Reinforcement Learning Package



I started to write a julia package for reinforcement learning:

Currently it contains a few algorithms for tabular environments and, on a separate branch, linear function approximation for a few TD algorithms. It is easy to use it with POMDPModels.jl of the POMDPs.jl universe, see examples. Whenever I have time I will continue to add support for linear function approximation, which is straightforward, and try to extend it with non-linear function approximation, e.g. using Knet.jl.

Any feedback is highly appreciated.
Please let me know, if you want to collaborate.


The package has grown quite a bit in the meantime and features now linear function approximation and deep reinforcement learning (currently (double)DQN and deep actor critic policy gradient) thanks to Flux. Included are classic environments like Mountain Car and CartPole written in Julia (way faster than using OpenAI Gym through PyCall), Atari Games and MDPs/POMDPs.

I moved the package to