[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gnugo-devel] linear_eye_space() considered harmful
From: |
Gunnar Farneback |
Subject: |
Re: [gnugo-devel] linear_eye_space() considered harmful |
Date: |
Tue, 09 Oct 2001 17:23:26 +0200 |
User-agent: |
EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.2 (Yagi-Nishiguchi) APEL/10.3 Emacs/20.7 (sparc-sun-solaris2.7) (with unibyte mode) |
Inge wrote:
> Is the function linear_eye_space still needed? It is a very complex
> function and the same functionality could easily be created with a few
> methods in eyes.db. Or perhaps, the necessary patterns are already
> there?
Not easily. The special case of topological false eyes coinciding with
opponent stones inside the eyespace doesn't have support in eyes.db
yet.
Dan wrote:
> It is still used. I am very skeptical that we have all the
> patterns implied by this function. But one way to try is
> to take it out and run the regressions. I expect some
> errors will pop up.
The test cases optics:312,325 in particular, which have the feature I
described above. It's possible that some other patterns handled by
linear_eye_space() are also missing, but there's a reasonable chance
that the heuristics for guessing nonmatched eyespace values will get
those right.
Btw, those heuristics are complex enough that they should be broken
out of compute_eyes_pessimistic() and be put in a function of their
own.
/Gunnar