On Wed, Dec 3, 2008 at 05:17, Mark Boon <[EMAIL PROTECTED]> wrote: > I had set myself as an arbitrary goal that it should do at least 20K > playouts. But with real liberties, AMAF and a RAVE formula I got stuck in > the 16K-17K range. According to my profiler that is mostly due to the > expensive formula used to compare nodes, where it says it spends 25% of > total time. The formula I'm using is: > > beta * (virtual-win-ratio + RAVE) + (1-beta) * (win-ratio + UCT) > > beta = 1 - log(parent-visits) / 20 > UCT = exploration-factor *sqrt( log(parent-visits) / (nr-visits+1) ) > RAVE = exploration-factor *sqrt( log(parent-visits) / > (nr-virtual-visits+1) )
Hi, This is probably something you already do, but just in case you don't, here comes a simple optimization trick: since parents-visits is an integer value and in a somewhat bounded range (0..max_playouts), the logarithm can be precomputed in a table. Then even sqrt(log()) can be precomputed also. Similarly, a table with sqrt(integer()) can be used for the other values. best regards, Vlad _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
