Aja I don't know if you have looked at Hendrik's masters thesis (as opposed to the shorter paper with Peter Drake) but it describes a lot of variations on LGR which look sensible and should be better but are in fact much worse. This suggests that LGR1 and LGRF2 are quite hard to improve on.
Hendrik - did you look at any metrics on the variations to see if you could establish why most of them were not successful? I was wondering if looking at the percentage of suggestions made by the policy or the refresh rate would suggest what the problem is with some of the others. For example, a policy which is providing a "good reply" nearly 100% of the time with a low refresh rate is probably too narrow for good exploration. Oliver On Sun, Jan 30, 2011 at 4:47 PM, Aja <[email protected]> wrote: > > Thanks for your hint and encouragement. I will try. > > Aja > > > You will most likely get very few matches of all three patterns, the >> probability is just too low. Try matching only the surroundings of the last >> move, for example. Try to go from general to more specific conditions, as >> there are many ways to formulate the specific ones. Good luck! >> > > _______________________________________________ > Computer-go mailing list > [email protected] > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
