[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gnugo-devel] eyespace reading
From: |
Gunnar Farneback |
Subject: |
Re: [gnugo-devel] eyespace reading |
Date: |
Wed, 09 May 2007 23:12:55 +0200 |
George wrote:
> I restrict the considtion
>
> optics.c
>
> +1698 || eye[pos].neighbors > 2 || eye[pos].neighbors > 1 &&
> !is_edge_vertex(pos))
> -1698 || eye[pos].neighbors > 1)
This has quite big effects on the regressions, 28 PASS / 40 FAIL, see
http://trac.gnugo.org/gnugo/ticket/184 and the details at the bottom of
http://trac.gnugo.org/gnugo/regressiontest
> By the way, the amend above can solve position like
>
> (;AB[be][bc][cc][dc][ec][fc][hc][gb]AW[ab][bb][cb][db][eb]SZ[19]GM[1]CA[UTF-
> 8]FF[4])
>
> That is,
>
> +------------------------------
> |........
> |OOOOO.X.
> |.XXXXX.X
>
> yields 1-1-2-2 instead of 1-2-2-2 before amendation.
>
> Ya, it is tricky.
I think the only way to make significant progress with the optics code
is to introduce (shallow) reading. It's not at all obvious how to do
that efficiently, so it would be a major project.
/Gunnar
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- Re: [gnugo-devel] eyespace reading,
Gunnar Farneback <=