A novel heuristic for building reduced-set SVMs using the self-organizing map

Abstract

We introduce a novel heuristic based on the Kohonen's SOM, called Opposite Maps, for building reduced-set SVM classifiers. When applied to the standard SVM (trained with the SMO algorithm) and to the LS-SVM method, the corresponding reduced-set classifiers achieve equivalent (or superior) performances than standard full-set SVMs. © 2011 Springer-Verlag.