BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Optimal control and optimal sampling: A statistical physics perspe
 ctive. - Bert Kappen\, Radboud University Nijmegen\, and UCL London
DTSTART:20141021T133000Z
DTEND:20141021T143000Z
UID:TALK55411@talks.cam.ac.uk
CONTACT:Tim Hughes
DESCRIPTION:Intelligent systems\, whether natural or artificial\, must act
  in a world that is highly unpredictable. To plan actions with uncertainty
  is a stochastic optimal control problem. However\, there are two fundamen
 tal problems: the optimal control solution is intractable to compute and i
 ntractable to represent due the non-trivial state dependence of the optima
 l control. This has prevented large scale application of stochastic optima
 l control theory sofar. The path integral control theory describes a class
  of control problems whose solution can be computed as an inference comput
 ation. In this presentation we formalize the intuitive notion that the eff
 iciency of the inference computation\nis related to the proximity of the s
 ampling control to the optimal control. Secondly\, we show new results tha
 t allow approximate computation\nof state dependent optimal controls in te
 rms of basis functions. These two ingredients together suggest a novel ada
 ptive sampling procedure. We illustrate the results on a few examples.
LOCATION: Cambridge University Engineering Department\, LR12
END:VEVENT
END:VCALENDAR