[Top][All Lists]

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

Re: [Bug-gnubg] An evalutaion of the pruning nets

From: Joseph Heled
Subject: Re: [Bug-gnubg] An evalutaion of the pruning nets
Date: Tue, 26 Oct 2004 12:54:07 +1300
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7) Gecko/20040616

As the person who "independently discovered" the reduced evaluation (I am sure others have thought about it before), I will be happy to see them go. Everything you can do with the reduced evaluation, you can do faster and with better accuracy with the prune nets. The whole reason I developed the prune nets was my dissatisfaction with reduced evaluations. I am not sure I have the time to prove to you they are bad, but I am very confident that is the case.

As to your idea of half ply, it is a good one. It certainly improves on 1ply and I certainly would love to see it in GNUbg. The way things work, this will probably not happen before I implement it myself ...


Nis Jorgensen wrote:
Øystein Johansen wrote:

Jim Segrave wrote:
| Given the enormous speed increase from the pruning nets, I am inclined
| to agree with Joseph Heled that we should simply remove reduced
| analysis.

Me too! I also agree that the reduced evaluations can be removed. I
believe that the "old" reduced evaluation gives more deviations from the
"ordinary" nets than the pruning net reduction.

I still have my suggested "fractional ply" code lying around, which uses
the "reduced" settings. Not sure whether they can be merged with the
current code though.

I never got around to actually testing whether it was improving on the
playing strength. I have a hope that it can work in combination with the
pruned nets - together allowing us to achieve better playing strength in
the same timespan.

My compilation environment is broken, so it might take some time before
I can get this working again. And I still don't have a framework for
testing playing strength. But I would definitely be sad to see reduced
evaluation being removed from the code.

Bug-gnubg mailing list

Bug-gnubg mailing list

reply via email to

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