Hi!
I'm sorry to rehash such an old topic and keep bothering you with my
ineptitude, but I only recently got back to tuning the gamma-player and
I still have trouble achieving good performance...
On Tue, Dec 15, 2009 at 05:35:16PM +0100, Rémi Coulom wrote:
> Petr Baudis wrote:
> >I wonder now, do you use separate set of gammas for simulations and node
> >biasing? Since I've found that the performance seems very bad if I don't
> >include some time-expensive features, since the gammas are then very
> >off; I will probably simply generate two gamma sets, but perhaps it's
> >enough to do some trick like merging features by computing weighted
> >(geometric?) averages?
>
> I learn two sets of gammas separately for the two sets of features.
One thing that would help me tremendously for sanity-checking of my
algorithm would be listing of gamma values similar to those published in
your paper, but for the random simulation feature set. Would it be
possible to get these for some program that successfully implements the
CrazyStone algorithm?
(BTW, I have looked at libego's way of dealing with gammas yesterday
and it seems to include stone's atari status in the pattern bitmask
instead of having the capture/escape/selfatari features. I assume this
is an original improvement for libego?)
Thanks,
--
Petr "Pasky" Baudis
The true meaning of life is to plant a tree under whose shade
you will never sit.
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
