EECS Joint Colloquium Distinguished Lecture Series

Wednesday, March 30, 2005
Hewlett Packard Auditorium, 306 Soda Hall
4:00-5:00 p.m.

Bruno Olshausen

University of California at Davis Department


Principles of Image Representation in the Visual Cortex



Nervous systems have evolved impressive abilities to extract useful information about the environment from images. Here, I shall present the results of recent efforts aimed at understanding how images are represented in the mammalian visual cortex. The driving hypothesis behind this work is that the cortex contains powerful inferential machinery - i.e., probabilistic models for combining incoming sensory information together with prior knowledge in order to infer what's "out there" in the environment. I will show that a simple version of a probabilistic model for natural images based on sparse coding can account for several of the response properties of neurons in the primary visual cortex (V1), such as the spatio-temporal structure of receptive fields, and the sparse activity of neurons in response to natural images. I will also discuss current efforts aimed at extending this model in a hierarchical fashion to capture more complex aspects of image structure, as well as its application to problems in image analysis such as denoising


Bruno Olshausen is an Associate Professor of Department of Neurobiology, Physiology and Behavior and Center for Neuroscience, UC Davis, and is the Principal Investigator for the Redwood Neuroscience Institute.