Thanks for taking a look. A couple of reactions below.
Darren Cook wrote:
Is anyone altering the playout policy during the course of a game? Is
seems like a huge potential win if done right (of course, I don't claim
to know how to do it right). If so, can you give any details?
Searching the archives the closest I could find was "Valkyria does not":
http://computer-go.org/pipermail/computer-go/2007-April/009714.html
Ah, I just managed to find this:
http://computer-go.org/pipermail/computer-go/2007-July/010432.html
It seems like you would have a problem when you need one strategy in one
area of the board and another strategy in another area of the board
(which would probably happen a lot, or maybe even all the time,
depending on how you define strategy).
http://computer-go.org/pipermail/computer-go/2007-July/010446.html
I'm assuming the mathematical methods in CrazyStone (and others) would
converge faster than GA methods. And fast convergence would be critical
for in-game adaptation of weights. To those of you who have
CrazyStone-like engines, how long does the offline learning take?
Darren
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
