Re: [computer-go] UCT outside of go?

2007-06-06 Thread Martin Møller Skarbiniks Pedersen
On 04/06/07, Darren Cook [EMAIL PROTECTED] wrote: Does anyone know of UCT being used in games other than go, or outside games altogether, such as travelling salesman problem, or some business-related scheduling/optimizing/searching problem domain? I am trying to use UCT for the game trax. I

Re: [computer-go] UCT outside of go?

2007-06-06 Thread Roland Illig
Darren Cook wrote: Does anyone know of UCT being used in games other than go, or outside games altogether, such as travelling salesman problem, or some business-related scheduling/optimizing/searching problem domain? I have used it in a connect4 engine.

Re: [computer-go] Efficiently selecting a point to play in a random playout

2007-06-06 Thread Peter Drake
This sounds a lot like the roulette wheel selection scheme used in genetic algorithms. The idea is that each candidate has a different slice of a roulette wheel, with better candidates getting bigger slices. Peter Drake http://www.lclark.edu/~drake/ On Jun 6, 2007, at 2:07 AM, Jacques

Re: [computer-go] Efficiently selecting a point to play in a random playout

2007-06-06 Thread Rémi Coulom
Álvaro Begué wrote: Actually, John had a better idea to do this. In two words: binary tree. The root represents the whole board, and it contains the sum of the probabilities of all the points (you don't need to force this to be 1, if you use non-normalized probabilities). This node points to

Re: [computer-go] UCT outside of go?

2007-06-06 Thread Nicholas Halderman
In case anyone else was interested, yes, the AAAI general game-playing competition has just started. You can find instructions to join the mailing list on the resources page at the site (games.stanford.edu) and evidently the games from the competition are being broadcast live online. If it's

Re: [computer-go] Efficiently selecting a point to play in a random playout

2007-06-06 Thread Jason House
On 6/6/07, Rémi Coulom [EMAIL PROTECTED] wrote: I wonder if other people had thought about this before... Álvaro. Yes, I did it in the beginning. But I found that it is faster to divide by more than two. Currently, I keep the probability of the whole board, each line, and each point. It is

Re: [computer-go] Efficiently selecting a point to play in a random playout

2007-06-06 Thread Graham Thomson
I would be weary of using java.util.Random - it is not that random: http://alife.co.uk/nonrandom/. A drop in Mersenne Twister replacement for java.util.Random is available at http://cs.gmu.edu/~sean/research/. Cheers, Graham. On 05/06/07, Peter Drake [EMAIL PROTECTED] wrote: Oddly, there

Re: [computer-go] Efficiently selecting a point to play in a random playout

2007-06-06 Thread Rémi Coulom
Jason House wrote: On 6/6/07, *Rémi Coulom* [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] wrote: I wonder if other people had thought about this before... Álvaro. Yes, I did it in the beginning. But I found that it is faster to divide by more than two. Currently, I

Re: [computer-go] Efficiently selecting a point to play in a random playout

2007-06-06 Thread Peter Drake
Thanks for the tip. It does seem a bit faster (5% speedup of the program overall), and I'm willing to accept the consensus that the randomness is better. Peter Drake http://www.lclark.edu/~drake/ On Jun 6, 2007, at 2:15 PM, Graham Thomson wrote: I would be weary of using java.util.Random

Re: [computer-go] Efficiently selecting a point to play in a random playout

2007-06-06 Thread Peter Drake
On Jun 6, 2007, at 2:41 PM, Rémi Coulom wrote: Also, if you have a clever probability distribution, the range of values for each move will be very large. For instance, here are two 3x3 shapes used by Crazy Stone (# to move): O O # # . . # O # Gamma = 143473; . # # . . . . . . Gamma