Re: computer-go: Perl Module for next move.

[Toni Cebrian <cebrian@xxxxxx>]

Perphaps what you all are looking for is the work from J. Baxter et al.
applied to the game of chess. Check out
"TDLeaf(Lambda): Combining temporal difference learning with game-tree
search"

    http://cs.anu.edu.au/~Lex.Weaver/pub_sem/publications/AJIIPS_98.pdf

Cheers,

    Toni.


"Staelin, Carl" wrote:

> The original work on using temporal difference
> learning for game players was by Gerald Tesauro
> to train a backgammon player.  A copy of the
> paper can be found at:
>
>         http://www.research.ibm.com/massive/tdl.html
>
> Others thought it might be a nice technique to
> apply to other games, such as checkers or GO.
> Chellapilla and Fogel trained checkers players
> using techniques similar to those used by
> Tesauro for his backgammon player.
>
>         http://vision.ucsd.edu/~kchellap/papers
> (Look for Chellapilla and Fogel, "Co-evolving
> checkers playing programs using only win, lose
> or draw", from SPIE 1999.
>
> There has also been some work done on applying
> similar techniques to GO players, such as Nici
> Schraudolphs work on applying temporal difference
> learning neural networks to GO:
>
>         http://www.idsia.ch/~nic/pubs.html#gochap
>
> Cheers,
>
> Carl
> _________________________________________________
> [(hp)]  Carl Staelin
>         Senior Research Scientist
>         Hewlett-Packard Laboratories
>         Technion City
>         Haifa, 32000
>         ISRAEL
>         +972(4)823-1237x221     +972(4)822-0407 fax
>         staelin@xxxxxxxxxxxxxxxxx
> _______http://www.hpl.hp.com/personal/Carl_Staelin_______
>
> > -----Original Message-----
> > From: Matthew Corey Brown [mailto:bromoc@xxxxxxxxxxxxxxxxx]
> > Sent: Wednesday, June 06, 2001 12:51 AM
> > To: computer-go@xxxxxxxxxxxxxxxxx
> > Subject: RE: computer-go: Perl Module for next move.
> >
> >
> > On Tue, 5 Jun 2001, Mark Boon wrote:
> >
> > > In a previous post a checkers project was said to use neural nets
> > > successfully, but I'm wondering how much was done by a fast
> > mini-max program
> > > which used the neural net in some way (as a move-selector
> > for example)
> > > instead of having the net learn the mini-max strategy as
> > well. I somehow
> > > feel the latter is necessary for neural networks to be
> > successful in Go.
> > >
> >
> > What i was told all it did was say which move you made result
> > in a better
> > position.. it never looked ahead.. just what was best in the here and
> > now and which move would it be to improve your position. I
> > learned it from
> > a SIG in AI that the local game dev community has started up.
> > I'm going to
> > a different SIG meeting tonight but the person who told me
> > about it will
> > be there and I'll try and get more reference. But the neural
> > net was the
> > only descion maker of the checkers game to my understanding. The idea
> > intrigued me when i heard it last week.
> >
> >
> > Matthew Corey Brown
> > bromoc@xxxxxxxxxxxxxxxxx
> >   "Death can not stop true love. All it can do is delay it
> > for awhile."
> >