Thread: [Toss-devel] GGP course at Coursera.
Status: Beta
Brought to you by:
lukaszkaiser
Menu
▾
▴
toss-devel
|
From: Lukasz S. <luk...@gm...> - 2013-10-06 14:34:24
|
Hi, I've missed the start of the course: https://www.coursera.org/course/ggp by Michael Genesereth (lectures), Michael Thielscher (help) and Sam Schreiber (programming). Here is the textbook for the course: http://arrogant.stanford.edu/ggp/chapters/cover.html I may enlist Toss as a player in the course, but I probably won't do my best toward Toss winning... Regards. |
|
From: Lukasz K. <luk...@gm...> - 2013-10-07 02:08:46
|
Hi. > I've missed the start of the course: > https://www.coursera.org/course/ggp > I may enlist Toss as a player in the course, but I probably > won't do my best toward Toss winning... I'm sure it is not too late to make up for stuff in the course :). As for Toss, there is one thing I'm recently interested in. I added a way to generate QBF formulas saying "player 1 can win in N steps" for "easy" games (2 locations only for now, and ?P not allowed in rules). This works quite ok e.g. for Connect-4. And in SatSolver ml we now have an interface to depqbf, a very good qbf solver. So e.g. for Connect4 it can check quite efficiently whether player 1 wins in a few steps (solving the whole game, i.e. 7*6 steps, is more difficult of course). But it seems that knowing "I can win in 4 steps" should be good for a player - esp. if we could combine it with a heuristic in case of not-yet-winning. And this should be quite easy in Toss now. So if you ever get to making a player again, you can consider adding such a check :). My best! Lukasz |
Thanks for helping keep SourceForge clean.
X