An optimization algorithm for WNN based on immune particle swarm

Abstract

Wavelet neural network (WNN) is a combination of wavelet analysis and neural network and has the strong fault tolerance, the strong anti-jamming and the strong adaptive ability. However, WNN is likely to trap local minimum and premature convergence. According to these shortcomings, particle swarm optimization (PSO) algorithm is applied to wavelet neural network (WNN) and has good effect. This paper presents a PSO algorithm based on artificial immune (AI). Through importing antibody diversity keeping mechanism, this algorithm can retain high fitness of particles and ensure the diversity of population. Then, the new algorithm is applied to the training of WNN and the parametric optimization. Through some simulation experiments, this paper concludes that the presented algorithm has stronger convergence and stability than the basic particle swarm optimization algorithm on optimizing WNN, and has the better performance of reducing the number of training and error. © 2011 Springer-Verlag.