Remarks on multi layer neural networks involving chaos neurons

Issue title: ICMIT '01

Article type: Research Article

Authors: Takahashi, Kazuhiko^a

Affiliations: [a] Yamaguchi University, 2-16-1 Tokiwadai Ube Yamaguchi 755-8611, Japan. Tel.: +81 836 85 9132; Fax: +81 836 85 9101; E-mail: kylyn@yamaguchi-u.ac.jp

Abstract: This paper investigates multi layer neural networks that involves the chaos neurons to achieve high efficiency of neural networks learning. Two types of neural networks are proposed. One of the multi layer neural network has the chaos neural network in its hidden layer. The other multi layer neural network is composed only using the chaos neurons. The learning ability of the neural networks involving the chaos neurons and/or chaos neural network is evaluated by benchmark experiment with XOR problems. Experimental results show the feasibility of the chaos neurons to improve the learning of neural networks. As an engineering application example, the multi layer chaos neural network is applied for identifying nonlinear dynamical systems and its effectiveness is demonstrated.

DOI: 10.3233/JAE-2003-283

Journal: International Journal of Applied Electromagnetics and Mechanics, vol. 18, no. 1-3, pp. 165-176, 2003