CORTICAL MEMORY DYNAMICS

Biological memories have a number of unique features, including (1) hi erarchical, reciprocally interacting layers, (2) lateral inhibitory in teractions within layers, and (3) Hebbian synaptic modifications. We i ncorporate these key features into a mathematical and computational mo del in which we derive and study Hebbian learning dynamics and recall dynamics. Introducing the construct of a feasible memory (a memory tha t formally responds correctly to a specified collection of noisy cues that are known in advance), we study stability and convergence of the two kinds of dynamics by both analytical and computational methods. A conservation law for memory feasibility under Hebbian dynamics is deri ved. An infomax net is one where the synaptic weights resolve the most uncertainty about a neural input based on knowledge of the output. Th e infomax notion is described and is used to grade memories and memory performance. We characterize the recall dynamics of the most favorabl e solutions from an infomax perspective. This characterization include s the dynamical behavior when the net is presented with external stimu li (noisy cues) and a description of the accuracy of recall. The obser ved richness of dynamical behavior, such as its initial state sensitiv ity, provides some hints for possible biological parallels to this mod el.