[computer-go] cgos3: Small fix
I noticed that the simple tcl client outputs many lines like this: 09:21:30S-C info Estimated time until next round: 06:53 09:21:30Estimated time until next round: 06:53 As those scroll interesting info out of my screen, I disabled the line that outputs the second line. All the info is already in the first one. Maybe the server could send those messages a bit less often too? Thanks again for cgos, it has turned out to be a fun thing. I am again playing with my 'halgo' series, going for some sort of search next... Regards Heikki -- Heikki Levanto In Murphy We Turst heikki (at) lsd (dot) dk ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Re: cgos3: Small fix
The info messages were designed for an eventual graphical client. The idea being to send this and perhaps other info that could be displayed in a separate window. The S-C is server to client for debugging. So really I could turn off the lines that have the messages going back and forth. You could pipe these messages through grep or make your own filter that just shows what you want. Since I already have cgosview it would be real easy to modify it to make an engine client that watches your programs games graphically, as they are being plays. The windows would only show the games of the client program. Could even have a provision to show the output of the engine if it were put to stderr. Keep up the good work with Halgo. If you add a search it makes a big difference. - Don On Mon, 2007-05-21 at 09:25 +0200, Heikki Levanto wrote: I noticed that the simple tcl client outputs many lines like this: 09:21:30S-C info Estimated time until next round: 06:53 09:21:30Estimated time until next round: 06:53 As those scroll interesting info out of my screen, I disabled the line that outputs the second line. All the info is already in the first one. Maybe the server could send those messages a bit less often too? Thanks again for cgos, it has turned out to be a fun thing. I am again playing with my 'halgo' series, going for some sort of search next... Regards Heikki ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Re: Amsterdam 2007 paper
On 5/18/07, Rémi Coulom [EMAIL PROTECTED] wrote: My idea was very similar to what you describe. The program built a collection of rules of the kind if condition then move. Condition could be anything from a tree-search rule of the kind in this particular position play x, or general rule such as in atari, extend. It could be also anything in-between, such as a miai specific to the current position. The strengths of moves were updated with an incremental Elo-rating algorithm, from the outcomes of random simulations. The obvious way to update weights is to reward all the rules that fired for the winning side, and penalize all rules that fired for the losing side, with rewards and penalties decaying toward the end of the playout. But this is not quite Elo like, since it doesn't consider rules to beat each other. So one could make the reward dependent on the relative weight of the chosen rule versus all alternatives. increasing the reward if the alternatives carried a lot of weight. Is that how your ratings worked? I'm not sure how that compares with TD learning. Maybe someone more familiar with the latter can point out the differences. TD learning (with linear function approximation) uses a gradient descent rule to update weights. The simplest gradient descent rule, LMS or Widrow-Hoff, does something like you describe: rules that are followed by positive reward (win) are increased in weight, and rules that are followed by negative reward (loss) are decreased. The exact update depends on the set of rules firing, and is proportional to the error between the estimated reward (based on all rules) and the actual reward. In other words, each weight is updated a little towards the value which would have made a correct overall prediction. TD learning is similar, except that it updates weights towards a subsequent prediction of the reward (e.g. on the next move), instead of the actual reward. Rich Sutton gives a much better explanation than me: http://www.cs.ualberta.ca/%7Esutton/book/ebook/the-book.html ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] 9x9 vs 19x19 (was: computer-go Digest)
On Mon, May 21, 2007 at 10:46:24AM -0600, David Silver wrote: But... in practice, I haven't got good results on larger boards. But to be honest, I've focused much more on 9x9, so perhaps I've missed some simple tricks. I think there has been a marked change of interest since the introduction of UCT, and - around the same time - the cgos 9x9 tournament page. I understand that most people do their experiments on 9x9, the results are available so much faster. Still, I think it might be time to loosen the focus on 9x9, and have some more things happening on other sizes. Would there be interest in a tournament system for 19x19 programs? Something like 30 mins / player sounds like a reasonable extrapolation. -H -- Heikki Levanto In Murphy We Turst heikki (at) lsd (dot) dk ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest)
There has been much talk of a 19x19 CGOS and I have had people offer systems to run it on. I think Dave Dyer also would let us run a 19x19 version. David Doshay has offered some space for it too - which is what I am leaning towards right now. I haven't implemented any of the handicap stuff, but CGOS would work otherwise without any change. Maybe that is good way to start. I still have this horrible fear that 9x9 would suffer if several programs moved over to 19x19. Or perhaps BOTH would suffer from a lack of a variety of competition. What might be useful is if there was an organized system of volunteers who might be willing to run an instance of your program on request for a few days? CGOS has not been particularly stable until just recently. I fixed a bug about 10 days ago and it has been running without a problem since then, so I think it's reasonably stable now.I was not willing to think about another server until I can be sure that it's possible to run for weeks in a row without bugs. But maybe the time is just about right. - Don On Mon, 2007-05-21 at 19:58 +0200, Heikki Levanto wrote: On Mon, May 21, 2007 at 10:46:24AM -0600, David Silver wrote: But... in practice, I haven't got good results on larger boards. But to be honest, I've focused much more on 9x9, so perhaps I've missed some simple tricks. I think there has been a marked change of interest since the introduction of UCT, and - around the same time - the cgos 9x9 tournament page. I understand that most people do their experiments on 9x9, the results are available so much faster. Still, I think it might be time to loosen the focus on 9x9, and have some more things happening on other sizes. Would there be interest in a tournament system for 19x19 programs? Something like 30 mins / player sounds like a reasonable extrapolation. -H ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest)
There has been much talk of a 19x19 CGOS and I have had people offer systems to run it on. I think Dave Dyer also would let us run a 19x19 version. ... I still have this horrible fear that 9x9 would suffer if several programs moved over to 19x19. Or perhaps BOTH would suffer from a lack of a variety of competition. Why not 13x13 before 19x19? ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest)
I have a dual-core AMD64 which is unused and connected to the internet for a most of the day, and would be delighted to volunteer it for running an instance of a 19x19 go program for cgos. Terry McIntyre UNIX for hire software development / systems administration / security [EMAIL PROTECTED] - Original Message From: Heikki Levanto [EMAIL PROTECTED] To: [EMAIL PROTECTED]; computer-go computer-go@computer-go.org Sent: Monday, May 21, 2007 12:13:43 PM Subject: Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest) On Mon, May 21, 2007 at 02:58:25PM -0400, Don Dailey wrote: I still have this horrible fear that 9x9 would suffer if several programs moved over to 19x19. Or perhaps BOTH would suffer from a lack of a variety of competition. I see your point. What might be useful is if there was an organized system of volunteers who might be willing to run an instance of your program on request for a few days? Let me be the first one to volunteer for that. I have a dual-core AMD-64 as my workstation, and as long as I have to be at the office every day, it has lots of unused cpu power... The next few days I may be running my Halgo on it, but I'm sure it can handle another program without me noticing any ill effects. CGOS has not been particularly stable until just recently. I fixed a bug about 10 days ago and it has been running without a problem since then, so I think it's reasonably stable now.I was not willing to think about another server until I can be sure that it's possible to run for weeks in a row without bugs. I know the feeling - I am as much a sysadmin as a programmer by profession, and have great understanding about not putting code into production settings before it is ready for it. But maybe the time is just about right. Maybe - your decision! -Heikki -- Heikki Levanto In Murphy We Turst heikki (at) lsd (dot) dk ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ Get the free Yahoo! toolbar and rest assured with the added security of spyware protection. http://new.toolbar.yahoo.com/toolbar/features/norton/index.php___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest)
On Mon, 2007-05-21 at 15:13 -0400, Chris Fant wrote: There has been much talk of a 19x19 CGOS and I have had people offer systems to run it on. I think Dave Dyer also would let us run a 19x19 version. ... I still have this horrible fear that 9x9 would suffer if several programs moved over to 19x19. Or perhaps BOTH would suffer from a lack of a variety of competition. Why not 13x13 before 19x19? Whatever is done should be permanent. Do you think 13x13 would be as popular as 19x19? I'm leaning towards 19x19 since it is the standard board and would optionally consider doing 13x13 in addition. - Don ___ 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] Re: 9x9 vs 19x19 (was: computer-go Digest)
I suggest that it would be more convenient for everyone if various sizes of cgos all ran on the same server. If you want to donate horsepower to the project, a good use of the resource would be to run anchorman type clients. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest)
On Mon, May 21, 2007 at 03:13:09PM -0400, Chris Fant wrote: Why not 13x13 before 19x19? Because the next step would be 15x15, and then 17x17, and when (if) we get to 19x19, there are so few competitors around that the whole tournament won't make any sense. I think it is better to stick to 9x9 as the beginners tournament, where it is easy to test new ideas in quick games, and 19x19 as the serious tournament where we can see how good computers are at playing the game like we humans do. Just my humble opinion, of course. - Heikki -- Heikki Levanto In Murphy We Turst heikki (at) lsd (dot) dk ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest)
I agree 9x9 is wonderful, but a 19x19 for deep testing would be nice. To many variations and you risk the threat of diluting the engine pool. -Josh On 5/21/07, Heikki Levanto [EMAIL PROTECTED] wrote: On Mon, May 21, 2007 at 03:13:09PM -0400, Chris Fant wrote: Why not 13x13 before 19x19? Because the next step would be 15x15, and then 17x17, and when (if) we get to 19x19, there are so few competitors around that the whole tournament won't make any sense. I think it is better to stick to 9x9 as the beginners tournament, where it is easy to test new ideas in quick games, and 19x19 as the serious tournament where we can see how good computers are at playing the game like we humans do. Just my humble opinion, of course. - Heikki -- Heikki Levanto In Murphy We Turst heikki (at) lsd (dot) dk ___ 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/
Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest)
You missed 11x11. I used to test a lot with 11x11. I think it's a great size, a big step up from 9x9 and more go-like than 9x9 but still easy to test. But I agree with Heikki - we probably don't want too many variants. Perhaps I set up 19x19 tomorrow on Dave Dyers server. I'm think 20 minutes per player - 30 minutes is better but there would be a lot of waiting around - up to 1 hour per game. Taking votes now. 15, 20 or 30? - Don On Mon, 2007-05-21 at 23:10 +0200, Heikki Levanto wrote: On Mon, May 21, 2007 at 03:13:09PM -0400, Chris Fant wrote: Why not 13x13 before 19x19? Because the next step would be 15x15, and then 17x17, and when (if) we get to 19x19, there are so few competitors around that the whole tournament won't make any sense. I think it is better to stick to 9x9 as the beginners tournament, where it is easy to test new ideas in quick games, and 19x19 as the serious tournament where we can see how good computers are at playing the game like we humans do. Just my humble opinion, of course. - Heikki ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest)
Two variants will be difficult enough to support. If it's possible to build an infrastructure to permit volunteers to put their spare cycles to work for a various periods of time ( something like [EMAIL PROTECTED] ), perhaps there would be enough spare capacity to test a variety of programs. Terry McIntyre UNIX for hire software development / systems administration / security [EMAIL PROTECTED] - Original Message From: Don Dailey [EMAIL PROTECTED] To: computer-go computer-go@computer-go.org Sent: Monday, May 21, 2007 3:58:18 PM Subject: Re: [computer-go] 9x9 vs 19x19 (was: computer-go Digest) You missed 11x11. I used to test a lot with 11x11. I think it's a great size, a big step up from 9x9 and more go-like than 9x9 but still easy to test. But I agree with Heikki - we probably don't want too many variants. Perhaps I set up 19x19 tomorrow on Dave Dyers server. I'm think 20 minutes per player - 30 minutes is better but there would be a lot of waiting around - up to 1 hour per game. Taking votes now. 15, 20 or 30? - Don On Mon, 2007-05-21 at 23:10 +0200, Heikki Levanto wrote: On Mon, May 21, 2007 at 03:13:09PM -0400, Chris Fant wrote: Why not 13x13 before 19x19? Because the next step would be 15x15, and then 17x17, and when (if) we get to 19x19, there are so few competitors around that the whole tournament won't make any sense. I think it is better to stick to 9x9 as the beginners tournament, where it is easy to test new ideas in quick games, and 19x19 as the serious tournament where we can see how good computers are at playing the game like we humans do. Just my humble opinion, of course. - Heikki ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ Take the Internet to Go: Yahoo!Go puts the Internet in your pocket: mail, news, photos more. http://mobile.yahoo.com/go?refer=1GNXIC___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Re: 9x9 vs 19x19 (was: computer-go Digest)
On Mon, 2007-05-21 at 14:01 -0700, Dave Dyer wrote: I suggest that it would be more convenient for everyone if various sizes of cgos all ran on the same server. If you want to donate horsepower to the project, a good use of the resource would be to run anchorman type clients. Are you saying that you believe there should be a lot of anchor players? If so 1. If there are too many, it decreases the variety, you might be playing this single opponent a lot more. 2. I think a single anchor works. Having said that, I originally intended for there to be a couple of FatMan clones playing (as anchors) on other peoples computers.That would take the pressure off of me to always keep FatMan running. You wouldn't have to run it all the time, just hopefully a lot of the time. - Don ___ 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] Re: 9x9 vs 19x19 (was: computer-go Digest)
On Mon, 2007-05-21 at 16:52 -0700, Dave Dyer wrote: I figured that a credible anchor player for 19x19 might need a lot of cycles, and need to play a lot of games at first, so spreading the load would be a good idea. Yes, that's true. When the server first goes up, there are no rated players and the games of 2 unrated players have no affect on each other.It takes a long time for ratings to converge in this case. When I started the 9x9 server I seeded it before opening it up to everyone. I played a few hundred matches between the anchor, gnugo and other players I had access to so that there would be reasonable rating pool to start with. It's actually not a big problem if you pick a default rating that is close to the average - unfortunately you don't know what the average will be in advance! I will probably make some stable and common version of gnugo be the anchor for 19x19. If others want to run an instance of it I can set them up as anchors too - I just need to know the names of any bots set up to be anchors. I'll put up a 19x19 version tomorrow on boardspace at 30 minutes per game. - Don ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Re: Amsterdam 2007 paper
Rémi, May I ask you some more questions? (1) You define Dj as Dj=Mij*ci+Bij. Is it not Aij but Bij? What does this mean? (2) You have relatively few shape patterns. How large is each pattern? 5x5, 7x7, or more? (3) You say the nth move is added when 40*1.4^(n-2) simulations have been run. How did you determine these numbers? Thanks -- Yamato ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/