CRPSO-Based Integrate-and-Fire Neuron Model for Time Series Prediction

Abstract

Single Integrate-and-Fire neuron (IFN) model is used for time series prediction recently in which a multilayer neural network is always utilized. An improved particle swarm optimization (PSO) algorithm named cooperative random learning particle swarm optimization (CRPSO) algorithm is put forward to training the IFN model in order to enhance its approximation and generalization capabilities. The proposed CRPSO-based IFN model is used for Mackey-Glass time series prediction problem. The experimental results demonstrate the superiority of CRPSO-based model in efficiency and robustness over the PSO algorithm, BP algorithms and GA.