Re: [computer-go] f(score) instead of sign(score)
> These ideas are all old,
I never said they were new. I wanted to give a mathematical argument on
them.
What would have been new would have been methods with filters applied on
the \hat{p}_i. However, though I am pretty sure I could make them more
efficient with little data, that's certainly not the case with much
data. Now the question would be whether this could make better programs,
with work more or less on the leaves. That's something I cannot even
make suggestions on without real data and undertanding precisely what
the bot does (and much work).
> see for example "Old fashioned Computer Go
> vs Monte Carlo Go":
>
> http://ewh.ieee.org/cmte/cis/mtsc/ieeecis/tutorial2007/Bruno_Bouzy_2007.pdf
That's really interesting ! That's a whole directory of ideas ! Thank
you for the link !
At least my next one still seems to be original.
Jonas
___
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Way MC plays
Jonas Kahn wrote: > I mean, if Crazy Stone played against himself from the position where > Katsunari was thought to be ahead, would he win with its original color, > or with Katsunari's ? That is a very big question. I hope it would win with Katsunari's stones *if* Katsunari was really ahead. I have been now working for some months in knowledge extracted from over 50K games played by pros and high dan players (both players), enough time, no handicap. I don't know for sure if that was not in vane yet, that's why I write "I hope". I certainly don't have a high opinion on any of the strong program's fuseki. These programs are not strong for their fuseki, I think, they are strong for many reasons and their fuseki is not really as bad as it looks. But I guess there is a lot of margin for improvement there. After the success of MC based programs some people argue that computers should play intrinsically different from humans and that computer play is superior. I don't share that opinion. I think that all what humans have elaborated on fuseki makes a lot of sense, when you apply it, you improve a lot. I don't see a reason why that would not apply to computers. Jacques. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Way MC plays
Jacques Basaldúa wrote: > Jonas Kahn wrote: > > > I mean, if Crazy Stone played against himself from the position where > > Katsunari was thought to be ahead, would he win with its original > color, > > or with Katsunari's ? > > That is a very big question. I hope it would win with Katsunari's > stones *if* > Katsunari was really ahead. I have been now working for some months in > knowledge extracted from over 50K games played by pros and high dan > players (both players), enough time, no handicap. I don't know for sure > if that was not in vane yet, that's why I write "I hope". I personally have serious doubts about knowledge extraction from human games, but I hope you have success.I think you can get more from computer games of strong players even though the level is weaker. Here is why I say that: 1. A strong computer still plays a lot of good moves - so the delta between human and computer games is not as high as you think. 2. A certain consistency in computer games that humans don't possess. 3. You have access to the internals, such as a score that quantifies moves. I still believe however there is something to be learned from these games - I want to see how your research comes out. You will let us know? > > I certainly don't have a high opinion on any of the strong program's > fuseki. > These programs are not strong for their fuseki, I think, they are > strong for > many reasons and their fuseki is not really as bad as it looks. But I > guess > there is a lot of margin for improvement there. > > After the success of MC based programs some people argue that > computers should play intrinsically different from humans and that > computer play is superior. Who ever said that? Clearly humans are still superior - no question. As far as playing differently, every non-trivial game playing entity has it's own unique style but as a class there are clear differences.I don't think that matters, it's possible to be equally strong and have a much different style, even between 2 perfect playing entities. If computers really get strong, they will be strong for different reasons than humans and I suspect humans will still perceive them as weaker because we judge strength mostly on how well it does things WE do well, not how well it does other things better. The most typical example is that rarely does someone judge the strength of a computer on it's low blunder rate. People don't even judge their own strength based on their blunder rate - they just slap themselves and pass it off as a fluke that should "never happen" and doesn't have anything to do with their skill as a player or their ability to win. We like to ignore what seems like "noise." > I don't share that opinion. I think that all > what humans have elaborated on fuseki makes a lot of sense, when > you apply it, you improve a lot. I don't see a reason why that would > not apply to computers. Clearly humans play better in the opening from everything I've seen and read about this. - Don > > 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/
Re: [computer-go] 9x9 scalability study
Michael Williams wrote: > Has the full game collection from this study been posted anywhere? I > couldn't find any reference to it in the email list. And I didn't see > any reference to it on the 929 scalability results page. Here are all the games I could collect. I'm not sure whether this included every single game or not as this depended on the testers sending me the games. The games are in one giant text file, I don't know if game readers can handle that, but if not you separate them easily enough with a text editor, or a simple script. Also, please note that the game numbers are not unique, for instance there may be many games labeled as game number 7. http://hsrf-mact.cse.ucsc.edu/~drd/scale9x9.zip - 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] really long thinking times
I'm playing Leela a 13x13 game at level 20 -- and I'm curious, has anyone ever tried either Leela or Mogo at very long thinking times, as in one-or-two-moves-per-day? ( Level 20 isn't that long -- it appears to be a few minutes per move -- I'll have to watch the clock ) Terry McIntyre <[EMAIL PROTECTED]> Wherever is found what is called a paternal government, there is found state education. It has been discovered that the best way to insure implicit obedience is to commence tyranny in the nursery. Benjamin Disraeli, Speech in the House of Commons [June 15, 1874] Never miss a thing. Make Yahoo your home page. http://www.yahoo.com/r/hs ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
