On the nonlinearity of pattern classifiers

1 January 1996

conference paper
Published by Institute of Electrical and Electronics Engineers (IEEE)

Vol. 4 (10514651) , 271-275 vol.4
https://doi.org/10.1109/icpr.1996.547429

Abstract

This paper presents a novel approach to the analysis of the overtraining phenomenon in pattern classifiers. A nonlinearity measure /spl Nscr/ is introduced which relates the shape of the classification function to the generalization capability of a classifier. Experiments using the k-nearest neighbour rule, a neural classifier and the quadratic classifier show that the introduced measure /spl Nscr/ can be used to study the overtraining behaviour of a classifier. Moreover /spl Nscr/ shows to be a predictor for the local sensitivity of a classifier. Classifiers that have a small local sensitivity are shown to have a low nonlinearity whereas an increased nonlinearity indicates an increase in local sensitivity.

Keywords

This publication has 2 references indexed in Scilit:

On the approximate realization of continuous mappings by neural networks
Neural Networks, 1989
Multilayer feedforward networks are universal approximators
Neural Networks, 1989