ABSTRACT
This chapter explores the functions of lateral inhibition in the transformation and short-term storage of patterns in model neural networks. It focuses on the dynamics of competition between inputs, particularly at the level of sensory areas of the cerebral cortex. Contrast enhancement is an outgrowth of decision or competition between inputs. Competition can be biased in favor of either more intense or less intense inputs by nonlinear interactions. Short-term memory in recurrent lateral inhibitory networks has been modeled since the early 1970s, particularly by D. Wilson and J. D. Cowan, S. Grossberg, S. I. Amari, and their colleagues. The distance-dependent networks of Wilson and Cowan have the same range of limiting behavior as those of their earlier article, despite greater mathematical complexity. The implementation of shunting recurrent lateral inhibition in Grossberg was motivated by some heuristics relating to shunting nonrecurrent inhibition.
