CMAC, based on a cerebellar model initiated by Eccles,
was proposed by Albus in 1975 [7]. The model was established
by simulating the structure and functions of human cerebella.
CMAC is a self-organizing neural network with the type of
form enquiry, describing complex nonlinear functions. CMAC
regards the input mode of a system as a pointer and stores
relevant information in a storage unit. Its essence is similar to
the technique of look-up-table that is used to map complex
nonlinear functions. The method is to divide the input space
into many sections and each section is allocated to a memorizer
address. Every section stores the information that has learned
in its adjacent section(s) in a distributed way. In other words, it
is to map several sections onto one memorizer address. The
structure of CMAC is shown in Fig.3.