Friday, April 11, 2014
217 Perkins Library - 11:00 a.m.
Marc Sommer, PhD
Associate Professor, Department of Biomedical Engineering, Duke University
Our visual sensors are mobile. We move our eyes about twice per second throughout our waking lives. Each quick eye movement, or saccade, causes the image to leap across the retinas. How does our brain disregard the jumpy visual inputs to create the stable, continuous visual percept that we enjoy? My laboratory models the neuronal circuits that mediate stable visual perception across saccades, based on empirical data from behaving monkeys. Two approaches are used. In the first, sheets of simulated neurons attached in recurrent layers receive visual images from a camera as well as motor signals from the camera controller. The networks are trained to maintain a constant representation of an external visual object regardless of camera movements. The second approach is a Bayesian model with the striking emergent property of recreating saccade-related visual illusions that people experience. The technical goal of our work is to build a biologically plausible model for use in robots to provide them with eye movements just like ours (same frequency, size, and speed) without sacrificing the speed and stability of visually-guided reaches. Our scientific goal is to generate predictions about sensorimotor processing in the brain that can be tested in new physiological experiments.
Marc Sommer studies neuronal circuits in the primate brain from an engineering perspective. He studied Electrical Engineering and Biological Sciences at Stanford before earning a PhD in Systems Neuroscience at MIT. After postdoctoral training at the National Eye Institute of the NIH, he started his first lab at the University of Pittsburgh in 2004. He moved to Duke in 2010 and runs two closely linked labs, one devoted to computational modeling and robotics and the other to neurophysiology of the primate brain.