Stability conditions of Hopfield ring networks with discontinuous piecewise-affine activation functions

Abstract : Ring networks, a particular form of Hopfield neural networks, can be used in computational neurosciences in order to model the activity of place cells or head-direction cells. The behaviour of these models is highly dependent on their recurrent synaptic connectivity matrix and on individual neurons' activation function, which must be chosen appropriately to obtain physiologically meaningful conclusions. In this article, we propose some simpler ways to tune this synaptic connectivity matrix compared to existing literature so as to achieve stability in a ring attractor network with a piece-wise affine activation functions, and we link these results to the possible stable states the network can converge to.
Type de document :
Communication dans un congrès
56th IEEE Conference on Decision and Control, CDC 2017, Dec 2017, Melbourne, Australia
Liste complète des métadonnées

Littérature citée [14 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-01645410
Contributeur : Amélie Aussel <>
Soumis le : jeudi 23 novembre 2017 - 09:13:09
Dernière modification le : jeudi 30 novembre 2017 - 13:52:39

Fichier

CDC_Conference_Paper_VF_.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01645410, version 1

Citation

Amélie Aussel, Laure Buhry, Radu Ranta. Stability conditions of Hopfield ring networks with discontinuous piecewise-affine activation functions. 56th IEEE Conference on Decision and Control, CDC 2017, Dec 2017, Melbourne, Australia. 〈hal-01645410〉

Partager

Métriques

Consultations de la notice

86

Téléchargements de fichiers

8