[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Forwarded suggestion from the website

From: Joseph Heled
Subject: Re: [Bug-gnubg] Forwarded suggestion from the website
Date: Wed, 27 Oct 2004 11:30:23 +1300
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7) Gecko/20040616

Jim Segrave wrote:
On Tue 26 Oct 2004 (23:58 +0200), ??ystein Johansen wrote:

Hash: SHA1

- ----8<----
Maik Stiebler wrote:
Congratulations to this new improvement!
Have you considered to use the pruning net as evaluation engine for
variance reduction on 0-ply rollouts? Right now I don't use VR on those
because it's so expensive relative to the rollout itself. If the pruning
net is good enough, that might be much more effective.
- ----8<----

Heeeeyyyy! What a great idea!

I know for sure I won't have the capacity to take this task right now.
(Even though I would love to). Anyone?

It sure sounds like a great idea to me!

I thought Joseph said they only come into play on .gt. 0 ply rollouts
(they speed up the selection of candidates for the next ply), so
they'd have no effect on zero ply rollouts.

Yes. but I think the suggestion is to use them for VR. fortunately, they are not good enough for that.

BTW, one day I may understand VR and find a method that works. While I have seen many people discuss it and lots of hand waving, I never saw it work - i.e. where for a fixed time T it produces a better result than simply doing more trials.

Please, someone prove me wrong ...


reply via email to

[Prev in Thread] Current Thread [Next in Thread]