An integrated microcircuit model
of attentional processing in the neocortex
Xiao-Jing Wang
I will discuss a joint work with Salvador Ardid and Albert Compte on a
spiking
neuron model of feature-based selective attention. We found that a wide
range
of physiological phenomena induced by selective attention arise
naturally in
a reciprocally connected loop of two (sensory and working memory)
networks.
Our model instantiates the `feature-similarity gain principle', and
provides a synthetic account for biased competition and multiplicative
gain
modulation. The underlying circuit mechanism critically depends on an
interplay between selective top-down excitation from the working memory
network, feedback inhibition in the sensory network, and power-law
input-output relationship of neurons. This work supports the notion
that a
common circuit subtrate subserves both working memory and selective
attention.