Input Maping and Simulation Analysis using Adaptive Network Based Fuzzy Inference System

Fuzzy logic control systems are structured numerical estimators. They combine both the numerical process and human like reasoning. Neural networks are numerical trainable dynamical systems that are able to emulate human brain functions; their connectionist structure can be used to find the proper parameters and structures that resemble human thinking rules for fuzzy logic controllers. Generally fuzzy logic is best applied to non linear, time varying, illdefined systems, which are too complex for conventional control systems to apply. In this paper a new combinational connectionist structure is proposed which exploits the advantages of both the fuzzy and neural networks avoiding the rule-matching time of the inference engine in the traditional fuzzy logic system. Some examples are presented using MATLAB simulation to illustrate the performance and applicability of the proposed connectionist model. Keywords— Fuzzifier, membership function, receptive field, hybrid learning, adaptivity, input-output mapping, ANFIS,training, epoch