A novel cluster method in fuzzy neural networks

Ching-chang Wong et. al(1999) proposed a novel cluster method to make training sample data stepwise converge to cluster centers regardless of the predetermination of center number. This paper improves the novel cluster method, and proves its convergence by using Brouwer Fixed Point Theorem. Based on the result of the novel cluster method, one first order TSK fuzzy neural network is established and hybrid algorithm is implemented to tune network parameters. Finally, simulation results are given to demonstrate the effectiveness of this novel cluster method in fuzzy neural networks.