BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Measuring Game Temperature With UCT-Monte Carlo - Philipp Hennig (
 University of Cambridge)
DTSTART:20080512T101500Z
DTEND:20080512T111500Z
UID:TALK11844@talks.cam.ac.uk
CONTACT:Carl Scheffler
DESCRIPTION:Exhaustive search in reasonably complex trees\, like e.g. the 
 board game Go\, is extremely expensive. A particular Monte Carlo policy ca
 lled _upper confidence bound for trees_ (UCT) has emerged over the last th
 ree years as a very promising lever on such reinforcement learning problem
 s\, but has recently run into scaling problems when attempting to beat hum
 ans on large Go boards. It represents a very heuristic approach to Games.\
 n\n_Combinatorial Game theory_ on the other hand\, is a branch of pure mat
 hs providing a sturdy framework for sub-divisible full-information games. 
 It uses an abstract concept called "Temperature" to develop approximate st
 rategies that have a bounded error on the perfect line of play. Unfortunat
 ely\, it is extremely tedious to discover the temperature of games like Go
  using traditional exhaustive search.\n\nIn this talk I will present preli
 minary results on an attempt to combine the two worlds of Monte Carlo plan
 ning and Combinatorial Game Theory to produce a UCT algorithm that measure
 s Temperature and simultaneously searches for good moves on small (sub) ga
 mes. There's faint hope that this could lead to "divide-and-conquer" solut
 ions for search in general AND/OR trees with bounded rewards. \n\nThis tal
 k is about a work in progress and part of my preparations for my first yea
 r report.
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Department of Physics
END:VEVENT
END:VCALENDAR