On Tue, Jul 6, 2010 at 10:02 AM, Petr Baudis <[email protected]> wrote:
>  I can do this already, and it's rather easy to do it incrementally;
> moreover, usually you are interested only in liberties of low-liberty
> groups, so I use the GNUGo method and keep explicit list only of up to
> five liberties of a group. I think I have found that it speeds
> 1-lib/2-lib tactics a lot.

You mean, it speeds 'your' 1-lib/2-lib tactics a lot. I found the
opposite to be the case with mine.

>  But please note that what I'm after is something completely different.
> I'm not after liberties, but after the opposite - the neighboring groups,
> and I need a complete list to find all atari'd neighbors. Also, I do need
> to do this incrementally since I need to have this information about all
> atari'd groups at each move (for the probability distribution).

That was included  in the same discussion thread, although I can't
claim the ultimate answer was found. Obviously, all known methods come
with a certain expense. You'll have to come up with something
completely new if you're looking for dramatic speed-up. So far every
attempt to make a general solution for liberty administration to be
faster than a custom one for tactics comes up short by a good factor
of two or three AFAIK. Also, the solution may be radically different
depending on whether you need to be able to go back or not.

Mark
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go

Reply via email to