BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:KL control theory and decision making under uncertainty - Bert Kap
 pen ( Radboud University Nijmegen\, The Netherlands)
DTSTART:20091110T110000Z
DTEND:20091110T120000Z
UID:TALK21197@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:KL control theory consists of a class of control problems for 
 which\nthe control computation can be solved as a graphical model inferenc
 e\nproblem. In this talk\, we show how to apply this theory in the context
  of\na delayed choice task and for collaborating agents.  We first introdu
 ce\nthe KL control framework. Then we show that in a delayed reward task\n
 when the future is uncertain it is optimal to delay the timing of your\nde
 cision. We show preliminary results on human subjects that confirm\nthis p
 rediction. Subsequently\, we discuss two player games\, such as the\nstag-
 hunt game\, where collaboration can improve or worsten as a result\nof rec
 ursive reasoning about the opponents actions.  The Nash equilibria\nappear
  as local minima of the optimal cost to go\, but may disappear when\nmonet
 ary gain decreases. This behaviour is in agreement with\nexperimental find
 ings in humans.\n
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR