Eliminating Spurious Memories in a Network of Chaotic Elements

Issue title: Chaos Engineering

Article type: Research Article

Authors: Ishii, Shin

Affiliations: ATR Human Information Processing Research Laboratories, 2-2 Hikaridai, Seika-cho, Soraku-gun, Kyoto 619-02, Japan, e-mail: ishii@hip.atr.co.jp

Abstract: A globally coupled map (GCM) model is a network of chaotic elements that are globally coupled with each other. We previously proposed an associative memory system based on GCM, which has a better ability than the Hopfield network. This result indicates that the dynamics of our system is more efficient than that of the Hopfield network. However, even in our system, spurious memories, that is, the system's equilibria that do not correspond to any of the proper memories, do exist. In this article a modified associative memory system is proposed in which spurious memories are noticeably reduced. This is achieved by modifying the chaotic dynamics of the system. With this improvement, our system's memory capacity and basin volumes are expanded a great deal. Some experimental results in comparison with those of a neural network that employs a nonmonotonic output junction are also shown.

DOI: 10.3233/IFS-1997-5106

Journal: Journal of Intelligent and Fuzzy Systems, vol. 5, no. 1, pp. 69-83, 1997