Home Publications A feature-binding model with localized excitations
Personal tools

Hecke Schrobsdorff, J. Michael Herrmann, and Theo Geisel (2007)

A feature-binding model with localized excitations

Neurocomputing 70(10-20):1706-1710 .  (export entry)

We study a model of feature binding in prefrontal cortex which defers specific perceptual information to lower areas and merely maintains the identity of the combination. The model consists of three layers of pulse-coupled leaky integrate-and-fire neurons. Features are encoded by the location of sustained activity in the subordinate layers. The feature layers are excitatorily coupled to a superordinate layer that represents combinations of features by means of an oscillatory dynamics. The model accounts for effects such as the memorization of an object that was perceived only for a short period, illusory binding of simultaneous stimuli, and the limit of attentional capacity. The present paper discusses conditions for localized excitations in networks of integrate-and-fire neurons and considers the application to a dynamic link architecture.