> One might consider heuristics like AMAF, pattern knowledge, etc. to be > simply a more effective way to guide exploration. The UCB term has no > domain-specific knowledge. It works surprisingly well but it should be > no surprise that one can do better with domain-specific knowledge.
The problem of knowledge like AMAF or patterns is that it is a static heuristic. They are fine when you have no more information (from the search). I find that it works best to quickly scale them out as soon as search (or simulation) information is available. So, I think I disagree here. UCB has no domain-specific knowledge, but it is able to use search results. On the other hand, Sylvain Gelly posted Mogo's current system here, and it seems they always use AMAF information to control exploration. I couldn't get this to work. Maybe that's my problem :-) -- GCP _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
