Fuzzy Systems with Learning Capability

Shigeo Abe

Abstract: In this paper, we discuss fuzzy systems with a learning capability that realize high speed training and high generalization ability. First fuzzy classifiers with ellipsoidal regions, hyperbox regions, and polyhedron regions are discussed and their performance and that of the neural network classifier are compared. Then the rule extraction for the fuzzy classifiers is extended to function approximation. Finally performance of one fuzzy system for a water purification plant is compared with that of the neural network.