>I agree with much of what you say (to the degree that anyone needs to "agree" >with questions).
Good of you not respond as Kosh might have: "Yes" (warble sound effects;-) > The discussions on this list dealing with "ownership maps", RAVE and AMAF have > to do with using additional information from the playouts. I've found the thread for "ownership maps": http://computer-go.org/pipermail/computer-go/2007-February/008897.html and I guess the explanations and origins of RAVE/AMAF are somewhere in the >100 references each brings up in the archives? Perhaps the archives should be published in printed form, to make "catching up" easier (I wonder whether David Brin was a member of this list when he wrote his Uplift series)?-) > Playouts can't be "unbiased." Picking a move with uniform probability is a > bias too, and not a good one. I was thinking of "unbiased" in the sense of not excluding valid moves (no blind spots), but I guess after one accounts for trying to search an effectivly infinite space with finite resources, the difference between "not there" and "not likely" isn't all that big. And since one is sure to miss some important moves, one has to try an tune the heuristics so that the available resources are spent in the most likely areas of the search space. It all comes down to experiments, probably, but the experimentation tends to be limited to self-play, play against a couple of freely available opponents, and low-frequency tournaments. Not exactly ideal for exposing unfortunate biases in experimental heuristics, no matter how much math one throws at it. I assume everyone is aware of how computer go is self-similar, the research/development following the same patterns as the go engines? Where there used to be deep study followed by few deliberate moves or engine releases, there is now a need for frequent tournaments as playouts for testing many possible playout heuristics:-) > Computer go papers here: http://www.citeulike.org/group/5884/library Thanks! Together with the list archives, I guess I'll not run out of reading material for a while. I've already found some papers relevant to my questions, but having the right search terms and reference lists is just as important. What I haven't yet seen is a FAQ for this list (search terms, topics, terminology, links, and the like). Is there one? Thanks for your comments, keywords, and library url, Claus _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
