A neural model for attentional modulation of lateral interactions in the visual cortex

Abstract

Neurophysiological investigations showed that attention influences neural responses in the visual cortex by modulating the amount of contextual interactions between cells. Attention acts as a gate that protects cells from lateral excitatory and inhibitory influences. A recurrent neural network based on dendritic inhibition is proposed to account for these findings. In the model, two types of inhibition are distinguished: dendritic and lateral inhibition. Dendritic inhibition regulates the amount of impact that surrounding cells may exert on a target cell via dendrites of excitatory neurons and dendrites of subpopulation of inhibitory neurons mediating lateral inhibition. Attention increases the amount of dendritic inhibition and prevents contextual interactions, while it has no effect on the target cell when there is no contextual input. Computer simulations showed that the proposed model reproduces the results of several studies about interaction between attention and horizontal connections in the visual cortex. © 2007 Springer-Verlag Berlin Heidelberg.