2010-05-24

POMDP

Unpated (July 24, 2010)
Now, their paper is uploaded:
KANDORI, M. and Obara, I. (2010) "Towards a Belief-Based Theory of Repeated Games with Private Monitoring: An Application of POMDP"
(paper can be downloaded from the author's website)

When I had a chat with Prof. Kandori on Friday, he talked about his recent joint-work with Ichiro Obara (UCLA) on repeated games with private monitoring. The key idea is to use a well-known method in dynamic programming called POMDP (partially observable Markov decision process).
Then, they derived the (right notion of) one-shot deviation principle for private monitoring, and constructed a method to check whether an arbitrary strategy, expressed by finite automaton, constitutes an equilibrium or not. I couldn't follow the arguments in detail, but what they have found sounds really path-breaking.
Look forward to see their paper soon!

Information about POMDP websites
Useful site includes tutorials and codes. link
Lecture movie at Stanford (Machine learning: Lecture 20) link

No comments: