Wavelet neural network based on improved particle swarm algorithm

In allusion to the shortcoming, easily falling into the local optimum, of basic particle swarm algorithm, this paper proposes an improved particle swarm algorithm, and applies it to wavelet neural network to optimize each parameter of the wavelet neural network. New algorithm improves basic particle swarm algorithm from three aspects: firstly, introduce inertial weight factor, and use linearly decreasing weight strategy to weigh two aspects, the convergence precision and convergence rate, of the search capability; secondly, use individual average extremum instead of individual extrema to expand the cognition scope of the particles, which makes the particles can obtain more information to adjust own state; finally, introduce the thought of cross in the genetic algorithm to keep diversity of particle swarm, in order to ensure that it is not easy to fall into the local optimum for the algorithm. The simulation results show that the wavelet neural network based on improved particle swarm algorithm has very good approximation ability and convergence speed.