Feature Selection Using Radial Basis Function Networks

Abstract

A new method of feature selection using a Radial Basis Function network is described. The parameters of the radial basis function network, in general, form a compact description of class structures. The intraclass and interclass distances are expressed in terms of the parameters of the trained network, and two different feature evaluation indices are derived from these distances. The effectiveness of the algorithm is demonstrated on Iris and speech data, and a comparative study is provided with several existing techniques.