BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:From tuning curves to behaviour - Mandana Ahmadi (Gatsby Computati
 onal Neuroscience Unit\, UCL)
DTSTART:20100524T093000Z
DTEND:20100524T103000Z
UID:TALK25029@talks.cam.ac.uk
CONTACT:Prof Máté Lengyel
DESCRIPTION:Trading off speed and accuracy in decision-making is a hard co
 mputational problem. This is partly because whenever we make a decision we
  are forgoing future rewards\, whose value is hard to estimate\, and partl
 y because feedback comes only when we make a decision\, which means we oft
 en have to learn from relatively sparse data.\n\nWe use reinforcement lear
 ning to find the optimal tradeoff between speed and accuracy in the random
  dot kinematogram task (Newsome et al\,1989)\, a task in which subjects es
 timate the direction of moving dots against a noisy background. We employ 
 a biologically plausible learning algorithm\, Temporal Difference learning
 \, to learn the optimal behavior through trial and error\, as animals do. 
 Based on this\, we propose a rate-based neuronal network that could\, in p
 rinciple\, explain the activity observed in area LIP in monkeys performing
  the random dot kinematogram task.
LOCATION:Cambridge University Engineering Department\, CBL Rm #438 (http:/
 /learning.eng.cam.ac.uk/Public/Directions)
END:VEVENT
END:VCALENDAR