On 10/15/07, Erik S. Steinmetz <[EMAIL PROTECTED]> wrote: > Greetings all, > > I have been looking through the literature (many thanks to Markus's > wonderful online bibliography) on existing strategies in the opening > game, and have not found too many articles on the specifics outside > of a few papers on neural net learning applied to the opening. There > are some vague references to 'pattern matching' to generate moves, > but no information about how those patterns and moves were created. > > I am wondering if anyone knows of any attempts made to run pattern > recognition (for example, clustering) algorithms over a library of > games in order to learn reasonable opening moves. If so, and there > are any papers about the success (or failures) of such an effort, I > would really appreciate a pointer!
Large-scale pattern harvesting from real games between strong and pro players is the raison d'etre of Frank de Groot's commercial go analysis tool Moyo Go Studio (www.moyogo.com). Moyo Go Studio's pattern database includes millions of patterns of a wide variety of sizes, definitely includes opening patterns. I don't know if de Groot has contributed to the literature beyond making a number of posts to this mailing list and commercial announcements on rec.games.go. Remi wrote that for CrazyStone, he culled some larger patterns that did not appear to add strength (in particular opening patterns, though maybe he put them back in for his standalone pattern recognizer). It is just a different emphasis -- probably less complete but more useful for making a strong go-playing program. I can't say anything about any of the others who have tried similar things. _______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
