>> more nodes. It will be the same as if I did 100,000 nodes instead >> of 50,000 nodes. Or put another way, it will be the same as if >> I spent 20 seconds on this move instead of 10 seconds. > ... > Consider move 20 (for example). If you saved every "move 20" node > you ever encountered, how often do you think you'd encounter a...
Hi Dave, You are right that by move 20 you are unlikely to get a repeated game. But by having that information in the tree, the decision made at move 5 is more accurate. Basically what Don's suggestion means is for the first few moves of the game (while you are playing a game that has been seen before) it is as if your program is playing on hardware that is N times quicker (where N is the number of times this position has been seen before). I think a UCT tree seeded with a large pro/strong-player game library would be useful, and is basically the same idea. It would encourage the computer to play well-known fuseki/joseki (without having to work them out from first principles) while still having the UCT tree behind it. Meaning that the level of play would not suddenly drop when the opponent leaves your opening library, and meaning it would not suddenly find itself in a situation that it does not understand. Darren -- Darren Cook, Software Researcher/Developer http://dcook.org/mlsn/ (English-Japanese-German-Chinese-Arabic open source dictionary/semantic network) http://dcook.org/work/ (About me and my work) http://dcook.org/blogs.html (My blogs and articles) _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
