You are here

Working memory for multidimensional functions

This project is an extension of the dynamic working memory example presented in section 8.3 of Neural Engineering (Eliasmith and Anderson, 2003), much of the system description overlaps. The subpopulation under consideration in that case was the Lateral Intraparietal Area (LIP) of the neocortex of macaque monkeys. This population exhibits the behaviour of storing memories of salient stimuli, and has been studied extensively by researchers. The results of these studies indicate that multiple bumps of varying heights can be encoded by the LIP to represent multiple stimuli in the spatial field (represented by v), as well as a non-spatial characteristic of each stimulus f(v). Research by Colby and Goldberg[2] suggests that f(v) represents the attentional resources given to the stimuli, while research by Andersen et al.[3] suggests that it represents intention to move to the object.
Peer Reviewed: 
Corneil and Gevaert, 2009, Working memory for multidimensional functions. Tech Report for SYDE 556. University of Waterloo
Publication files: 

This model and report compares 1D and 2D working memory in terms of RMS and stability. It also incorporates methods for running function representation simulations in Nengo. Here are some cool movies referred to in the report: spikes movie, 2000 neurons, 40 dimensions and 20 dimensions.

spikes movie

2000 neurons 40 dimensions

40 dimensions

20 dimensions