Instance Selection for Kernel Classifiers

Gerhard Paass
GMD - German National Research Center for Information Technology 
Schloss Birlinghoven, 
D - 53754 Sankt Augustin, 
Germany 
paass@gmd.de

The active selection of instances can significantly improve the
generalization performance of a training algorithm. Kernel classifiers
such as support vector machines orBayesian kernel classifiers
classify data using the most informative data instances (support
vectors). This makes them natural candidates for instance selection
procedures. We survey instance selection strategies proposed in the
literature and discuss their application to Bayesian kernel
classifiers.