Tuning of the Structure and Parameters of Dynamic Process Neural Network Using Improved Chaotic PSO

This paper presents the tuning of the structure and parameters of a dynamic process neural network(DPNN) using a improved chaotic particle swarm optimization (ICPSO), the ICPSO approach is a method of combining the improved particle swarm optimization (IPSO), which has a powerful global exploration capability, with the chaotic strategy , which can exploit the local optima. By introduced a new strategy to the ICPSO, it will also be shown that the ICPSO performs better than the traditional PSO and GA based on some benchmark test functions. A PNN with switches introduce to links is proposed. By tuning the structure and improving the connection weights of PNN simultaneously, a partially connected PNN can be obtained. By doing this, it eliminates some ill effects introduced by redundant in features of PNN. An application example on iris forecasting is given to show the merits of the improved DPNN using ICPSO.