IMPLEMENTATION OF A FUZZY INFERENCE SYSTEM USING A NORMALIZED FUZZY NEURAL-NETWORK

Abstract

In this paper, we present a normalized fuzzy neural network (NFNN) to implement fuzzy inference systems. The proposed NFNN architecture makes an effective rule combination technique possible and thus enables us to significantly reduce the number of rules in the NFNN. We also derive a sufficient condition for rule combination and provide an algorithm to perform rule combination. Simulation results show that when combined with a rule elimination method the proposed rule combination method can greatly reduce the number of rules in the NFNN.