Optimizing the fuzzy adaptive learning by the gradient descent approach

摘要

A systematic approach to optimize the fuzzy rules and the corresponding membership functions to solve the local minimum problem is proposed. Without resorting to a good choice of the initial parameters, the presented technique can satisfactorily reach the desired results. The simulation results demonstrate that the proposed approach outperforms those exploit either neural modeling or symmetric adjustment methods. Examples are provided to verify the superiority of the proposed technique.