A RBF network modeling approach combining GA and orthogonal optimum seeking method

This paper proposes a novel approach for the training of RBF neural networks. It combines the genetic algorithm (GA) and orthogonal optimum seeking method, taking advantage of both the GA´s global optimization ability and the orthogonal optimum seeking method that can quickly determine the numbers of the hidden nodes, to correctly determine the centers of the RBF network. Simulation results show that it is a simple method, and is more reliable and efficient than orthogonal optimum seeking method in training the RBF neural network for dynamic modeling