Friday, June 25, 2004

12 noon

Redwood Neuroscience Institute

 

Title:   “A Sparse Distributed Model of Episodic and Semantic  Spatiotemporal Memory”

 

Rod Rinkus

Previous at Enkidu Research, Inc.

 

Abstract:

A neural network model is proposed that forms sparse spatiotemporal memory traces of spatiotemporal events given single occurrences of the events. The traces are distributed in that each individual cell and synapse participates in numerous traces. This sharing of representational substrate provides the basis for similarity-based generalization and thus semantic memory.

 

 Simulation results are provided demonstrating that similar spatiotemporal patterns map to similar traces. The model achieves this property by  measuring, on each time slice, the degree of match, G, between the actual  current input pattern and the expected input pattern given the preceding  time slices (i.e., temporal context) and then adding an amount of noise,  inversely proportional to G, to the process of choosing the internal  representation for the current time slice. Thus, if G is small, indicating novelty, we add much noise and the resulting internal representation of the current input pattern has low overlap with any preexisting representations of time slices. If G is large, indicating a familiar event, we add very little noise. This lets previously learned associations, i.e., patterns of  increased synaptic weights, predominate in the choice of internal  representation, resulting in reactivation of all or most of the preexisting  representation of the input pattern.