Fuzzy Neural Networks Design Methods Based on Swarm Intelligent Optimization Algorithms and Its Application

Abstract

The structure and parameters of fuzzy neural networks (FNN) are analyzed and a Boolean variable is proposed to network as the structure parameter. Then the question of FNN design is transformed to a function optimization question with multi-parameters. A new hybrid swarm intelligent optimization algorithm is proposed, and its binary-coded form (BIOA) and real-coded form (RIOA) are presented. BIOA and RIOA are applied to cooperative optimize the structure parameters and premise parameters of the FNN. The conclusion parameters of the FNN are optimized by least square error algorithm after the premise parameters are obtained. In the experiment, the number of sunspots is modeled by FNN and the results show that the designed FNN not only has simpler structure, but has higher precision and generalization.

Keywords

• |fuzzy neural network (fnn)|swarm intelligent optimization algorithm|cooperative optimization|least square error algorithm|structure of network