I think I learned it from Remi's paper "Computing ELO Patterns in the Game of Go". I use it in the UCT RAVE bias, but it makes little difference to playing strength for ManyFaces.
I think many strong programs now use a topological distance measure, where all the stones in an adjacent enemy group are distance 1, and its liberties are distance 2, etc. I use this measure in the traditional MF engine, used to bias the UCT search, but I dont use it yet otherwise. David > -----Original Message----- > From: [email protected] [mailto:computer-go- > [email protected]] On Behalf Of Harald Johnsen > Sent: Sunday, June 12, 2011 9:47 AM > To: [email protected] > Subject: Re: [Computer-go] "gridcircular" - author/reference help requested > > Arthur Cater a écrit : > > I cannot remember who/what I read that mentioned a "gridcircular" > > measure of > > distance - contrasting with Euclidean and Manhattan. It is something like > > max(Hseparation,Vseparation) + ( 1/2 * min(Hseparation,Vseparation) ) > > > > If anybody can help, with claiming it, or naming its inventor, or > > especially giving > > a reference I can cite in a paper - I'd be very grateful. > > > > (Google is often my friend; but not this time.) > > > > Arthur > > > > _______________________________________________ > > Computer-go mailing list > > [email protected] > > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go > > > Petr Baudis used this word and used that kind of patterns in > experimental code of Pachi.I'm also using that now. > > http://www.mail-archive.com/[email protected]/msg13324.html > > > hj. > _______________________________________________ > Computer-go mailing list > [email protected] > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
