One more thing ;) Ofcourse one could also extract them by doing a fourier transformation, but I have no idea how fast this will be.
________________________________ Van: [email protected] namens [email protected] Verzonden: di 2-4-2013 8:10 Aan: [email protected] Onderwerp: Re: [Computer-go] Weight of moves I'd like to add one thing: The program can know if it's on the right track by noting that the secondary modes are removed from the histogram. It might work it's way to the solution incrementally by monitoring the histograms. I remember something about CLOP recently. I only looked at it very briefly, but my impression was that it had to do with extracting modes from a distribution. I may be wrong though, I don't follow this list actively nowadays. ________________________________ Van: [email protected] namens [email protected] Verzonden: di 2-4-2013 7:58 Aan: [email protected] Onderwerp: Re: [Computer-go] Weight of moves By the way, when I was tring to get this working I was also looking at terminal node score histograms like Ingo Althofer (http://www.althofer.de/crazy-shadows.html). It seems that a multimodal distribution hints at some larger area that is more or less independant of the rest of the board, like a semeai or a arge ko fight. In a sense every histogram will be multimodal, but in a position whithout large local fights the modes are small and close together. While it is interesting to know from the distribution that some large local fight is going on, I'm not aware of a way for the program to use this information. It's sort of a telltale sign of a horizon effect that might be solved by: 1 - finding the local fights causing secondary modes in the distribution by statistical analysis accross terminal nodes 2 - solving the local fight "separately" 3 - twining the solution (the local game tree) in the global search (game tree) in a way that looks like Conway's methods in Winning Ways. But instead of using he exact methods of Conway, it would be a MC version of it. I think 1 and 3 are the hardest parts to do, but on the other hand, as it is now, this issue is sort of a horizon effect that troubles MC players in general. I think dealing with this will be a major next step in computer go and I'm convinced it can be done, I just don't know how. Dave ________________________________ Van: [email protected] namens Gabriel .Santos Verzonden: ma 1-4-2013 22:07 Aan: [email protected] Onderwerp: Re: [Computer-go] Weight of moves Thank you Dave, I will take a look in Winning Ways. ;) On Mon, Apr 1, 2013 at 4:55 PM, <[email protected]> wrote: Gabriel, I don't think that MC players are aware of "local fights". It would be very nice if a program could divide the board in separate locations, because the combinatoric explosion would be reduced by a huge factor (4 areas with 16 empty intersections has a much smaller game tree than 1 area with 64 empty intersections). There is a method of combining the results of local endgame fights in a global result (Winning Ways by Conway, it can be viewed as a way to determine the optical merging of sepatate game trees), but in earlier stages of the game it is hard to separate out areas of the board that have low interaction. Perhaps it could be derived in an MC way (statistically) from cross correlations of board occupance at playout terminal nodes. I gave that a try a couple of years ago, but I gave up when it didn't seem to give useful results. Could be due to bugs in my code though. Dave ________________________________ Van: [email protected] namens Gabriel .Santos Verzonden: ma 1-4-2013 19:42 Aan: [email protected] Onderwerp: Re: [Computer-go] Weight of moves Álvaro, When I say "think like a human player ", I mean regarding to the strategy. For example, when there are several fights happening simultaneously at the board, a human player can identify them and decide which one worth more to invest, I thinks this is a really difficult task in Go. How does he do this judge ? Which features does he analyze? And there are cases which "try" to mimic the biological solution is worth. See Neural Networks, Ant Colony Optimization Algorithm, Genetic Algorithm, etc. Santos, Gabriel. On Mon, Apr 1, 2013 at 2:30 PM, Álvaro Begué <[email protected]> wrote: On Mon, Apr 1, 2013 at 11:10 AM, Gabriel .Santos <[email protected]> wrote: I know that it is a lot of questions, but in order to get a computer go machine to outperform a human player I think that the machine should to ratiocinate like a human player. Do you also think a machine that carries people very fast should have strong legs like a horse? And a machine that can fly should flap its wings like a bird? And a closer example: Do you think the same thing about chess machines? In all those cases the engineering solution to the problem was very different from the biological solution, and I expect the same will happen with computer go. Actually, it's already happening. Álvaro. _______________________________________________ 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
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
