BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Craik Club
SUMMARY:Statistically optimal inference and learning: from
  behavior to neural representations - Professor Jó
 zsef Fiser\, Brandeis University
DTSTART;TZID=Europe/London:20090317T130000
DTEND;TZID=Europe/London:20090317T140000
UID:TALK17217AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/17217
DESCRIPTION:In recent years\, a growing number of human psycho
 physical studies and physiological experiments wit
 h behaving animals supported the notion that the b
 rain encodes both the value and the uncertainty of
  the stimulus during perception.  As a result\, co
 rtical sensory processing has been increasingly vi
 ewed in the framework of probabilistic inference. 
  However\, there have been much fewer attempts to 
 extend this probabilistic notion to learning even 
 though perception and learning are clearly two fun
 damental aspects of cortical processing.  In addit
 ion\, present theoretical probabilistic models of 
 cortical inference cannot handle easily important 
 aspects of cortical activity such as trial-to-tria
 l variability\, and structured spontaneous activit
 y.  I will present a framework and some results th
 at address these issues and potentially link learn
 ing behavior to probabilistic neural activity in t
 he cortex.  First\, I provide evidence that human 
 learning of chunks\, new visual feature combinatio
 ns is close to statistically optimal.  Next\, I pr
 esent a generative framework of cortical processin
 g and learning based on the assumption that neural
  activity represents samples of the posterior prob
 ability distribution defined by the uncertainty of
  the stimulus.  This framework can explain trial-t
 o-trial variability\, gives a functional role to s
 pontaneous activity in the cortex\, and provides s
 ome clearly testable physiological predictions.  I
 n the final part of my talk\, I will demonstrate h
 ow we confirmed one of these predictions by showin
 g that with accumulating visual experience\, the d
 istribution of spontaneous activity in the cortex 
 approximates the distribution of evoked activity a
 veraged over natural stimuli.
LOCATION:Kenneth Craik Room\, Craik-Marshall Building\, Dow
 ning Site
CONTACT:John Mollon
END:VEVENT
END:VCALENDAR