Marc Lanctot on Sat, 24 Jul 2010 10:01:49 -0700 (MST)


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: [game-lang] a survey of previous work


On 07/24/2010 06:59 AM, Joel Uckelman wrote:
Thus spake Marc Lanctot:

I don't mean to sound like a broken record, but... I'd really like to
know why we wouldn't want to form a superset of GDL.


I'm not saying we don't. I'm not saying either way yet, as I haven't
had a chance to seriously study GDL yet.


Alright. Maybe we should start more like a reading group; each week we can discuss what we read, why it's good for our purpose, or why it's not, and then eventually we'll have the ground knowledge to know where to go from there. There isn't a huge literature out there, but there are some very relevant places to start. I'm not a GDL expert, but since it's the most-used and extensively studied general game language, we should probably have a good grasp on its strengths and weaknesses.

Any objections? This seemed to be the direction we were going anyway, so I hope not.

Before that, we should probably agree on what our goal is :)

I'll now follow-up on Gala.

  http://voidstar.boldlygoingnowhere.org/lanctot/tmp/koller97gala.pdf

Gala is a language for representing extensive-form games with imperfect information (the mathematical description of 'game'). It was designed by Daphne Koller et. al., the same people who designed a popular algorithm called "Sequence-Form Linear Programming" for finding Nash Equilibria in imperfect information, circa 1994. As far as I know, they designed Gala specifically because they released an implementation of their algorithm and wanted people to use it to 'solve' games. I don't know how much we can gain from this (I don't know how many people are using it, I don't know much about it.. ), but it's a start.

Marc

_______________________________________________
game-lang mailing list
game-lang@xxxxxxxxx
http://lists.ellipsis.cx/mailman/listinfo/game-lang