I'm just surprised to hear that the program that introduced UCT (and got
so many others to use it), isn't using UCT any more. Combining RAVE and
UCT as described in the PDF still sounds like UCT to me, but with no
sqrt(log) term, it no longer is. I'll certainly have to think about the
trades being made and what I'd expect the outcome to be.
I agree that this is Bandit Based Monte-Carlo planning much more than
UCT. When the RAVE paper has been published, the UCT term was present,
but I think Sylvain can confirm that it has been removed a short time
later - this has been posted in the mailing list a long time ago.
(no extensive study yet of the precise
computation time required by mogo for removing all bad moves - we
are mainly analyzing _if_ some moves are still weak in spite of large
computation times - this is already quite hard, as guessing
which move(s) is (are) bad is not easy - finally, this is a 9x9 study
with humans in the loop!)
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
