Learning the Kernel Matrix 1

olivier chapelle — Mon, 14 Jan 2008 08:50:37 -0000

Code for automatically selecting the kernel parameters of an SVM. It is based on a gradient descent minimization of either the radius/margin bound, the leave-one-out error, a validation error or the marginalized likelihood.

Included is also a special code for learning a linear combination of kernels.

mloss.org Learning the Kernel Matrix

Learning the Kernel Matrix 1