Does anyone else here really nerd out about minimax/α-β/MCTS and generally AI for two-player games of perfect information?
-
-
Replying to @nelhage
when you say it like that, I feel like I can't possibly meet your standards of nerding out about it.
1 reply 0 retweets 2 likes -
Replying to @avibryant @nelhage
but if you're starting a paper reading group count me in.
2 replies 0 retweets 0 likes -
Replying to @avibryant
.
@avibryant I “accidentally” wrote the strongest player for http://cheapass.com/tak/ and periodically find myself in need of a rubber duck :)1 reply 0 retweets 5 likes -
Replying to @nelhage
I mean, I'm also down for nerding out about
@PatrickRothfuss, so that works.1 reply 0 retweets 2 likes -
Replying to @avibryant
Any papers offhand that talk about how one tunes an MCTS? Feels like the parameter space for implementors is substantial.
1 reply 0 retweets 0 likes -
Replying to @nelhage
not that I'm aware of beyond "use UCB" which afaik everyone does anyway.
1 reply 0 retweets 0 likes -
Replying to @avibryant
UCT seems to be the only game in town. But e.g. how do I tune a rollout policy? How do I handle nodes with no data yet? etc.
2 replies 0 retweets 0 likes -
Replying to @nelhage
same as any hyperparameter search I guess? Grid search, or bayesian things (link to follow).
1 reply 0 retweets 0 likes
this is the specific paper I was thinking of: https://arxiv.org/pdf/1206.2944.pdf …
-
-
Replying to @avibryant
this project will probably inevitably result in me learning some ML, which I really should have done by now anyway...
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.