The Bayesian fly: Inference based on natural sensory signals

by Rob de Ruyter, Indiana University

Date

Tomorrow, November 21

Time

3:00 p.m. — 4:00 p.m.

Place

Lindley Hall 102

Abstract:  As animals navigate the world their sense organs measure a continuous stream of complex and noisy physical signals. These measurements are used to make inferences about events in the animal's environment. Bayes' rule in statistics is the basis of a general theory of how to make such inferences, and we would like to explore biological information processing in terms of this general approach. However, in order to implement the rule one needs a statistical description of the relevant signals. In this talk I will present a strategy for sampling the joint statistics of visual signals and motion necessary for building an optimal estimator of motion based on visual data. This approach leads to specific predictions on the structure of the optimal motion estimator. Those predictions will be compared to the real biological example of motion estimation in the blowfly, where we can make recordings from motion sensitive neurons.
Biography: Robert R. de Ruyter van Steveninck is the Linda and Jack Gill Professor of Physics, and Program in Neuroscience at Indiana University, Bloomington. He received his B.S. and Ph.D. at the University Groningen, The Netherlands, and has held post doctoral positions at the Cambridge University, UK, and the University of Groningen.  He is interested in understanding basic principles underlying coding, computation and inference in the sensory nervous system.

Colloquium Provided By:

the School of Informatics