[Top][All Lists]

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

Re: [gnugo-devel] GNU Go Development Idea

From: David G Doshay
Subject: Re: [gnugo-devel] GNU Go Development Idea
Date: Tue, 1 Mar 2005 12:02:49 -0800

On 1, Mar 2005, at 9:40 AM, Bowdridge, Steven James wrote:

It basically ponders the top 10 expected
responses, and checks back for user input between pondering each move. If the opponent has made one of the moves that has been pondered, GNU Go doesn't need to consider a
response, and instantly responds.  The idea is to save some time,

I have not checked the top 10, as you propose, but I have checked the single top expected response when our SlugGo (a wrapper over GNU Go) plays Many Faces of Go. In the admittedly very small sample of one game, GNU Go's top response to SlugGo's chosen move was first chosen by Many Faces on move 50, i.e. 49 misses before a hit. It seemed so seldom that I did not keep checking after that. But I do expect more hits out of the top 10.

In the case of Many Faces, I doubt that GNU Go could ponder the replies to the 10 possible responses in less time than Many Faces takes to play. But if you have 10 cpus at hand to do that pondering in parallel (one time step rather than 10) you might save some time.


reply via email to

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