2007/12/4, Álvaro Begué <[EMAIL PROTECTED]>: > > > > > Also there is not much published information evaluation functions in > > Go. Obviously a go programming is a business and giving out such > > information does not make sense. Best publicly available thingy is > > GnuGo and it does not even have one. > This was true of chess too, and there is more money to be made in computer > chess, so my guess is that people are more secretive there. > But there was much more open research and with published results. And ther are extremely good open source chess programs - you right there were none in early eighties, but still lots off useful information was available.
> > So selective search is part of eval. > What? That doesn't follow at all. I'm not even sure what it means. It means that "expert" part of selective search takes out lots of situation that evaluation function would analyse wrongly. kind of knowledge that chess program has in its eval go program typically has in its move generator. > Poor performance of current programs doesn't mean anything. People in the > 80s could have made similar arguments saying that alpha-beta searchers with > a simple evaluation function that considers material and positional values But they were good already then! I played computer chess in eighties. Yeah they were not masters yet but still could beat the average Joe pretty easily. Road ahead was pretty clear. And MC programs do scale really badly, better than expert programs though. You add more simulations and they get only a little bit better - on full sized board that is. Making computer 100 times faster would probably not make them measurably stronger i openeing and middle-game right now. So I think this MC thing is a passing phase that will leave go-programming with some new knowledge but will not be the solution. Better evaluation functions are needed together with global search. Then once computers get faster more candidates can be included into search. > only cannot possibly be the right approach, as evidenced by their funny > moves and their poor results against humans. It turns out that an evolution > of that approach on much faster hardware plays better than humans. > > Álvaro. > > > > > _______________________________________________ > computer-go mailing list > computer-go@computer-go.org > http://www.computer-go.org/mailman/listinfo/computer-go/ > -- Petri Pitkänen e-mail: [EMAIL PROTECTED] _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
