Neural Network Research Using Particle Swarm Optimization

Author

Wang, Yahui ; Xia, Zhifeng ; Huo, Yifeng

Author_Institution

Sch. of Electron. & Inf. Eng., Beijing Univ. of Civil Eng. & Archit., Beijing, China

fYear

2011

fDate

17-18 Sept. 2011

Firstpage

407

Lastpage

410

Abstract

In view of the artificial neural network weights training problem, this paper proposed a method to optimize the network´s structure parameters and regularization coefficient using two-layer Particle Swarm Optimization (PSO). This algorithm was applied to train Adaline network. Compared with fixed regularization coefficient method and Sliding Mode Variable Structure optimization method, the result showed that it had the advantages of high precision and strong ability of generalization.

Keywords

generalisation (artificial intelligence); learning (artificial intelligence); neural nets; particle swarm optimisation; variable structure systems; Adaline network; artificial neural network weight training problem; fixed regularization coefficient method; generalization; network structure parameter optimization; particle swarm optimization; sliding mode variable structure optimization method; Algorithm design and analysis; Educational institutions; Optimization; Particle swarm optimization; Signal processing algorithms; Testing; Training; Neural network; Regularization; Two-layer Particle Swarm;

fLanguage

English

Publisher

ieee

Conference_Titel

Internet Computing & Information Services (ICICIS), 2011 International Conference on

Conference_Location

Hong Kong

Print_ISBN

978-1-4577-1561-7

Type

conf

DOI

10.1109/ICICIS.2011.106

Filename

6063283