Brown University Department of Neuroscience, Division of Biology and Medicine Seminar
Dr. Redish is a candidate for a faculty position in the Department of Neuroscience. If you wish to meet with Dr. Redish contact Susan Troy at X3-9524. Please send written comments to: J. Donoghue, Theoretical Search Committee, Box 1953, John_Donoghue@brown.edu |
Brown University Joint Seminar, Department of Neuroscience, Division of
Biology and Medicine and
Division of Applied Mathematics
Abstract: The probabilistic notion of departure from conditional independence may be useful when studying the brain on each of the following three levels: cognition; physiology; anatomy. The first part of the talk will outline an approach to high-level vision in which composition expresses a departure from conditional independence. Thus, given that two collections of pixels, L1 and L2, have been interpreted separately as instances of the symbol LINE, specific relationships between these lines may signal that L1 and L2 did not occur independently of each other. We then compose L1 and L2 into a higher-level object, an instance of the symbol RIGHT ANGLE, itself available for further composition. The second part of the talk will describe a two-tiered probabilistic model for the spiking activity of a collection of neurons. The model allows us to formulate a null hypothesis H(r), which roughly says that (a) the temporal variation of firing rates is bounded by r, and (b) neurons fire independently of each other given their rates. H(r) can be tested using a tailor-fit "jitter" method, and this allows us to get a lower bound on the "time resolution" of neural activity. The third part of the talk will discuss data from the literature indicating a departure from randomness, i.e., conditional independence, in the graph of synaptic connections between individual cortical pyramidal cells. I will conclude with (a) the proposal to view these three occurrences of departure from conditional independence as three facets of one phenomenon, and (b) a proposed experiment for a partial test of this conjecture.
***Dr. Bienenstock is a candidate for a faculty position in the Department of Neuroscience and the Division of Applied Mathematics. If you wish to meet with Dr. Bienenstock contact Susan Troy at X3-9524. Please send written comments to: J. Donoghue, Theoretical Search Committee, Box 1953, John_Donoghue@brown.edu
<--- 2000 Index