[Top][All Lists]

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

Re: [Bug-gnubg] Some suggestions, what do you think?

From: Joseph Heled
Subject: Re: [Bug-gnubg] Some suggestions, what do you think?
Date: Tue, 26 Oct 2004 13:01:32 +1300
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7) Gecko/20040616

I am assuming you are talking about the defaults, since anyone can change the setting in their .gnubgautorc file. I say make the default all on.

I am running a check on the contact benchmark at 2ply. It is a slow task.
So far I have 82 of 9396 cases where the two (pruning vs non-pruning) disagreed. The error rate (per move) is 0.00725802 vs. 0.00725238

This is not conclusive, but you can see what I am talking about. A difference of 0.000006 per move. Blimey.


Øystein Johansen wrote:
Hi all!

I'm working on adding the pruning neural net option into the user
interface. (That was quite easy, actually). I have also removed the
reduction pull down menu in the evaluation setting widget.

The pruning net code has now replaced all the reduction code so you have
to choose at compile time if you want the one or the other. (The
reduction code can be compiled with #define REDUCTION_CODE 1).

So, now comes all the nice questions to all of you:

Where should the pruning be used and where should it not be used.

What about the "World Class" predefined setting? Should it be used there?

What about "Supremo" predefined setting?  Should it be used there?

And what about "Grandmaster" setting?  Should it be used there?

And what about the tool buttons in the hint window. Should that
evaluation use the pruning nets?

Any opinions?

(I have not commited the code to the CVS yet. Since I'm still missing
some small details. Hopefully these details is done tomorrow. )


Bug-gnubg mailing list

reply via email to

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