The Center for Cognitive Computation (CCC) invites you to the lecture of Professor Gašper Tkačik (Institute of Science and Technology Austria).
Title: Statistical analysis, optimality, and efficient attentional modulation
Abstract: This talk consists of two parts. In the first part, I will briefly outline a way to unify normative (optimization) theories of biological system function with statistical inference. This is achieved a coherent Bayesian framework that embeds a normative theory into a family of maximum-entropy ‘‘optimization priors.’’ This family defines a smooth interpolation between a data-rich inference regime and a data-limited prediction regime. I will use simple examples from neuroscience to show how this framework allows one to address a number of fundamental statistical challenges relating to inference in high-dimensional, biological problems.
In the second part, I will present recent work on efficient sensory coding in systems with top-down feedback. In a normative model of dynamic population coding in visual cortex, a perceptual observer constantly and optimally tunes the sensory population to solve a perceptual inference task while maintaining coding efficiency. This optimal behavior leads to attention-like modulation and explains a number of seemingly disparate cortical phenomena.