Optimization Principles, Revisited
William Bialek, PhD
John Archibald Wheeler/Battelle Professor in Physics
Princeton University
Abstract
Although we often focus on our failings, the brain can perform with remarkable efficiency. In some cases, this performance approaches fundamental limits, as with photon counting or diffraction-limited imaging in vision. Many people have explored turning these observations around and using optimization as a principle from which the behaviors and mechanisms of neural circuits can be derived, but this has been controversial. I will review arguments for and against these ideas, then show how new measurements on the statistical structure of natural signals sharpen the discussion in the case of visual motion estimation. Finally, I will present progress on the landscape of optimization, arguing that classical objections to this principle may be misplaced. Hopefully, these examples help to renew your interest in optimization as the basis for theories of neural coding and computation.
Bialek’s research has addressed a wide range of theoretical physics problems motivated by the phenomena of life, across scales ranging from single molecules to flocks of birds. He is the recipient of numerous honors, including the Max Delbruck Prize in Biological Physics from the American Physical Society and the Swartz Prize for Theoretical and Computational Neuroscience from the Society for Neuroscience. He is a Fellow of the American Physical Society and a member of the National Academy of Sciences.