Dynamic Neural Field with Local Inhibition

Nicolas P. Rougier (1)

1 CORTEX - Neuromimetic intelligence

A lateral-inhibition type neural field model with restricted connections is presented here and represents an experimental extension of the Continuum Neural Field Theory (CNFT) by suppression of the global inhibition. A modified CNFT equation is introduced and allows for a locally defined inhibition to spatially expand within the network and results in a global competition extending far beyond the range of local connections by virtue of diffusion of inhibition. The resulting model is able to attend to a moving stimulus in the presence of a very high level of noise, several distractors or a mixture of both.

Domaines: Réseau de neurones [cs.NE]

Nicolas P. Rougier. Dynamic Neural Field with Local Inhibition. Biological Cybernetics (Modeling), 2006, 94 (3), pp.169-179. ⟨10.1007/s00422-005-0034-8⟩. ⟨inria-00000244⟩