Re: [computer-go] Other uses of dynamic komi
> is dynamic komi not also usefull for > > - early pass, (only play if a move is better than a pass) Maybe, as you can define "better move" by "there is a dynkomi x where one loses while the other wins". But you'd need O(log(x_max)) runs to find it out. > - preventing horizon effect. (pushing losses over the search horizon) > (not so sure about this but MTSC/ utc doesn't seem to prevent it on > its own, what i find s bit puzelling on its self) I don't know how to do this - do you have a concrete idea? > -strong play in lost games. preventing braindead moves > (even if the game is lost the program can still play strong, just > adjust the komi so that some moves are better than others) > resignation can than be defined as points behind. (resign if more > than 5 points behind but keep playing if 1-4 points behind) > > -strongest play at won positions, (if all moves lead to a win, dynamic > komi can decide what is the best move) Definitely. > - whole number komi. > for player just play like the komi higher (for white) or lower (for > black) so for the program a dfraw become a win for the player. > (but even then a win is better than a draw, so it is all a bit > fidely but whith the progress on the 9x9 front, when solving small > board go comes into view the problems around whole number komi needs > addresing. "If all moves lose with -0.5 dynkomi, play the best move at +0.5 dynkomi", so yes. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] any mac programmers out there?
I tried both llvm-gcc and CLANG. I did not have any trouble getting them to work for my 64 bit chess program. I didn't try too hard, but neither is producing executables as fast as gcc. llvm-gcc is the slowest about 20% slower than gcc and clang is only a little slower than gcc. Since I developed with gcc it is very likely that the program and the way I write code is "tuned" to work well with gcc. Perhaps I will try this with the GO program, which is not heavily optimized. I grabbed and compiled the latest llvm and clang - so I cannot be accused of using outdated versions. And I didn't use the debug versions either. But I will keep my eye on llvm and clang. - Don 2009/9/6 Mark Boon > > On Sep 5, 2009, at 4:41 AM, terry mcintyre wrote: > > Found an interesting article on Snow Leopard at Ars Technica ... 20-some > pages. > > http://arstechnica.com/apple/reviews/2009/08/mac-os-x-10-6.ars > > Of interest to Computer Go programmers: the addition of blocks to C, which > allow closures and other fun stuff, much like Lisp. LLVM, which allows JIT > compilation to multiple architectures, including GPUs; Grand Central > Dispatch, which provides very light-weight concurrency; and CLANG, a new > compiler which is said to be quite an improvement over GCC. Open CL, which > leverages LLVM to program GPUs. > > > > Seems interesting indeed. Does anyone know how Objective-C 2.0 compares in > speed to C? I like the promise of abstracting the CPU to the point where you > can execute either on the CPU or the GPU, depending on which is available > and which is suitable. I also like the blocks, it seems a little more > elegant and more flexible than anonymous functions in Java. Combined with > light-weight concurrency makes for an interesting combination. > > Mark > > > ___ > computer-go mailing list > computer-go@computer-go.org > http://www.computer-go.org/mailman/listinfo/computer-go/ > ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Parameter Tuning by the Cross-Entropy Method
I've just been reading the paper /Parameter Tuning by the Cross-Entropy Method/ from Guillaume Chaslot et al. Unless I'm missing it, the paper doesn't say what step size (their 'alpha') they used in the experiments with Mango. Has anyone here tried this technique? What step size did you use? -M- ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
