ePrints@IIScePrints@IISc Home | About | Browse | Latest Additions | Advanced Search | Contact | Help

A neural network for storing individual patterns in limit cycles

Deshpande, Varsha and Dasgupta, Chandan (1991) A neural network for storing individual patterns in limit cycles. In: Journal of Physics A : Mathematical and General, 24 (21). pp. 5105-5119.

[img] PDF
A_neural_network_for_storing.pdf - Published Version
Restricted to Registered users only

Download (711kB) | Request a copy
Official URL: http://iopscience.iop.org/0305-4470/24/21/022/

Abstract

A neural network model in which individual memories are stored in limit cycles is studied analytically and numerically. In this model there are two kinds of interactions: a Hopfield-like term that tends to stabilize the system in a memorized state and a second term with a time delay that acts to induce transitions between a memorized state and its complement state. For a proper choice of the values of the parameters, this model exhibits limit cycle behaviour in which the overlap with a target pattern oscillates in time. An asymmetrically diluted version of the model is studied analytically in the limit of extreme dilution. We find that the model with cycles performs better than a similarly diluted version of the Hopfield model. The performance of the fully connected model is studied by numerical simulations. We find a behaviour qualitatively similar to that of the dilute model. The model with cycles is found to perform better than the Hopfield model as a pattern classifier if the memory loading level and the degree of corruption of the input patterns are high.

Item Type: Journal Article
Publication: Journal of Physics A : Mathematical and General
Publisher: IOP Publishing
Additional Information: Copyright of this article belongs to IOP Publishing.
Department/Centre: Division of Physical & Mathematical Sciences > Physics
Date Deposited: 13 Jul 2007
Last Modified: 27 Jan 2012 06:33
URI: http://eprints.iisc.ac.in/id/eprint/11548

Actions (login required)

View Item View Item