News for hackers. Stuff that matters.

A comment on Mastering the Game of Stratego with Model-Free Multiagent Reinforcement Learning. Parent comment

Re:Mastering the Game of Stratego with Model-Free Multiagent... by joe_the_user on Monday July 11, 2022 @07:16PM (#32063419)
POMDPs? Are those considered to be much simpler problems?

Well, solving a Partially Observed Markov Decision Processes in general isn't just NP-complete but actually undecidable. So I'm not sure how one measures SOTA (state of the art).

[ Parent | on HN ]
Re:Mastering the Game of Stratego with Model-Free Multiagent... by thomasahle on Monday July 11, 2022 @07:33PM (#32063574)
> I'm not sure how one measures SOTA (state of the art).

You measure it by practical performance on a test suite of important POMDPs and games.

[ Parent | on HN ]