Neural networks, clustering techniques, and function approximation problems

Abstract

To date, clustering techniques have always been oriented to solve classification and pattern recognition problems. However, some author shave applied them unchanged to construct initial models for function approximators. Nevertheless, classification and function approximation problems present quite different objectives. Therefore it is necessary to design new clustering algorithms specialized in the problem of function approximation. © Springer-Verlag Berlin Heidelberg 2002.