Re: [computer-go] Kinds of Zobrist hashes
Hi! Thanks to all for replies. On Tue, Dec 08, 2009 at 09:30:47PM -0500, Eric Boesch wrote: You can mathematically prove the two systems are almost the same, so there's no need to test. Yes, this was my line of thought, but I wasn't sure if I'm not missing anything... Petr Pasky Baudis ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Bursaries for Computer Go
I have some bursary funding available for a computer Go project at Stellenbosch University in South Africa, at both masters and doctoral level. Although they would be useful, knowledge of machine learning/artificial intelligence techniques and skill in the game of Go are not necessary. However, applicants should be good programmers, and preferably have some background in mathematics and statistics. Applicants require an Honours or 4-year degree in a suitable field (e.g. Computer Science, Statistics, Applied Maths, Electrical Engineering). Bursary amounts (fulltime per year): Doctoral: R45000; Masters: R3. If you might be interested, please contact me. If you know of other people who might be interested, please forward this to them. [I also have some funding for robotic soccer strategy - see http://www.cs.sun.ac.za/postgrad-opportunities/ .] Regards, -- Steve Kroon | Computer Science Division, Stellenbosch University (084) 458 8062 (Cell) | (086) 655 4386 (Fax) | kroonrs (Skype) http://www.cs.sun.ac.za/~kroon | kr...@sun.ac.za ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Kinds of Zobrist hashes
On Tue, Dec 08, 2009 at 09:30:47PM -0500, Eric Boesch wrote: You can mathematically prove the two systems are almost the same, so there's no need to test. Yes, this was my line of thought, but I wasn't sure if I'm not missing anything... If you ever decide to test which is faster, please post the results, I'm curious about how expensive the branch prediction miss is when using two values :-) Christian ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Kinds of Zobrist hashes
2009/12/9 Christian Nentwich christ...@modeltwozero.com: On Tue, Dec 08, 2009 at 09:30:47PM -0500, Eric Boesch wrote: You can mathematically prove the two systems are almost the same, so there's no need to test. Yes, this was my line of thought, but I wasn't sure if I'm not missing anything... If you ever decide to test which is faster, please post the results, I'm curious about how expensive the branch prediction miss is when using two values :-) I don't think there is any branching involved. When you place a stone, you add zobrist_table[point][color]. When you remove it, you subtract it. That's all you need to do. If you had a value for empty, you would have to add and subtract zobrist_table[point][color]-zobrist_table[point][empty]. Nothing else changes. Álvaro. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Kinds of Zobrist hashes
On Tue, Dec 8, 2009 at 8:37 PM, David Fotland fotl...@smart-games.com wrote: I use two values. I never even occurred to me to use three. I use one, it never occurred to me to use two :) At the time I'd never heard of Zobrist, but I used to use a single value for each point to look up Joseki. By using white-value = -black-value I could recognise the same joseki with colors inverted simply by using -value. Mark ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Fuego parameter question
To run Fuego in Windows, there the build I posted in: http://www.mail-archive.com/computer-go@computer-go.org/msg12208.html There is both source and binary. It is 40% faster than the cygwin version and more efficient in all senses, just because it is a native build. cygwin has serious limitations. When I did a lot of evaluation using Mogo as an oracle I finally had to use a Linux box, because cygwin is not able to run more than 15M sims (from Mogo) reliably. (If you run less than 15M kill the process and run it again you can use it without limits.) Jacques. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Kinds of Zobrist hashes
1. The 3 hashes: As Eric Boesch and Álvaro said, you can get the same codes as in the Black[], White[], Empty[] scenario using only 2 BlackXorEmpty[] and WhiteXorEmpty[] 2. What does make sense is making a difference between a POSITIONAL hash and a SITUATIONAL hash. The positional hash is the typical Zobrist hash of just the stones. The situational hash is the XOR of the positional hash and all moves forbidden by ko or superko (if any). I use another table of 361 64 bit codes for each forbidden move. At least, the tree search must have this in consideration, because the same stones but different legal moves, it is not the same position. I guess most programs do this one way or another. Jacques. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Kinds of Zobrist hashes
On Wed, Dec 9, 2009 at 1:19 PM, Jacques Basaldúa jacq...@dybot.com wrote: 1. The 3 hashes: As Eric Boesch and Álvaro said, you can get the same codes as in the Black[], White[], Empty[] scenario using only 2 BlackXorEmpty[] and WhiteXorEmpty[] 2. What does make sense is making a difference between a POSITIONAL hash and a SITUATIONAL hash. The positional hash is the typical Zobrist hash of just the stones. The situational hash is the XOR of the positional hash and all moves forbidden by ko or superko (if any). I use another table of 361 64 bit codes for each forbidden move. At least, the tree search must have this in consideration, because the same stones but different legal moves, it is not the same position. You can take this further and say that if it's the same stone configuration, the same legal moves but a different set of visited stone configurations, it is not the same position. The legal moves might be the same now, but they might be different after I perform a particular move. I usually understand that the Zobrist key just hashes the stone configuration, and I have to mess with it if I want to add more information (side to move, ko...). Álvaro. I guess most programs do this one way or another. Jacques. ___ 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] Kinds of Zobrist hashes
Álvaro Begué wrote: The legal moves might be the same now, but they might be different after I perform a particular move. You are right. Looks like my solution is not perfect, but it works good enough. I have seen trouble, before I implemented it, because the transposition table identified the same position in different situations. My solution worked, but, as you say it may not be perfect. Anyway, it is good enough for ko and my version of superko is a simplification restricted to the last 8 moves. Basically, triple ko and double ko of a group with one eye are the only superko sequences I have seen in serious games. I though it was good, now I have added a remark to the source code just in case some weird behavior appears. Thanks. Jacques. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Congratulations to Zengg!
Congratulations to Zengg, winner of Sunday's KGS bot tournament. My report is at http://www.weddslist.com/kgs/past/54/index.html Nick -- Nick Weddn...@maproom.co.uk ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Re: Congratulations to Zengg!
Nick Wedd: pjmml6zb9ailf...@maproom.demon.co.uk: Congratulations to Zengg, winner of Sunday's KGS bot tournament. My report is at http://www.weddslist.com/kgs/past/54/index.html Thank you Nick. The tournament was quite interesting and exciting. Zengg9 Zengg, platform unspecified but probably a cluster Zengg9 run on a (mini) pc cluster, four quad-core (an Intel Core i7 920, two Intel Core2 Q9550 and a Q6600; all run at 3 GHz) handcraft computers connected via a Gigabit Ethernet LAN. Eight threads run on the i7. Hideki -- g...@nue.ci.i.u-tokyo.ac.jp (Kato) ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/