An Improved Hybrid Algorithm Based on PSO and BP for Feedforward Neural Networks

International Journal of Digital Content Technology and its Applications. Volume 5, Number 2, January 2011

An Improved Hybrid Algorithm Based on PSO and BP for Feedforward Neural Networks

Fei Han, Tong-Yue Gu, Shi-Guang Ju

hanfei@ujs.edu.cn

gtyglx@tom.com

jushig@ujs.edu.cn

School of Computer Science and Telecommunication Engineering, Jiangsu University, Zhenjiang, Jiangsu, 212013, China

Abstract

An improved hybrid algorithm combining particle swarm optimization (PSO) with back propagation algorithm (BP) is proposed to train feedforward neural networks (FNN). While PSO is a global search algorithm, it is easy for the swarm in PSO to lose its diversity, resulting in premature convergence. Conversely, BP algorithm is a gradient-descent-based method which has good local search ability around the global minima. The algorithm presented in this study combines PSO with BP to perform double search. In addition, in order to improve the diversity of the swarm in the PSO, each particle in the swarm and its best position are disturbed by a random function, and the best position of all particles are reset as the optimum weights of FNN obtained by BP.

The proposed algorithm improves the diversity of the swarm as well as reducing the likelihood of the swarm being trapped into local minima on the error surface. Compared with the traditional learning algorithms, the improved learning algorithm has much better convergence accuracy and rate.

[EXCERPT]

To demonstrate the improved convergence performance of the proposed algorithm, in this section we conduct two experiments which are function approximation and modeling the true boiling point curve of crude oil

Free Full Text Source: http://www.aicit.org/jdcta/ppl/12_%20Binder1-14.pdf