On Fuzzy Controllers Having Radial Basis Transfer Functions

Abstract

This works establish a link between certain fuzzy controllers and radial basis function networks. We deal with the two soft computing techniques and compare them from the approximation capability point of view. The fact that the approximation behaviour of these systems are similar establishes the ground of analysing their possible functional equivalence. We show that certain fuzzy systems can be considered as radial basis function (RBF) approximation scheme, having the same transfer function. This allows to implement those fuzzy systems by RBF networks, and hence, establishing a learning algorithm for them and improving their approximation capabilities. We also points out that the classes of the corresponding fuzzy system can be generalized to some extent if we use general basis function with weighted norm.