BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Balanced spiking networks can implement dynamical systems with pre
 dictive coding
DTSTART:20121206T133000
DTSTAMP:20260407T212736Z
UID:81d0642872025089a50da8f5945a5e89a7f7e53b81fdf22952872c74
CATEGORIES:Conferences - Seminars
DESCRIPTION:Sophie DENEVE\; Group for Neural Theory\, LNC\, DEC\, ENS\nNeu
 ral networks can integrate sensory information and generate continuously v
 arying outputs\, even though individual neurons communicate only with spik
 es---all-or-none events. Here we show how this can be done efficiently if 
 spikes communicate "prediction errors" between neurons. We focus on the im
 plementation of linear dynamical systems and derive a spiking network mode
 l from a single optimization principle. Our model naturally accounts for t
 wo puzzling aspects of cortex. First\, it provides a rationale for the tig
 ht balance and correlations between excitation and inhibition. Second\, it
  predicts asynchronous and irregular firing as a consequence of predictive
  population coding\, even in the limit of vanishing noise. We show that ou
 r spiking networks have error-correcting properties that make them far mor
 e accurate and robust than comparable rate models. Our approach suggests s
 pike times do matter when considering how the brain computes\, and that th
 e reliability of cortical representations could have been strongly under-e
 stimated.
LOCATION:SG 0213 http://plan.epfl.ch/?room=sg0213
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR