Jonathan Pillow
(http://pillowlab.cps.utexas.edu/~pillow/)
Tuesday 21st August 2012
Time: 4pm
B10 Seminar Room, Basement,
Alexandra House, 17 Queen Square, London, WC1N 3AR
Efficient coding without information theory: a Bayesian theory of efficient neural coding
The “efficient coding hypothesis” considers neural codes to be efficient if they maximize information transfer between stimuli and neural responses. This idea, first articulated by Attneave and Barlow in the 1950s and 60s, has provided a guiding theoretical framework for the study of coding in neural systems, and has motivated a large number of experimental and theoretical studies. More recently, theories of probabilistic neural coding and Bayesian inference, often called the "Bayesian brain" hypothesis, has attracted great interest in systems neuroscience. However, there does not appear (as yet) to be any clear connection between these two paradigms. In this talk, I will introduce a more general "Bayesian" theory of efficient coding, which has classic efficient coding as a special case. I will argue that there is nothing privileged about information-maximizing codes; they are ideal for some tasks but sub-optimal for many others.
Bayesian efficient coding substantially enlarges the family of normatively optimal codes and provides a general framework for understanding the principles of sensory encoding. I will derive Bayesian efficient codes for a few simple examples, show an application to neural data, and suggest several avenues for future research.