BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Deducing Principles in Natural Perception - Mike Lewicki - CMU\, B
 erlin\, and Case Western
DTSTART:20090415T130000Z
DTEND:20090415T140000Z
UID:TALK17519@talks.cam.ac.uk
CONTACT:David MacKay
DESCRIPTION:What are the underlying computational principles that biology 
 uses to transform the raw sensory signal into a hierarchy of representatio
 ns that subserve higher-level perceptual tasks?  One hypothesis is that bi
 ological representations are optimal from the viewpoint of statistical inf
 ormation processing\, and are adapted to the statistics of the natural sen
 sory environment.  In the initial stages of sensory coding\, information h
 as to be coded in a way that makes best use of the available resources.  I
  will show how the optimal solution to coding natural images with a popula
 tion of noisy neurons predicts many properties of retinal coding.  I will 
 also show that the same approach can be applied to natural sounds to expla
 in many aspects of the auditory code at the cochlear nerve.  Finally\, I w
 ill present work that extends this general principle to higher levels of v
 isual processing.  I will show that efficient representations of the stati
 stics of local image regions can form stable\, invariant representations o
 f edges\, contours\, and textures.  These results also provide a novel fun
 ctional explanation for non-linear effects of complex cells in the primary
  visual cortex.\n
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Department of Physics
END:VEVENT
END:VCALENDAR