Fuzzy CMAC and its application in function learning

To avoid the shortcoming of normal CMAC which only works on some discrete points, a fuzzy cerebellar model articulation controller (FCMAC) with fuzzifying and defuzzifying process is proposed. The fuzzifying process of the input layer enables the FCMAC to accept a continuous quantum as its input. A concept mapping algorithm and learning algorithm of the FCMAC, which makes the association space rather small and the velocity of convergence rather high respectively. This is illustrated by the results of simulations given in the multiple-inputs case