An improved particle swarm optimization algorithm for radial basis function neural network

Abstract

An improved particle swarm optimization (IMPSO) which synthesizes the existing models of constriction factor approach (CFA PSO) is proposed. In the proposed method, an adaptive algorithm based on the search space adjustable is applied to solve the problem that conventional particle swarm optimization (PSO) algorithm easily falls into local optimal and occur premature convergence. Then, the IMPSO is used to optimize the parameters of RBF neural network. The new training algorithm is used to approximate polynomial function and predict chaotic time series, compared with PSO, and CFA PSO, the algorithm speed up the speed of convergence, and has much greater accuracy.