, Volume 51, Issue 2, pp 195-205

Context-dependent associations in linear distributed memories

Rent the article at a discount

Rent now

* Final gross prices may vary according to local VAT.

Get Access

Abstract

In this article we present a method that allows conditioning of the response of a linear distributed memory to a variable context. This method requires a system of two neural networks. The first net constructs the Kronecker product between the vector input and the vector context, and the second net supports a linear associative memory. This system is easily adaptable for different goals. We analyse here its capacity for the conditional extraction of features from a complex perceptual input, its capacity to perform quasi-logical operations (for instance, of the kind of “exclusive-or”), and its capacity to structurate a memory for temporal sequences which access is conditioned by the context. Finally, we evaluate the potential importance of the capacity to establish arbitrary contexts, for the evolution of biological cognitive systems.

Part of this study has been presented in a preliminary version at the XVI Reunión Científica de la Sociedad Argentina de Biofísica, Tigre, Argentina, December 1987.