Fuzzy CMAC structures

Abstract

Cerebellum Model Articulation Controller (CMAC) is known as a feedforward Neural Network (NN) with fast learning and performance. Many improvements have been introduced to it which fuzzy CMAC (FCMAC) is the most important one. Fuzzy CMAC as a neuro fuzzy system increases precision, reduces memory size and makes CMAC differentiable. In addition FCMAC converts CMAC NN as a black box to a white box that its operation is interpretable using fuzzy rules. Fuzzy CMAC has not a unique structure in literature and there are differences in many aspects as membership function, memory layered structure, deffuzification and the fuzzy system applied. Discussing these, this paper reviews fuzzy CMAC different structures in literature.