On Fri, Mar 28, 2008 at 3:10 PM, Jaonary Rabarisoa <[EMAIL PROTECTED]> wrote: > So to sum up we have the following pseudo code : > > at a given node : > - find the child (among the visited child only) that maximizes de UCT-RAVE > value > - if this maximum UCT-RAVE value is less than FPU value and if there still > exisits unvisited nodes : > > choose one unvisited node- continue > > Is this correct ?
Maybe, but it depends on how exactly you compute your Upper Confidence Bounds. If you don't add UCB's to the rave values you may either have to compare based on the UCT part alone, or do as GCP suggests (which sort of turns FPU in an UCT prior). However, I would suggest that you first start out with the standard UCT (always explore unvisited nodes) and see how you can improve from that as a baseline. If your branching factor becomes too big maybe try some kind of metabandit. Erik _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
