Bruno Olshausen (University of California, Davis):
Sparse coding as a principle of image representation in visual cortex.
I shall discuss the principle of sparse coding as it has been applied
toward modeling the response properties of V1 neurons. I shall also
present evidence from functional magnetic resonance imaging showing
that activity in V1 is reduced as a result of perceptual grouping.
These results together with the modeling studies suggest that cortical
neurons attempt to provide a succinct description of the structures in
natural scenes using a small number of active units.