Re: [Computer-go] Congratulations to CrazyStone!
Rémi, Petr, Thank you for the corrections. I have update the report. Nick On 7 April 2015 at 04:24, Petr Baudis pa...@ucw.cz wrote: Hi, sorry for not mentioning it, Pachi was running on all threads of AMD FX(tm)-8350 Eight-Core Processor with 24GB RAM this time. On Mon, Apr 06, 2015 at 06:48:22PM +0100, Nick Wedd wrote: Congratulations to CrazyStone, winner of yesterday's KGS bot tournament on 13x13 boards! My report is at http://www.weddslist.com/kgs/past/111/index.html As usual I welcome your comments and corrections. -- Petr Baudis If you do not work on an important problem, it's unlikely you'll do important work. -- R. Hamming http://www.cs.virginia.edu/~robins/YouAndYourResearch.html ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go -- Nick Wedd mapr...@gmail.com ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
On Sat, Mar 28, 2015 at 11:51 AM, Petr Baudis pa...@ucw.cz wrote: On Sat, Mar 28, 2015 at 11:45:48AM +0100, Urban Hafner wrote: Good to know Petr! Where does the strength come from? Sophisticated playouts or a search algorithm or both? Frankly, I don't know for sure! The downside of Michi's slowness is that playtesting takes too many resources so configuration performance exploration is awkward. But my guess based on ad hoc tests during the development is that the contribution of basic playout heuristics and RAVE+priors may be about 1:1 (with large pattern priors giving further extra boost). Now that I have a bit of time again, what would be a good starting point to improve upon UCT and light playouts? RAVE definitely comes to mind, as well as enhancing the playouts with heuristics like the MoGo 3x3 patterns. Are there are any good papers on adding priors to the search tree (and where the underlying data is coming from)? I'm sure there must be, but I think I just don't know how to search for it. Urban -- Blog: http://bettong.net/ Twitter: https://twitter.com/ujh Homepage: http://www.urbanhafner.com/ ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
Urban Hafner: cahmxpnnkr-ixqou4stxi_dfojfhj7j_qccj_duit6uqw4o8...@mail.gmail.com: On Sat, Mar 28, 2015 at 11:51 AM, Petr Baudis pa...@ucw.cz wrote: On Sat, Mar 28, 2015 at 11:45:48AM +0100, Urban Hafner wrote: Good to know Petr! Where does the strength come from? Sophisticated playouts or a search algorithm or both? Frankly, I don't know for sure! The downside of Michi's slowness is that playtesting takes too many resources so configuration performance exploration is awkward. But my guess based on ad hoc tests during the development is that the contribution of basic playout heuristics and RAVE+priors may be about 1:1 (with large pattern priors giving further extra boost). Now that I have a bit of time again, what would be a good starting point to improve upon UCT and light playouts? RAVE definitely comes to mind, as well as enhancing the playouts with heuristics like the MoGo 3x3 patterns. Are there are any good papers on adding priors to the search tree (and where the underlying data is coming from)? I'm sure there must be, but I think I just don't know how to search for it. For prior values in the tree, almost(?) all strong programs use Remi's method these days. http://remi.coulom.free.fr/Amsterdam2007/MMGoPatterns.pdf Hideki -- Hideki Kato mailto:hideki_ka...@ybb.ne.jp ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
On Tue, Apr 7, 2015 at 12:56 PM, Petr Baudis pa...@ucw.cz wrote: Hi! On Tue, Apr 07, 2015 at 12:03:12PM +0200, Urban Hafner wrote: Now that I have a bit of time again, what would be a good starting point to improve upon UCT and light playouts? RAVE definitely comes to mind, as well as enhancing the playouts with heuristics like the MoGo 3x3 patterns. Well, my suggestions would be in the form of Michi (and its git history). ;-) I suspected you'd say something like this. ;) It is definitely on my list of things to steal a few things from Michi. But maybe I'll start with simpler and/or well defined things like RAVE or the hand picked MoGo 3x3 patterns. That way it's easy to see if I really screwed something up. The bot is still rather weak so adding some of those features should really improve the strength. Urban -- Blog: http://bettong.net/ Twitter: https://twitter.com/ujh Homepage: http://www.urbanhafner.com/ ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
On Tue, Apr 7, 2015 at 12:20 PM, Hideki Kato hideki_ka...@ybb.ne.jp wrote: For prior values in the tree, almost(?) all strong programs use Remi's method these days. http://remi.coulom.free.fr/Amsterdam2007/MMGoPatterns.pdf Thank you! I will put that one on my reading list! Urban -- Blog: http://bettong.net/ Twitter: https://twitter.com/ujh Homepage: http://www.urbanhafner.com/ ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
Hi! On Tue, Apr 07, 2015 at 12:03:12PM +0200, Urban Hafner wrote: Now that I have a bit of time again, what would be a good starting point to improve upon UCT and light playouts? RAVE definitely comes to mind, as well as enhancing the playouts with heuristics like the MoGo 3x3 patterns. Well, my suggestions would be in the form of Michi (and its git history). ;-) there are any good papers on adding priors to the search tree (and where the underlying data is coming from)? I'm sure there must be, but I think I just don't know how to search for it. Basically, you can either do progressive widening / unpruning / bias. The terminology is rather confusing. In case of progressive bias, you can either initialize the winrate with N wins (positive bias) and M losses (negative bias) with N and M determined by various heuristics (Fuego, Pachi, Michi use this), or have (1-alpha)*winrate + alpha*bias with bias being a hypothetical winrate determined by the heuristics and alpha: 1 - 0 as #simulations: 0 - infty (e.g. alpha=sqrt(c/n) or some other random formula like that). I know about no good survey papers personally. On Tue, Apr 07, 2015 at 07:20:37PM +0900, Hideki Kato wrote: For prior values in the tree, almost(?) all strong programs use Remi's method these days. http://remi.coulom.free.fr/Amsterdam2007/MMGoPatterns.pdf Do you mean all the strong programs do progressive widening? -- Petr Baudis If you do not work on an important problem, it's unlikely you'll do important work. -- R. Hamming http://www.cs.virginia.edu/~robins/YouAndYourResearch.html ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
Petr Baudis: 20150407105648.gp6...@machine.or.cz: On Tue, Apr 07, 2015 at 07:20:37PM +0900, Hideki Kato wrote: For prior values in the tree, almost(?) all strong programs use Remi's method these days. http://remi.coulom.free.fr/Amsterdam2007/MMGoPatterns.pdf Do you mean all the strong programs do progressive widening? No. I meant that strong programs use prior in various ways (just an initial value for a move, for example) but (almost) all use Remi's method to compute the prior. Hideki -- Hideki Kato mailto:hideki_ka...@ybb.ne.jp ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
On Apr 7, 2015, at 4:34 AM, Urban Hafner cont...@urbanhafner.com wrote: I suspected you'd say something like this. ;) It is definitely on my list of things to steal a few things from Michi. But maybe I'll start with simpler and/or well defined things like RAVE or the hand picked MoGo 3x3 patterns. That way it's easy to see if I really screwed something up. The bot is still rather weak so adding some of those features should really improve the strength. How many playouts (per move) does 'stop_0.9-2b’ do? Christoph ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
I suspected you'd say something like this. ;) It is definitely on my list of things to steal a few things from Michi. But maybe I'll start with simpler and/or well defined things like RAVE or the hand picked MoGo 3x3 patterns. That way it's easy to see if I really screwed something up. The bot is still rather weak so adding some of those features should really improve the strength. How many playouts (per move) does 'stop_0.9-2b??? do? 0 (none, zero, keine, geen) Folkert van Heusden ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
On Tue, Apr 7, 2015 at 4:04 PM, Christoph Birk b...@obs.carnegiescience.edu wrote: On Apr 7, 2015, at 4:34 AM, Urban Hafner cont...@urbanhafner.com wrote: I suspected you'd say something like this. ;) It is definitely on my list of things to steal a few things from Michi. But maybe I'll start with simpler and/or well defined things like RAVE or the hand picked MoGo 3x3 patterns. That way it's easy to see if I really screwed something up. The bot is still rather weak so adding some of those features should really improve the strength. How many playouts (per move) does 'stop_0.9-2b’ do? I wouldn't know, Christoph. My (and Igor's) bot is called Iomrascálaí. :P It's running as the various Imrscl-XYZ bots on CGOS due to the username length restriction and the fact that the current CGOS can't handle Unicode characters. This bot however does around 4.5k pips on 9x9 and 1k apps on 19x19 running on a 2,2 GHz Intel Core i7 (6 months only MacBook pro). The versions on CGOS run using 8 threads and I get a speedup of about 4.5x. Urban -- Blog: http://bettong.net/ Twitter: https://twitter.com/ujh Homepage: http://www.urbanhafner.com/ ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
On Apr 7, 2015, at 7:16 AM, Urban Hafner cont...@urbanhafner.com wrote: I wouldn't know, Christoph. My (and Igor's) bot is called Iomrascálaí. :P It's running as the various Imrscl-XYZ bots on CGOS due to the username length restriction and the fact that the current CGOS can't handle Unicode characters. This bot however does around 4.5k pips on 9x9 and 1k apps on 19x19 running on a 2,2 GHz Intel Core i7 (6 months only MacBook pro). The versions on CGOS run using 8 threads and I get a speedup of about 4.5x. thanks, I agree 1400 is about as far as simple UCT will get you. My simple UCT implementation (myCtest-xxk-UCT) gets about 1200, but it does not do any adjustments to the number of playouts per move depending on the time remaining, so I have to limit it to 40k playouts per move. Have you thought about using the partial tree of the previous move as a bias? Chrisoph ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
It doesn't matter with simple UCT, but it's a huge difference when you have more targeted playouts because you can find a subtree you really like and spend 99%+ of your time exploiting that particular line (for example, not dying in one move). I see Fuego re-use most of its existing tree on a 9x9 almost every move. On 2015-04-07 11:16, Urban Hafner wrote: On Tue, Apr 7, 2015 at 4:44 PM, Christoph Birk b...@obs.carnegiescience.edu mailto:b...@obs.carnegiescience.edu wrote: thanks, I agree 1400 is about as far as simple UCT will get you. My simple UCT implementation (myCtest-xxk-UCT) gets about 1200, but it does not do any adjustments to the number of playouts per move depending on the time remaining, so I have to limit it to 40k playouts per move. Have you thought about using the partial tree of the previous move as a bias? You mean keeping the subtree and not starting from scratch every time? That sounds interesting, but I don't think it would lead to a huge strength boost. But I guess I may give that a try. Pondering on the opponent's time would be the natural follow up for that of course. Urban -- Blog: http://bettong.net/ Twitter: https://twitter.com/ujh Homepage: http://www.urbanhafner.com/ ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] What's a good playout speed?
On Tue, Apr 7, 2015 at 4:44 PM, Christoph Birk b...@obs.carnegiescience.edu wrote: thanks, I agree 1400 is about as far as simple UCT will get you. My simple UCT implementation (myCtest-xxk-UCT) gets about 1200, but it does not do any adjustments to the number of playouts per move depending on the time remaining, so I have to limit it to 40k playouts per move. Have you thought about using the partial tree of the previous move as a bias? You mean keeping the subtree and not starting from scratch every time? That sounds interesting, but I don't think it would lead to a huge strength boost. But I guess I may give that a try. Pondering on the opponent's time would be the natural follow up for that of course. Urban -- Blog: http://bettong.net/ Twitter: https://twitter.com/ujh Homepage: http://www.urbanhafner.com/ ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
