Yes, you are right. Now I understand how those Monte Carlo programs works. I modified my question (and email subject) as:
On average how many board updates/sec can top conventional Go programs do these days? On Jan 14, 2008 8:48 PM, David Doshay <[EMAIL PROTECTED]> wrote: > The problem here is that you asked mutually contradictory things. You > defined what you meant by a board update, in which you specified a > list of things, and you also asked about "top programs." The top > programs do not do the kinds of evaluations you specify, although > older conventional programs do. The newer programs that are now the > strongest are variations of the Monte Carlo method, which does > statistical sampling, not the kinds of evaluation you specify. > > Cheers, > David > > > > On 14, Jan 2008, at 7:41 PM, mingwu wrote: > > > On Jan 14, 2008 6:15 PM, Jason House <[EMAIL PROTECTED]> > > wrote: > > > > slow. UCT (or generically Monte Carlo) can "evaluate" a position > > fairly > > quickly (maybe 1k-100k per second depending on how heavy the playout > > is), they don't give a reliable estimate. To improve this, they > > end up > > > > 1K ~ 100 K / sec is much faster than "a dozen" / sec of a > > conventional program. > > > > Do they calculate dragon safety (eyes, connections, patterns ...)? > > if not, the estimate will be VERY unreliable. > > But if they do, how can they be this fast compared to the more > > conventional programs? > > > > reevaluating positions more than once (maybe 100 times?) to get a more > > reliable estimate. > > > > why "reevaluating" the same position? > > > > Sorry, I didn't go into their papers, can people who knows UCT, or > > actually working on UCT programs explain in a way that a layman can > > understand. Thanks. > > > > _______________________________________________ > > computer-go mailing list > > [email protected] > > http://www.computer-go.org/mailman/listinfo/computer-go/ > > _______________________________________________ > computer-go mailing list > [email protected] > http://www.computer-go.org/mailman/listinfo/computer-go/ >
_______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
