Friday, January 28, 2005

12 noon

Redwood Neuroscience Institute

 

Title "An information-theoretic min cut approach to population coding.”

 

Michael Gastpar

Department of Electrical Engineering and Computer Sciences

University of California, Berkeley

 

Abstract:

In this talk, we first present a review/overview of some of the key results in network information theory and discuss how some of these results could be relevant to the understanding of population codes. To simplify the problem, we advocate a cut-set approach in which we partition the entire network into a few blocks and study the information transfer between these blocks.

    In the second part of the talk, we discuss sensor networks, currently one of the most active research topics in engineering. The standard concepts of information theory seem insufficient to answer the fundamental coding and performance questions. We briefly outline some of the recent results, and again outline how these may be important in the analysis of population codes.