An Improved RBF Neural Network Based on Evolutionary Programming

Abstract

For the problem of local minimum for gradient descent method used to train an RBF (Radial Basis Function) neural network, EP (evolutionary programming) is introduced to the training of RBF neural network in this paper. The combination method of EP and gradient descent method can effectively avoid local minimum, and provides a more reasonable network design. The effectiveness of the proposed scheme is demonstrated by the simulation of a nonlinear system control.