On the training of a multi-resolution CMAC neural network

Several artificial neural network architectures have been proposed to solve problems in control systems and system identification. However, not all neural network structures are equally attractive for real-time adaptive situations. In this paper, some advantages and disadvantages of various structures are highlighted, especially in the class of associative memory networks. Particular attention is given to the CMAC neural network and its potential for real-time applications. A hierarchical multi-resolution approach is investigated through experimentation as a possible approach to alleviate the CMAC's main drawback: the exponential growth of storage as a function of the number of inputs. Relevant issues are discussed and suggestions for future improvement are given.