Re: [Bug-gnubg] GNU Backgammon overview/background

[Thomas Hauk]

On Tue, 11 Nov 2003, Øystein Johansen wrote:
> > http://www.cs.ualberta.ca/~hauk/public/wrapper.ps
> 
> \begin{Minor silly comment}
> Jørn and Øystein, can be written i \LaTeX{} like this:
> J\oe rn and \Oe ystein
> 
> It's a long time since I used \LaTeX{} now, so correct me if I'm wrong!
> 
> \end{Minor silly comment}

Fair nuff... it's actually just \o and \O, but I added them in all the 
same. I didn't think to do it since I just wanted to get the thing out.

> I'm not sure about the start time of gnubg. It started with Garys toys 
> costello and abbott.

He actually sent me a link to a detailed interview with lots of history, 
so I'll revamp based on that:

http://gtw.freeshell.org/interview.html

> I don't like the sound of the crashed neural net definition. What about 
> rewriting to:

Changed it a bit.

> About the training:
> Temporal difference learning was only used in the first stages of the 
> program. With this kind of training the program reached a intermediate 
> playing strength. Later the nets have been trained with supervised 
> training with positions from a position database. The desired outputs of 
> the training positions in the database are based on rollouts. The latest 
> contact and crashed nets are trained by a slighly different technique. 
> It's like the net is trained to play like the 2-ply evaluation.

So... let me get this straight.

TD learning was employed to train a NN, which then produced a position 
database along with what the NN thought to be the best move for that 
position. Then a second NN was trained with only this resulting position 
DB? In other words, the second NN is trained with expertly-labelled data, 
where the expert is an older NN trained only with self-play?

> Race net stuff

Answered by Joseph Heled.

> If you don't understand the bearoff database calculation and OSR, you 
> should also look at the article by M. Buro.
> Efficient Approximation of Backgammon Race Equities.
> ICCA Journal, Vol. 22, No. 2, pp. 98-102. ISSN 0920-234X

Michael Buro is actually one of my two supervisors ;-)

He told me he was going to use the weekend to look through his race 
papers. We might do some work using race doubling decisions and/or use of 
ProbCut techniques.

> \begin{quote}
> Unfortunately, gnubg has an unusual definition of ply
> \end{quote}
> 
> I strongly disagree. This is the way most chess engines count their 
> plies. This is also the way JellyFish counts it's plies. I really wonder 
> why "everybody" thinks it's GNU Backgammon that's the sinner of this 
> naming convention, since the truth is that the sinner is Snowie?
> As I type this, I feel this discussion has been typed before. I suggest 
> you search the archives of the mailing list.

Compared to the rest of the AI world, yes, it _is_ an unusual definition 
of ply. For example, a quick search reveals

"A computer program which evaluates no further than its own legal moves 
plus the legal responses to those moves is searching to a depth of 
two-ply."
http://chess.about.com/library/glossary/bldefply.htm

Basically, a ply is a new level of nodes in a search tree, or simply a 
move by a player. This is the accepted definition in AI.

Ply is also normally the depth of the tree. But in games where the search 
trees have intermediary levels of chance nodes -- like backgammon -- then the 
defintion of ply is [ply = (depth+1)/2]; a depth=1 search would be 1-ply, 
a depth=3 search would be 2-ply, etc.

> Strongest in the World?
> I actually feel more confident with the analysis done by Michael Depreli 
> at GammOnLine. Search net for his results. He's probably reading this 
> mail, so I guess he can mail you his methods and final results.

I hope he does. I don't subscribe to GammonLine so someone will have to 
help me out here ;-)

> Albert Silver have also run a series of 7-point matches agains Snowie4. 
> (That was before the latest 0.14 net). Check this thread out:
> http://mail.gnu.org/archive/html/bug-gnubg/2003-06/msg00096.html
> Jørn did a luck analysis of Alberts matches and concludes:
> http://groups.google.com/address@hidden

Thanks, I've added those to my list.

And thanks for all the feedback. 8)

-- 
Some people, when confronted with a problem, think "I know, I'll use 
regular expressions."  Now they have two problems.
  --Jamie Zawinski, in comp.lang.emacs