Computing That Serves

Skill and Billiards: Game Theory in Complex Domains


Thursday, January 31, 2013 - 11:00am


Chris Archibald

Post-doctoral Fellow, University of Alberta


Mike Goodrich

Colloquium is in 1170 TMCB.


Game-theoretic principles have lately been increasingly useful
in designing strategies for agents in interesting artificial
intelligence domains, beginning with games.  This talk will focus on
developments in computational pool, a relatively recent entrant to the
group of games played by computer agents.  This game features a unique
combination of properties that distinguish it from previous games,
including continuous action and state spaces, uncertainty in execution, a
unique turn-taking structure, and of course an adversarial nature. I
will discuss initial game-theoretic work in this domain, as well as the
practical work done to create CueCard, the most recent world champion
computational billiards program. I will also discuss insights into
execution skill that were gleaned from the computational pool domain,
and close with the challenges that lie ahead in this research area.



Christopher Archibald is a Postdoctoral Fellow at the
University of Alberta, where he works in the Alberta Innovates Centre
for Machine Learning.  He received his PhD from Stanford University in
2011, where he was advised by Professor Yoav Shoham.  His research
focuses on developing theoretically grounded methods for designing
successful strategies in complex adversarial environments.  He was the
lead AI programmer for CueCard, the agent that won the 2008 Computer
Olympiad Computational Pool tournament in Beijing, China. In the past he
was involved with designing a vision-guided robot at BYU, where he
received a B.S. in Computer Engineering.