That's exactly the issue. You don't know if it's an underestimate or
overestimate, but you can be sure that the RAVE and UCT values will
not match... Even if you run millions of simulations (without
expanding the tree), the values still will not match. I expect the
RAVE bias is the expected magnitude of that difference.
Sent from my iPhone
On Jul 28, 2009, at 7:28 PM, Yung-Pin Chen <yc...@lclark.edu> wrote:
Jason,
Thanks for your explanation. If RAVE "combines results from many
variations," how can we justify it is an overestimate or underestimate
of the true value of a move? Is it reasonable to assume that both
UCT and RAVE are equally-biased?
Yung-Pin
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/