What methods is everybody using?

[Thomas Kelly]

Hi what algorithms and libraries you guys using?

Im currently working my way through this article: https://deepmind.com/blog/article/Agent57-Outperforming-the-human-Atari-benchmark

and started from dqn -> ddqn -> duelingddqn with prioritised replay in pytorch.

Am at the moment trying to create some form of distributed learning setup (3 actors and 1 learner across my 4 processors) (from the Ape-x paper).

Haven't put much of time into policy gradient methods yet but will get there eventually. 

Oh and im also using 'simple115' observation with just linear layers as i dont have the computational muscle on my laptop to use conv layers.

So yeah what is everyone else's approach?

Tom