BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:LCN Seminar: Causes and effects of noise correlations in neural po
 pulations
DTSTART:20131010T133000
DTSTAMP:20260427T195822Z
UID:c0494a3fb758d52e7f8dc863de84995d596370b2709beca66667ea88
CATEGORIES:Conferences - Seminars
DESCRIPTION:Joel ZYLBERBERG\, Department of Applied Mathematics\, Universi
 ty of Washington\nSensory neurons are "noisy": if one repeats the same sti
 mulus\, the neural responses differ from trial-to-trial. This noise is oft
 en correlated between groups of neurons\, and much research has investigat
 ed how the noise correlations affect the ability of neural populations to 
 encode information about the stimulus. Typical work in this area ignores t
 he mechanistic origins of the noise correlations -- which include common i
 nput to multiple cells\, and recurrent coupling between cells -- and inste
 ad focuses on how the response statistics themselves (variances\, correlat
 ions\, etc.) relate to coding performance. Arguably the more important (ye
 t clearly related!) question is how the biophysical structure of the syste
 m affects its function. Naively\, one might expect that common input\, or 
 recurrent coupling\, help (hinder) coding performance\, if the noise corre
 lations that they generate help (hinder) coding performance. In my talk\, 
 I'll explain why this logic is flawed: recurrent coupling can improve codi
 ng performance while also creating deleterious noise correlations.\nThis w
 ork has implications for connecting structure to function with regards to 
 noise correlations in neuronal circuits\, and allows us to make testable p
 redictions for comparative physiology experiments.\nFinally\, I will discu
 ss our ongoing efforts to understand quantitatively the noise correlations
  in the mouse direction selective ganglion cell system.
LOCATION:SG0213 http://plan.epfl.ch/?room=sg0213
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR