A sparse implementation of dynamic competition in continuous neural fields
Résumé
This paper introduces a sparse implementation of the Continuum Neural Field Theory, promoting a trade-off in accuracy for higher computational efficiency and alleviated constraints on the underlying model. The sparse version reproduces the main properties of previous discrete 2D implementations, such as dynamic competition leading to localized focus activity or robustness to noise and distracters, with a much higher computational speed on standard computer architectures.
Domaines
InformatiqueOrigine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...