Ce court papier montre comment le modèle en double boucle est compatible avec les données physiologiques sur la mémoire.

Abstract

This article tackles the issue of the automatic constitution of mental images in a neural network, then deals with the use of these images in the two traditional physiological cases: short-term memory and long-term memory. The suggested model of learning, unsupervised and of reinforcement type, is compatible with physiological and pathological data and gives a consistent approach for both acquisition and recognition of the stimuli. The model is built around the mental object concept, which is made up of double loop signal flows circulating over a connectionist structure, and takes into account the functional role of the limbic system.

Keywords: associative memory, unsupervised learning, double loop signals flows, short and long term memory.

Reprints sur demande à cl AT mire.net