Linking perceptual uncertainty and response variability in the visual cortex
Fri, August 2, 2013 • 1:00 PM - 2:00 PM • SEA 4.242 (small conference room)
The University of Texas at Austin
Center for Perceptual Systems
"Linking perceptual uncertainty and response variability in the visual cortex"
Gergo Orban, Ph.D.
University of Cambridge
SEA 4.242 (small conference room)
Bag Lunch Talk
Find information about current and upcoming talks at CPS on our website:
Abstract: Neural responses in the visual cortex are variable, and the magnitude and structure of this variability is systematically controlled by stimulus attributes. Current theories of cortical computation either ignore variability altogether, or only focus on unstructured Poisson-like aspects of spike count statistics. We present a theory that links variability of neural responses directly to a fundamental class of computations: probabilistic inference of environmental features based on noisy and ambiguous inputs. In the theory, neural activation patterns in primary visual cortex (V1) represent stochastic samples from the probability distribution arising from probabilistic inference, the posterior probability distribution. The proposed theory contains classical theories of V1 function as a special case: it suggests that trial-averaged mean responses are equivalent to averages over a set of samples. We show that such a sampling-based probabilistic representation predicts several previously unexplained key features of the variability and covariability of membrane potentials and spike counts in V1. These include the modulation of neural variability by the onset, contrast, and orientation of stimuli, non-classical receptive field modulation, as well as the close correspondence between evoked and spontaneous activities.