A Research on Chaotic Recurrent Fuzzy Neural Network and Its Convergence

In this paper, a type of chaotic recurrent fuzzy neural network (CRFNN) model is proposed. The CRFNN model add chaotic map in the membership function layer of a RFNN. A generalized dynamic back propagation algorithm (DBP) is developed to automatically construct the CRFNN. To guarantee the convergence by Lyapunov function, the online learning rate adjusting range is given. Simulation results of identifying chaotic system show that, CRFNN has better performance than normal method and the adaptive learning rate could improve efficiency and decrease approximation errors.