinear Programming Codes. Lecture Notes in Economics and Mathematical Systems, Vol. 187, Springer-Verlag.

4 David M. Blei, Andrew Y. Ng, Michael I. Jordan (2003). Latent Dirichlet allocation. Journal of Machine Learning Research, Vol. 3, pp. 993-1022.

5 Tom L. Griffiths and Mark Steyvers (2004). Finding scientific topics. Proceedings of the National Academy of Sciences, Vol. 101, pp. 5228-5235.


March 3, 2007
gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabprogram.h000644 000765 000024 00000005501 11422240326 022263 0ustar00carlstaff000000 000000 #ifndef INCLUDE_MATLABPROGRAM #define INCLUDE_MATLABPROGRAM #include "mex.h" #include "program.h" #include "matlabscalar.h" #include "matlabstring.h" #include "arrayofmatrices.h" // Class MatlabProgram. // ----------------------------------------------------------------- // This is an implementation of the abstract class Program. class MatlabProgram: public Program { public: // On input, "variables" should contain the initial point for the // optimization routine. If no iterative callback function is // specified, "iterFunc" may be set to 0. Also, if no auxiliary data // is needed, it may also be set to 0. MatlabProgram (ArrayOfMatrices& variables, const ArrayOfMatrices& lowerbounds, const ArrayOfMatrices& upperbounds, const MatlabString* objFunc, const MatlabString* gradFunc, const MatlabString* iterFunc, mxArray* auxData, int m = defaultm, int maxiter = defaultmaxiter, double factr = defaultfactr, double pgtol = defaultpgtol); // The destructor. virtual ~MatlabProgram(); // These provide definitions for the pure virtual functions of the // abstract parent class. virtual double computeObjective (int n, double* x); virtual void computeGradient (int n, double* x, double* g); virtual void iterCallback (int t, double* x, double f); // Run the solver. Upon completion, the solution is stored in // "variables". SolverExitStatus runSolver(); protected: ArrayOfMatrices& variables; // Storage for the initial value and // solution. const MatlabString* objFunc; // The objective callback function. const MatlabString* gradFunc; // The gradient callback function. const MatlabString* iterFunc; // The iterative callback function. ArrayOfMatrices* varMatlab; // Inputs to the Matlab callback mxArray** varInputs; // functions representing the // current values of the variables. MatlabScalar* fMatlab; // Input to the Matlab callback // functions representing the // current value of the objective. MatlabScalar* tMatlab; // Input to the Matlab callback // functions representing the // current iteration. int numInputsObjFunc; // The number of inputs passed to the // objective callback function. int numInputsGradFunc; // The number of inputs passed to the // gradient callback function. int numInputsIterFunc; // The number of inputs passed to the // iterative callback function. mxArray** inputsObjFunc; // The inputs passed to the objective // callback function. mxArray** inputsGradFunc; // The inputs passed to the gradient // callback function. mxArray** inputsIterFunc; // The inputs passed to the iterative // callback function. }; #endif gpml-matlab-v3.0-2010-07-23/util/lbfgsb/arrayofmatrices.h000644 000765 000024 00000004144 11422240326 022630 0ustar00carlstaff000000 000000 #ifndef INCLUDE_ARRAYOFMATRICES #define INCLUDE_ARRAYOFMATRICES #include "array.h" #include "matlabmatrix.h" #include "mex.h" // Class ArrayOfMatrices. // ----------------------------------------------------------------- class ArrayOfMatrices : public Array { public: // This version of the constructor behaves just like its parent. explicit ArrayOfMatrices (int length) : Array(length) { }; // This constructor creates an array of matrices from a Matlab // array. It accepts either a matrix in double precision, or a cell // array with entries that are matrices. explicit ArrayOfMatrices (const mxArray* ptr); // This constructor creates an array of matrices from a collection // of Matlab arrays. The Matlab arrays must be matrices. ArrayOfMatrices (const mxArray* ptrs[], int numptrs); // This constructor creates an array of matrices and the // associated Matlab structures. The Matlab structures are // matrices. The second input argument acts as a template for the // creation of the matrices, but the data from "model" is not // actually copied into the new ArrayOfMatrices object. It is up // to the user to make sure that the array of mxArray pointers has // enough room for the pointers to the Matlab arrays. ArrayOfMatrices (mxArray* ptrs[], const ArrayOfMatrices& model); // This constructor creates an array of matrices using the second // input argument as a model. The input argument "data" contains // the element data. Note that the information is NOT copied from // the model! ArrayOfMatrices (double* data, const ArrayOfMatrices& model); // The copy constructor makes a shallow copy of the source object. ArrayOfMatrices (const ArrayOfMatrices& source); // The destructor. ~ArrayOfMatrices(); // Copy assignment operator that observes the same behaviour as // the Array copy assignment operator. ArrayOfMatrices& operator= (const ArrayOfMatrices& source); // Return the total number of elements in all the matrices. int numelems() const; protected: static int getnummatlabmatrices (const mxArray* ptr); }; #endif gpml-matlab-v3.0-2010-07-23/util/lbfgsb/array.h000644 000765 000024 00000010714 11422240326 020553 0ustar00carlstaff000000 000000 #ifndef INCLUDE_ARRAY #define INCLUDE_ARRAY #include "matlabexception.h" #include // Function definitions. // ----------------------------------------------------------------- template void copymemory (const Type* source, Type* dest, int length) { memcpy(dest,source,sizeof(Type)*length); } // Class Array // ----------------------------------------------------------------- // An Array object stores a collection of ordered elements. The key // aspect of Array objects is that they do not necessarily achieve // encapsulation; they do not necessarily retain independent // ownership of their data. That is, the data could be modified // externally by another agent. This behaviour is determined by the // data member "owner". If "owner" is false, the array does not take // care of allocation and deallocation of the data in memory. // // Copy assignment behaves different than usual---it copies the // data, but requires that the destination already have the proper // resources allocated in memory. // // The copy constructor performs a shallow copy. template class Array { public: // This constructor allocates memory for an array of elements of // the specificed length. explicit Array (int length); // This constructor is set to point to an already existing // array. As such, it does not gain ownership of the elements. Array (Type* data, int length); // The copy constructor performs a shallow copy, so that both // arrays share the same copy of the data. Array (const Array& source); // The destructor. ~Array(); // Set all the elements of the array to the specified value. void setvalue (const Type& value); // Copy the elements to from the source array. Both the source and // destination array must have the same length. void inject (const Type* source); void inject (const Array& source); Array& operator= (const Type* source); Array& operator= (const Array& source); // Get the number of elements in the array. int length() const { return n; }; // Elements of an array A can be accessed the subscript operator; // e.g. A[i]. For the sake of efficiency, no checks are made to // ensure that the subscripting is legal. It is up to the user to // ensure that i is non-negative and less than A.length(). Note // that subscripts, as per C++ convention, start at 0. Type& operator[] (int i); const Type& operator[] (int i) const; // Returns true if the two arrays have the same dimensions // (i.e. the same lengths). bool operator== (const Array& a) const { return n == a.n; }; bool operator!= (const Array& a) const { return !(*this == a); }; protected: Type* elems; int n; // The number of elements in the array. bool owner; // Whether the object has ownership of the data. }; // Function definitions for class Array. // ----------------------------------------------------------------- template Array::Array (int length) { n = length; owner = true; elems = 0; elems = new Type[length]; } template Array::Array (Type* data, int length) { n = length; owner = false; elems = data; } template Array::Array (const Array& source) { n = source.n; owner = false; elems = source.elems; } template Array::~Array () { if (owner && elems) delete[] elems; } template void Array::setvalue (const Type& value) { for (int i = 0; i < n; i++) elems[i] = value; } template void Array::inject (const Type* source) { // Check for self-assignment. If there is self-assignment, there // is no need to copy because the objects share the same data! if (elems != source) copymemory(source,elems,n); } template void Array::inject (const Array& source) { if (n != source.n) throw MatlabException("Unable to perform copy; arrays have \ different lengths"); inject(source.elems); } template Array& Array::operator= (const Type* source) { inject(source); return *this; } template Array& Array::operator= (const Array& source) { inject(source); return *this; } template Type& Array::operator[] (int i) { return elems[i]; } template const Type& Array::operator[] (int i) const { return elems[i]; } #endif gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabexception.cpp000644 000765 000024 00000001000 11422240326 023133 0ustar00carlstaff000000 000000 #include "matlabexception.h" // Function definitions for class MatlabException. // ----------------------------------------------------------------- MatlabException::MatlabException (const char* message) throw() : exception() { this->message = message; } MatlabException::MatlabException (const MatlabException& source) throw() : exception() { message = source.message; } MatlabException& MatlabException::operator= (const MatlabException& source) { message = source.message; return *this; } gpml-matlab-v3.0-2010-07-23/util/lbfgsb/arrayofmatrices.cpp000644 000765 000024 00000004655 11422240326 023172 0ustar00carlstaff000000 000000 #include "arrayofmatrices.h" #include "matrix.h" // Function definitions for class ArrayOfMatrices. // ----------------------------------------------------------------- ArrayOfMatrices::ArrayOfMatrices (const mxArray* ptr) : Array(getnummatlabmatrices(ptr)) { setvalue(0); if (mxIsCell(ptr)) // Fill in the entries of the array with the entries of the Matlab // cell array. for (int i = 0; i < n; i++) elems[i] = new Matrix(mxGetCell(ptr,i)); else // Fill in the single entry of the array with the Matlab matrix. elems[0] = new Matrix(ptr); } ArrayOfMatrices::ArrayOfMatrices (const mxArray* ptrs[], int numptrs) : Array(numptrs) { setvalue(0); for (int i = 0; i < numptrs; i++) elems[i] = new Matrix(ptrs[i]); } ArrayOfMatrices::ArrayOfMatrices (mxArray* ptrs[], const ArrayOfMatrices& source) : Array(source.length()) { setvalue(0); // Create the new Matlab matrices and copy the data from the // source. for (int i = 0; i < n; i++) { int h = source[i]->height(); int w = source[i]->width(); elems[i] = new Matrix(ptrs[i],h,w); *elems[i] = *source[i]; } } ArrayOfMatrices::ArrayOfMatrices (const ArrayOfMatrices& source) : Array(source) { } ArrayOfMatrices::ArrayOfMatrices (double* data, const ArrayOfMatrices& model) : Array(model.length()) { setvalue(0); // Repeat for each matrix. for (int i = 0; i < n; i++) { const Matrix* A = model[i]; elems[i] = new Matrix(data,A->height(),A->width()); data += A->length(); } } ArrayOfMatrices::~ArrayOfMatrices() { for (int i = 0; i < n; i++) if (elems[i]) delete elems[i]; } int ArrayOfMatrices::numelems() const { int m = 0; // The return value. // Repeat for each matrix. for (int i = 0; i < n; i++) m += elems[i]->length(); return m; } ArrayOfMatrices& ArrayOfMatrices::operator= (const ArrayOfMatrices& source) { // Copy each matrix. for (int i = 0; i < n; i++) *elems[i] = *source.elems[i]; return *this; } int ArrayOfMatrices::getnummatlabmatrices (const mxArray* ptr) { int n; // The return value. if (mxIsCell(ptr)) n = mxGetNumberOfElements(ptr); else if (mxGetNumberOfDimensions(ptr) == 2 && mxIsDouble(ptr)) n = 1; else throw MatlabException("The Matlab array must either be a cell \ array or a double precision matrix"); return n; } gpml-matlab-v3.0-2010-07-23/util/lbfgsb/program.h000644 000765 000024 00000011647 11422240326 021112 0ustar00carlstaff000000 000000 #ifndef INCLUDE_PROGRAM #define INCLUDE_PROGRAM // Type definitions. // ----------------------------------------------------------------- // This defines the possible results of running the L-BFGS-B solver. enum SolverExitStatus { success, // The algorithm has converged to a stationary // point or has reached the maximum number of // iterations. abnormalTermination, // The algorithm has terminated abnormally and // was unable to satisfy the convergence // criteria. errorOnInput // The routine has detected an error in the // input parameters. }; // Constants. // ----------------------------------------------------------------- const int defaultm = 5; const int defaultmaxiter = 100; const double defaultfactr = 1e7; const double defaultpgtol = 1e-5; const int defaultprintlevel = -1; // Function declarations. // ----------------------------------------------------------------- // This is the L-BFGS-B routine implemented in Fortran 77. extern "C" void setulb_ (int* n, int* m, double x[], double l[], double u[], int nbd[], double* f, double g[], double* factr, double* pgtol, double wa[], int iwa[], char task[], int* iprint, char csave[], bool lsave[], int isave[], double dsave[]); // Class Program. // ----------------------------------------------------------------- // This class encapsulates execution of the L-BFGS-B routine for // solving a nonlinear optimization problem with bound constraints // using limited-memory approximations to the Hessian. // // This is an abstract class since some of the class methods have not // been implemented (they are "pure virtual" functions). In order to // use this class, one needs to define a child class and provide // definitions for the pure virtual methods. class Program { public: // The first input argument "n" is the number of variables, and "x" // must be set to the suggested starting point for the optimization // algorithm. See the L-BFGS-B documentation for more information on // the inputs to the constructor. Program (int n, double* x, double* lb, double* ub, int* btype, int m = defaultm, int maxiter = defaultmaxiter, double factr = defaultfactr, double pgtol = defaultpgtol); // This is the same constructor as above, except that the // appropriate amount of memory is dynamically allocated for the // variables, the bounds, and the bound types. The destructor also // makes sure that the memory is properly deallocated, but not so // for the other constructor. Program (int n, int m = defaultm, int maxiter = defaultmaxiter, double factr = defaultfactr, double pgtol = defaultpgtol); // The destructor. virtual ~Program(); // An implementation of this method should return the value of the // objective at the current value of the variables "x", where "n" is // the number of variables. virtual double computeObjective (int n, double* x) = 0; // An implementation of this method should fill in the values of the // gradient "g" and the current point "x". virtual void computeGradient (int n, double* x, double* g) = 0; // The child class may optionally override this method which is // called once per iteration of the L-BFGS-B routine. The input // arguments are the current iteration "t", the current value of the // variables "x" and the current value of the objective "f". virtual void iterCallback (int t, double* x, double f) { }; // Run the solver. Upon completion, the solution is stored in "x". SolverExitStatus runSolver(); protected: // The copy constructor and copy assignment operator are kept // private so that they are not used. Program (const Program& source) { }; Program& operator= (const Program& source) { return *this; }; int n; // The number of variables. double* x; // The current point. double* lb; // The lower bounds. double* ub; // The upper bounds. int* btype; // The bound types. private: // Execute a single step the L-BFGS-B solver routine. void callLBFGS (const char* cmd = 0); // Initialize some of the structures used by the L-BFGS-B routine. void initStructures(); bool owner; // If true, then the blocks of memory pointed to by // x, lb, ub and btype must be deallocated in the // destructor. double f; // The value of the objective. double* g; // The value of the gradient. int iprint; // The print level. int maxiter; // The maximum number of iterations. double pgtol; // Convergence parameter passed to L-BFGS-B. double factr; // Convergence parameter passed to L-BFGS-B. int m; // The number of variable corrections to the // limited-memory approximation to the Hessian. // These are structures used by the L-BFGS-B routine. double* wa; int* iwa; char task[60]; char csave[60]; bool lsave[4]; int isave[44]; double dsave[29]; }; #endif gpml-matlab-v3.0-2010-07-23/util/lbfgsb/license.txt000644 000765 000024 00000003027 11422240326 021446 0ustar00carlstaff000000 000000 Copyright (c) 2009, Peter Carbonetto All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution * Neither the name of the University of British Columbia nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabscalar.h000644 000765 000024 00000002264 11422240326 022064 0ustar00carlstaff000000 000000 #ifndef INCLUDE_MATLABSCALAR #define INCLUDE_MATLABSCALAR #include "mex.h" // Class MatlabScalar // ----------------------------------------------------------------- // The main appeal of this class is that one can create a scalar // object that accesses a MATLAB array. // // Note that the copy assignment operator is not valid for this class // because we cannot reassign a reference. class MatlabScalar { public: // This constructor accepts as input a pointer to a Matlab array // which must be a scalar in double precision. explicit MatlabScalar (const mxArray* ptr); // This constructor creates a new Matlab array which is a scalar // in double precision. MatlabScalar (mxArray*& ptr, double value); // The copy constructor. MatlabScalar (MatlabScalar& source); // The destructor. ~MatlabScalar() { }; // Access the value of the scalar. operator const double () const { return x; }; // Assign the value of the scalar. MatlabScalar& operator= (double value); protected: double& x; // The copy assignment operator is kept protected because it is // invalid. MatlabScalar& operator= (const MatlabScalar& source) { return *this; }; }; #endif gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabmatrix.cpp000644 000765 000024 00000003156 11422240326 022457 0ustar00carlstaff000000 000000 #include "matlabmatrix.h" #include "matlabexception.h" // Function definitions. // ----------------------------------------------------------------- double* getMatlabMatrixDouble (const mxArray* ptr) { if (mxGetNumberOfDimensions(ptr) != 2) throw MatlabException("Matlab array must be a matrix"); if (!mxIsDouble(ptr)) throw MatlabException("Matlab array must be of type double"); return mxGetPr(ptr); } double* createMatlabMatrixDouble (mxArray*& ptr, int height, int width) { ptr = mxCreateDoubleMatrix(height,width,mxREAL); return mxGetPr(ptr); } // Function definitions for class Matrix. // ----------------------------------------------------------------- Matrix::Matrix (int height, int width) : Array(height*width) { h = height; w = width; } Matrix::Matrix (double* data, int height, int width) : Array(data,height*width) { h = height; w = width; } Matrix::Matrix (const mxArray* ptr) : Array(getMatlabMatrixDouble(ptr),mxGetNumberOfElements(ptr)) { h = mxGetM(ptr); w = mxGetN(ptr); } Matrix::Matrix (mxArray*& ptr, int height, int width) : Array(createMatlabMatrixDouble(ptr,height,width), height*width) { h = height; w = width; } Matrix::Matrix (const Matrix& source) : Array(source) { h = source.h; w = source.w; } Matrix& Matrix::operator= (const Matrix& source) { inject(source); return *this; } bool Matrix::operator== (const Matrix& X) const { return (h == X.h) && (w == X.w); } double& Matrix::entry (int r, int c) { return elems[h*c + r]; } double Matrix::entry (int r, int c) const { return elems[h*c + r]; } gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabmatrix.h000644 000765 000024 00000004562 11422240326 022126 0ustar00carlstaff000000 000000 #ifndef INCLUDE_MATLABMATRIX #define INCLUDE_MATLABMATRIX #include "array.h" #include "mex.h" // Class Matrix // --------------------------------------------------------------- // A matrix object stores its elements in column-major format, as in // Fortran and Matlab. This means that columns are stored one after // another. For example, the matrix // // 1 2 3 // 4 5 6 // // is stored in memory as // // 1 4 2 5 3 6 // // Like Array objects, Matrix objects are not necessarily // encapsulated, and they exhibit analogous behaviour. class Matrix : public Array { public: // This constructor allocates memory for matrix of the specified // heigth and width. Matrix (int height, int width); // This constructor basically follows the lead of the analagous // constructor for the Array class. Matrix (double* data, int height, int width); // This constructor retrieves a matrix from a Matlab array. This // particular constructor is only defined for Matrix. // Since Matlab handles storage, the object created by this // constructor is not encapsulated. explicit Matrix (const mxArray* ptr); // This constructor creates a new Matlab array as a side effect. // Since Matlab handles storage, the object created by this // constructor is not encapsulated. Matrix (mxArray*& ptr, int height, int width); // The copy constructor makes a shallow copy of the data. Matrix (const Matrix& source); // The destructor. ~Matrix() { }; // Copy assignment operator that observes the same behaviour as // the Array copy assignment operator. Matrix& operator= (const Matrix& source); // Get the height and width of the matrix. int height() const { return h; }; int width () const { return w; }; // Returns true if the two matrices have the same dimensions // (i.e. the same height and width). bool operator== (const Matrix& X) const; bool operator!= (const Matrix& X) const { return !(*this == X); }; // If X is an object of type Matrix, X.entry(r,c) accesses the // entry of the rth row and cth column. double entry (int r, int c) const; double& entry (int r, int c); double operator() (int r, int c) const { return entry(r,c); }; double& operator() (int r, int c) { return entry(r,c); }; protected: int h; // The height of the matrix. int w; // The width of the matrix. }; #endif gpml-matlab-v3.0-2010-07-23/util/lbfgsb/matlabprogram.cpp000644 000765 000024 00000016452 11422240326 022625 0ustar00carlstaff000000 000000 #include "matlabprogram.h" #include "array.h" // Function definitions. // ----------------------------------------------------------------- // Copy the matrix elements to the destination array. It is assumed // that sufficient memory has been allocated for the destination // array. The copying preserves the column-major ordering of the // matrix elements. void copyElemsFromMatrix (const Matrix& source, double* dest) { Matrix destmatrix(dest,source.height(),source.width()); destmatrix = source; } // Copy the elements from the source array into the matrix. void copyElemsToMatrix (const double* source, Matrix& dest) { dest.inject(source); } // Copy the elements contained in all the matrices to the destination // array. It is assumed that sufficient memory has been allocated for // the destination array. The copying preserves the column-major // ordering of the matrix elements. void copyElemsFromArrayOfMatrices (const ArrayOfMatrices& source, double* dest) { // Repeat for each matrix. for (int i = 0; i < source.length(); i++) { copyElemsFromMatrix(*source[i],dest); dest += source[i]->length(); } } // Copy the elements from the source array into the array of matrices. void copyElemsToArrayOfMatrices (const double* source, ArrayOfMatrices& dest) { // Repeat for each matrix. for (int i = 0; i < dest.length(); i++) { copyElemsToMatrix(source,*dest[i]); source += dest[i]->length(); } } int getBoundType (double& lb, double& ub) { bool haslb = !mxIsInf(lb); bool hasub = !mxIsInf(ub); int btype = haslb + 3*hasub - 2*(haslb && hasub); lb = haslb ? lb : 0; ub = hasub ? ub : 0; return btype; } // Function definitions for class MatlabProgram. // ----------------------------------------------------------------- MatlabProgram::MatlabProgram (ArrayOfMatrices& variables, const ArrayOfMatrices& lowerbounds, const ArrayOfMatrices& upperbounds, const MatlabString* objFunc, const MatlabString* gradFunc, const MatlabString* iterFunc, mxArray* auxData, int m, int maxiter, double factr, double pgtol) : Program(variables.numelems(),0,0,0,0,m,maxiter,factr,pgtol), variables(variables) { x = new double[n]; lb = new double[n]; ub = new double[n]; btype = new int[n]; this->objFunc = objFunc; this->gradFunc = gradFunc; this->iterFunc = iterFunc; // Copy some pieces of information from the input arguments to the // respective program structures. copyElemsFromArrayOfMatrices(variables,x); copyElemsFromArrayOfMatrices(lowerbounds,lb); copyElemsFromArrayOfMatrices(upperbounds,ub); // Set the bound types. for (int i = 0; i < n; i++) btype[i] = getBoundType(lb[i],ub[i]); // Create a Matlab array for the variables. int nv = variables.length(); varInputs = new mxArray*[nv]; varMatlab = new ArrayOfMatrices(varInputs,variables); // Set up the inputs to the objective callback function. numInputsObjFunc = nv + (bool) auxData; inputsObjFunc = new mxArray*[numInputsObjFunc]; copymemory(varInputs,inputsObjFunc,nv); if (auxData) inputsObjFunc[nv] = auxData; // Set up the inputs to the gradient callback function. numInputsGradFunc = numInputsObjFunc; inputsGradFunc = inputsObjFunc; // If necessary, set up the inputs to the iterative callback // function. The first two inputs are the iteration and the function // objective. The next of inputs are the values of the variables // then, optionally, the auxiliary data. if (iterFunc) { numInputsIterFunc = 2 + nv + (bool) auxData; inputsIterFunc = new mxArray*[numInputsIterFunc]; mxArray** ptr = inputsIterFunc; tMatlab = new MatlabScalar(*ptr++,0); fMatlab = new MatlabScalar(*ptr++,0); copymemory(varInputs,ptr,nv); ptr += nv; if (auxData) *ptr = auxData; } } MatlabProgram::~MatlabProgram() { delete[] x; delete[] lb; delete[] ub; delete[] btype; // Deallocate the Matlab arrays. int nv = variables.length(); for (int i = 0; i < nv; i++) mxDestroyArray(varInputs[i]); delete varMatlab; delete[] varInputs; // If necessary, deallocate the additional Matlab arrays that act as // inputs to the iterative callback function. if (iterFunc) { delete fMatlab; delete tMatlab; // delete[] inputsIterFunc; } } double MatlabProgram::computeObjective (int n, double* x) { int nlhs = 1; // The number of outputs from Matlab. mxArray* plhs[nlhs]; // The outputs from the Matlab routine. // Copy the current value of the optimization variables. copyElemsToArrayOfMatrices(x,*varMatlab); // Call the designated Matlab routine for evaluating the objective // function. It takes as input the values of the variables, and // returns a single output, the value of the objective function. if (mexCallMATLAB(nlhs,plhs,numInputsObjFunc,inputsObjFunc,*objFunc)) throw MatlabException("Evaluation of objective function failed in \ call to MATLAB routine"); // Get the result passed back from the Matlab routine. MatlabScalar matlabOutput(plhs[0]); double objective = matlabOutput; // Free the dynamically allocated memory. mxDestroyArray(plhs[0]); return objective; } void MatlabProgram::computeGradient (int n, double* x, double* g) { int nlhs = variables.length(); // The number of outputs from Matlab. mxArray* plhs[nlhs]; // The outputs from the Matlab routine. // Copy the current value of the optimization variables. copyElemsToArrayOfMatrices(x,*varMatlab); // Call the designated Matlab routine for computing the gradient // of the objective function. It takes as input the values of the // variables, and returns as many outputs corresponding to the // partial derivatives of the objective function with respect to // the variables at their curent values. if (mexCallMATLAB(nlhs,plhs,numInputsGradFunc,inputsGradFunc,*gradFunc)) throw MatlabException("Evaluation of objective gradient failed in \ call to MATLAB routine"); // Get the result passed back from the Matlab routine. ArrayOfMatrices matlabOutput((const mxArray**) plhs,nlhs); for (int i = 0; i < variables.length(); i++) if (matlabOutput[i]->length() != variables[i]->length()) throw MatlabException("Invalid gradient passed back from MATLAB \ routine"); copyElemsFromArrayOfMatrices(matlabOutput,g); // Free the dynamically allocated Matlab outputs. for (int i = 0; i < variables.length(); i++) mxDestroyArray(plhs[i]); } void MatlabProgram::iterCallback (int t, double* x, double f) { if (iterFunc) { // Copy the current iteration, the current value of the objective // function, and the current value of the optimization variables. *tMatlab = t; *fMatlab = f; copyElemsToArrayOfMatrices(x,*varMatlab); // Call the Matlab iterative callback routine. Since there are no // outputs, there is no memory to deallocate. if (mexCallMATLAB(0,0,numInputsIterFunc,inputsIterFunc,*iterFunc)) throw MatlabException("Call to MATLAB iterative callback function \ failed"); } } SolverExitStatus MatlabProgram::runSolver() { SolverExitStatus status; // The return value. status = Program::runSolver(); // Copy the solution to the class member "variables". copyElemsToArrayOfMatrices(x,variables); return status; } gpml-matlab-v3.0-2010-07-23/mean/meanConst.m000644 000765 000024 00000001135 11422240324 020070 0ustar00carlstaff000000 000000 function A = meanConst(hyp, x, i) % Constant mean function. The mean function is parameterized as: % % m(x) = c % % The hyperparameter is: % % hyp = [ c ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-01-10. % % See also MEANFUNCTIONS.M. if nargin<2, A = '1'; return; end % report number of hyperparameters if numel(hyp)~=1, error('Exactly one hyperparameter needed.'), end c = hyp; if nargin==2 A = c*ones(size(x,1),1); % evaluate mean else A = ones(size(x,1),1); % derivative end gpml-matlab-v3.0-2010-07-23/mean/meanSum.m000644 000765 000024 00000003424 11422240324 017551 0ustar00carlstaff000000 000000 function A = meanSum(mean, hyp, x, i) % meanSum - compose a mean function as the sum of other mean functions. % This function doesn't actually compute very much on its own, it merely does % some bookkeeping, and calls other mean functions to do the actual work. % % m(x) = \sum_i m_i(x) % % Copyright (c) by Carl Edward Rasmussen & Hannes Nickisch 2010-06-04. % % See also MEANFUNCTIONS.M. for ii = 1:numel(mean) % iterate over mean functions f = mean(ii); if iscell(f{:}), f = f{:}; end % expand cell array if necessary j(ii) = cellstr(feval(f{:})); % collect number hypers end if nargin<3 % report number of parameters A = char(j(1)); for ii=2:length(mean), A = [A, '+', char(j(ii))]; end; return end [n,D] = size(x); v = []; % v vector indicates to which mean parameters belong for ii = 1:length(mean), v = [v repmat(ii, 1, eval(char(j(ii))))]; end if nargin==3 % compute mean vector A = zeros(n,1); % allocate space for ii = 1:length(mean) % iteration over summand functions f = mean(ii); if iscell(f{:}), f = f{:}; end % expand cell array if needed A = A + feval(f{:}, hyp(v==ii), x); % accumulate means end else % compute derivative vector ii = v(i); % which mean function j = sum(v(1:i)==ii); % which parameter in that mean f = mean(ii); if iscell(f{:}), f = f{:}; end % dereference cell array if necessary A = feval(f{:}, hyp(v==ii), x, j); % compute derivative endgpml-matlab-v3.0-2010-07-23/mean/meanZero.m000644 000765 000024 00000000603 11422240324 017720 0ustar00carlstaff000000 000000 function A = meanZero(hyp, x, i) % Zero mean function. The mean function does not have any parameters. % % m(x) = 0 % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-01-10. % % See also MEANFUNCTIONS.M. if nargin<2, A = '0'; return; end % report number of hyperparameters A = zeros(size(x,1),1); % derivative and mean gpml-matlab-v3.0-2010-07-23/mean/meanPow.m000644 000765 000024 00000001721 11422240324 017550 0ustar00carlstaff000000 000000 function A = meanPow(mean, hyp, x, i) % meanPow - compose a mean function as the power of another mean function. % % The degree d has to be a strictly positive integer. % % m(x) = m_0(x) ^ d % % This function doesn't actually compute very much on its own, it merely does % some bookkeeping, and calls other mean function to do the actual work. % % Copyright (c) by Carl Edward Rasmussen & Hannes Nickisch 2010-06-18. % % See also MEANFUNCTIONS.M. d = mean{1}; d = abs(floor(d)); d = max(d,1); % positive integer degree mean = mean{2}; if ~iscell(mean), mean = {mean}; end if nargin<3 % report number of parameters A = feval(mean{:}); return end [n,D] = size(x); if nargin==3 % compute mean vector A = feval(mean{:},hyp,x).^d; else % compute derivative vector A = ( d*feval(mean{:},hyp,x).^(d-1) ).* feval(mean{:},hyp,x,i); endgpml-matlab-v3.0-2010-07-23/mean/meanMask.m000644 000765 000024 00000002371 11422240324 017700 0ustar00carlstaff000000 000000 function A = meanMask(mean, hyp, x, i) % Apply a mean function to a subset of the dimensions only. The subset can % either be specified by a 0/1 mask by a boolean mask or by an index set. % % This function doesn't actually compute very much on its own, it merely does % some bookkeeping, and calls another mean function to do the actual work. % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-15. % % See also MEANFUNCTIONS.M. mask = mean{1}(:); % either a binary mask or an index set mean = mean(2); % mean function to be masked nh_string = feval(mean{:}); % number of hyperparameters of the full mean if max(mask)<2 && length(mask)>1, mask = boolean(mask); end % convert 1/0->T/F if islogical(mask), D = sum(mask); else D = length(mask); end % masked dimension if nargin<3, A = num2str(eval(nh_string)); return; end % number of parameters if eval(nh_string)~=length(hyp) % check hyperparameters error('number of hyperparameters does not match size of masked data') end if nargin==3 A = feval(mean{:}, hyp, x(:,mask)); else % compute derivative A = feval(mean{:}, hyp, x(:,mask), i); endgpml-matlab-v3.0-2010-07-23/mean/meanLinear.m000644 000765 000024 00000001255 11422240324 020217 0ustar00carlstaff000000 000000 function A = meanLinear(hyp, x, i) % Linear mean function. The mean function is parameterized as: % % m(x) = sum_i a_i * x_i; % % The hyperparameter is: % % hyp = [ a_1 % a_2 % .. % a_D ] % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-01-10. % % See also MEANFUNCTIONS.M. if nargin<2, A = 'D'; return; end % report number of hyperparameters [n,D] = size(x); if any(size(hyp)~=[D,1]), error('Exactly D hyperparameters needed.'), end a = hyp; if nargin==2 A = x*a; % evaluate mean else A = x(:,i); % derivative endgpml-matlab-v3.0-2010-07-23/mean/meanOne.m000644 000765 000024 00000000601 11422240324 017520 0ustar00carlstaff000000 000000 function A = meanOne(hyp, x, i) % One mean function. The mean function does not have any parameters. % % m(x) = 1 % % Copyright (c) by Carl Edward Rasmussen and Hannes Nickisch, 2010-07-15. % % See also MEANFUNCTIONS.M. if nargin<2, A = '0'; return; end % report number of hyperparameters A = ones(size(x,1),1); % derivative and mean gpml-matlab-v3.0-2010-07-23/mean/meanProd.m000644 000765 000024 00000003655 11422240324 017717 0ustar00carlstaff000000 000000 function A = meanProd(mean, hyp, x, i) % meanProd - compose a mean function as the product of other mean functions. % This function doesn't actually compute very much on its own, it merely does % some bookkeeping, and calls other mean functions to do the actual work. % % m(x) = \prod_i m_i(x) % % Copyright (c) by Carl Edward Rasmussen & Hannes Nickisch 2010-06-04. % % See also MEANFUNCTIONS.M. for ii = 1:numel(mean) % iterate over mean functions f = mean(ii); if iscell(f{:}), f = f{:}; end % expand cell array if necessary j(ii) = cellstr(feval(f{:})); % collect number hypers end if nargin<3 % report number of parameters A = char(j(1)); for ii=2:length(mean), A = [A, '+', char(j(ii))]; end; return end [n,D] = size(x); v = []; % v vector indicates to which mean parameters belong for ii = 1:length(mean), v = [v repmat(ii, 1, eval(char(j(ii))))]; end A = ones(n,1); % allocate space if nargin==3 % compute mean vector for ii = 1:length(mean) % iteration over summand functions f = mean(ii); if iscell(f{:}), f = f{:}; end % expand cell array if needed A = A.*feval(f{:}, hyp(v==ii), x); % accumulate means end else % compute derivative vector ii = v(i); % which mean function j = sum(v(1:i)==ii); % which parameter in that mean for jj = 1:length(mean) f = mean(jj); if iscell(f{:}), f = f{:}; end % dereference cell array if necessary if jj==ii A = A .* feval(f{:}, hyp(v==jj), x, j); % multiply derivative else A = A .* feval(f{:}, hyp(v==jj), x); % multiply mean end end endgpml-matlab-v3.0-2010-07-23/mean/meanScale.m000644 000765 000024 00000001575 11422240324 020041 0ustar00carlstaff000000 000000 function A = meanScale(mean, hyp, x, i) % meanScale - compose a mean function as a scaled version of another one. % % m(x) = a * m_0(x) % % The hyperparameter is: % % hyp = [ a ] % % This function doesn't actually compute very much on its own, it merely does % some bookkeeping, and calls other mean function to do the actual work. % % Copyright (c) by Carl Edward Rasmussen & Hannes Nickisch 2010-07-15. % % See also MEANFUNCTIONS.M. if nargin<3 % report number of parameters A = [feval(mean{:}),'+1']; return end [n,D] = size(x); a = hyp(1); if nargin==3 % compute mean vector A = a*feval(mean{:},hyp(2:end),x); else % compute derivative vector if i==1 A = feval(mean{:},hyp(2:end),x); else A = a*feval(mean{:},hyp(2:end),x,i-1); end end