BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Stochastic control as an inference problem - Prof. Bert Kappen (Un
 iversity of Nijmegen)
DTSTART:20090217T140000Z
DTEND:20090217T150000Z
UID:TALK15879@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:Stochastic optimal control theory deals with the problem to co
 mpute an\noptimal set of actions to attain some future goal. Examples are 
 found\nin many contexts such as motor control tasks for robotics\, plannin
 g and\nscheduling tasks or managing a financial portfolio. The computation
  of\nthe optimal control is typically very difficult due to the size of th
 e\nstate space and the stochastic nature of the problem.\n\nWe introduce a
  class of stochastic optimal control problems that can\nbe mapped onto a p
 robabilistic inference problem. This duality between\ncontrol and inferenc
 e is well-known. The novel aspect of the present\nformulation is that the 
 optimal solution is given by the minimum of a\nfree energy and the link to
  graphical model inference. We can thus apply principled approximations su
 ch as the belief propagation or the Cluster Variation method to obtain eff
 icient approximations.\nWe will illustrate the method for the task stackin
 g blocks. If time permits we will discuss distributed (agent) solutions an
 d comment on the partial observable case.
LOCATION:Cambridge University Engineering Department\, LR5
END:VEVENT
END:VCALENDAR