BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Value Propagation: A Graphical Model for Bayesian Reinforcement Le
 arning - Philipp Hennig (University of Cambridge)
DTSTART:20081117T110000Z
DTEND:20081117T120000Z
UID:TALK14954@talks.cam.ac.uk
CONTACT:Carl Scheffler
DESCRIPTION:I will present Bayesian Reinforcement Learning\nmethods based 
 on the model-free approach\nused in the Temporal Difference family of\nalg
 orithms. Our implementations allow for\nthe incorporation of prior knowled
 ge in a\nprincipled way and automatically adapt their\nlearning rate and b
 ackup depth. In policy iteration\nsettings\, they can guide exploration\na
 nd converge on the optimal policy in an\nautomated fashion\, because they 
 track the\nuncertainty of evaluations.
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Department of Physics
END:VEVENT
END:VCALENDAR