Nonlinear System Modeling Based on IFCNN

摘要

This paper for the shortcomings of conventional BP algorithm which has slow convergence and falls into local minimum easily, the nonlinear self-feedback term is introduced into this algorithm. Thus chaotic BP algorithm(CBPA) is given. The weight of fuzzy neural network(FNN) is trained and learned by using it. Thus an introduction-type fuzzy chaotic neural network(IFCNN) is constituted. Then simulation of nonlinear system based on IFCNN given is proposed. Simulation results show that the designed IFCNN has the same and complex dynamic characteristics with chaotic system, which has good modeling capabilities for nonlinear system. And with the chaotic BP algorithm training parameters, it has fast convergence, mixed search capability, being able to be out of local minimum.