Self-structuring fuzzy systems for function approximation

Abstract

This paper presents an algorithm developed in a biological spirit and dedicated to the incremental building of fuzzy systems for function approximation. It is called EFUSS (evolving fuzzy systems structure) and aims at automatically and incrementally finding the minimal number of membership functions along with their appropriate shaping. The main mechanisms constituting our algorithm are to: observe the oscillatory tendency of the parameters defining the output part of the fuzzy rules, then detect the most oscillatory one, and finally supply the zone covered by the input of this strongly oscillating rule with a complementary fuzzy rule.<>