On the function approximation in Restricted Coulomb Energy neural network with Gaussian Radial Basis Function

The ability of function approximation in Restricted Coulomb Energy(RCE) neural network is explored in this paper. The output neurons in the presented RCE network adopt Gaussian Radial Basis Function (RBF) as their activation functions, thus the desired function can be smoothly approximated. The performance of the presented RCE network is investigated through approximations of two- and three-dimensional function, as compared with the conventional multilayer perceptron.