HAL : hal-00331551, version 1

Fiche détaillée

Récupérer au format BibTeX - EndNote - TEI - RefWorks -

Deuxième conférence française de Neurosciences Computationnelles, "Neurocomp08", Marseille : France (2008)

An Empirical Evaluation Framework for Qualifying Dynamic Neural Fields

Lucian Alecu (  Auteur correspondant ) 1, 2, Hervé Frezza-Buet 2

(08/10/2008)

In this paper, the behavior of dynamic neural fields is studied through the lens of performance. As an alternative to the currently available analytical instruments, an empirical evaluation methodology is proposed in order to examine the dynamic quality of a field. This consists of simulating the field through various key scenarios and compare the observed behavior to an optimal expected one. Some desired effects concerning the evolution of an ideal field are inspected, and a performance criterion is defined accordingly. Practically, this approach implements a generic benchmark framework for qualifying neural fields, allowing to inspect the evolution of the model in different key situations. The presented methodology provides a basis for a methodological computational approach towards adjusting the free parameters of the fields in order to satisfy specific desired properties.

1 :	CORTEX (INRIA Lorraine - LORIA)
	INRIA – CNRS : UMR7503 – Université Henri Poincaré - Nancy I – Université Nancy II – Institut National Polytechnique de Lorraine (INPL)
2 :	SUPELEC-Campus Metz
	SUPELEC

Domaine

:

Sciences du Vivant/Neurosciences Sciences cognitives/Neurosciences

Mots Clés : Neuronal models – Simulation

hal-00331551, version 1

http://hal.archives-ouvertes.fr/hal-00331551

oai:hal.archives-ouvertes.fr:hal-00331551

Contributeur : Laurent Perrinet <>

Soumis le : Vendredi 17 Octobre 2008, 10:33:46

Dernière modification le : Vendredi 24 Octobre 2008, 13:49:30