Function Approximation With Continuous-Valued Activation Functions in CMAC

CMAC (Albus, 1975b) is well known as a good function approximator with local generalization abilities. Depending on the smoothness of the function to be approximated, the resolution as the smallest distinguishable part of the input domain, plays a crucial role. If the usually used binary quantizing functions are dropped in favor of more general, continuous-valued functions, this drawback can be remedied. We introduce such a model, using gaussians instead of binary functions. We apply the method to approximate functions. It outperforms the traditional approach with respect to training time, approximation accuracy, parameter insensitivity and memory consumption.