[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gnugo-devel] arend_1_25.3: less (??) experimental influence
From: |
Arend Bayer |
Subject: |
Re: [gnugo-devel] arend_1_25.3: less (??) experimental influence |
Date: |
Thu, 14 Feb 2002 01:01:12 +0100 (CET) |
>
> It appears that the experimental influence option is definitely stronger
> but now the question is why it is so slow, and whether something
> can be done about this.
>
16% time cost is definitely too much.
Of course the additional followup influence computation costs some time,
mainly for pattern matching. This can be reduced by using Trevor's
matchpat_goal_anchor.
I had assumed this is the only part costing extra time, but that's not
relevant for semeai.tst. Instead I realize that the reduced default
attenuation might cost extra time, as more fields are reached until the
influence cut off has some effect. This could be changed by revising
the cutoff policy.
However, at the moment I find it hard to imagine that this costs 16% time
for semeai.tst. I will have a look at that tomorrow.
Arend