Chaotic Based PSOs Combined with BP for Training Fuzzy Neural Network

Training neural networks is a complex and important problem in the supervised learning field of research. In this work we tackle this problem bycombining back propagation and some chaoticparticle swarm optimization algorithms. This study considers a fuzzy neural network (FNN) structure foridentifying nonlinear dynamic systems. We used a multilayered connectionist network for realizing fuzzy inference using dynamic fuzzy rules. Fortraining, we used BP for tuning output layer weights and three different chaotic based PSO algorithms to adjust membership functions in the second layer.Then, the FNN is applied in several simulations (identification of a nonlinear time-delayed system) with regard to three different hybrid algorithms. Finally, we compared results of these training algorithms.