ReinforcementLearning.jl

https://github.com/JuliaReinforcementLearning/ReinforcementLearning.jl [ReinforcementLearning.jl] — this is a package for reinforcement learning on Julia.

Its device is based on the following principles.

Reusability and extensibility: Well-thought-out components and interfaces help users implement new algorithms.
Ease of experimentation: New users can easily conduct experiments to evaluate performance, compare different algorithms, and evaluate and diagnose agents.
Reproducibility: From traditional tabular methods to modern reinforcement learning algorithms.

, Getting started

julia> ] add ReinforcementLearning

julia> using ReinforcementLearning

julia> run(
           RandomPolicy(),
           CartPoleEnv(),
           StopAfterNSteps(1_000),
           TotalRewardPerEpisode()
       )

The above example demonstrates the four main components of a standard reinforcement learning experiment:

Politics. RandomPolicy is the simplest instance of the type https://juliareinforcementlearning.org/docs/rlbase/#ReinforcementLearningBase .AbstractPolicy[AbstractPolicy]. It creates a random action at each step.
Wednesday. CartPoleEnv — this is a typical object of the type https://juliareinforcementlearning.org/docs/rlbase/#ReinforcementLearningBase .AbstractEnv[AbstractEnv] for testing reinforcement learning algorithms.
The final condition. StopAfterNSteps(1_000) notifies that the experiment should end after 1_000 steps.
Handler. Structure TotalRewardPerEpisode — one of the most common subtypes AbstractHook. It serves to collect the total reward for each episode of the experiment.

In https://juliareinforcementlearning.org/docs/tutorial /[guide] you can learn how these four components together allow you to solve many interesting problems. We also periodically share in our https://juliareinforcementlearning.org/blog /[blog post] features of the implementation of some algorithms. We especially recommend that you familiarize yourself with the introduction to the ReinforcementLearning.jl, which explains how the package was designed.

✋ Getting help

Need help with ReinforcementLearning.jl? Here’s how to get it:

Read the documentation on the Internet! Most likely, the answer is already in the example or in the API documentation. Perform the search using the field in the upper left corner.

Contact us at https://julialang.org/slack /[Julia Slack] in the #reinforcement-learnin channel.
Ask a question on https://discourse.julialang.org /[Julia discussion forum] in the Machine Learning category by adding the reinforcement-learning tag.
If the ReinforcementLearning package.jl does not work as expected, or you have identified inaccuracies in it, create a problem on https://github.com/JuliaReinforcementLearning/ReinforcementLearning .jl[his GitHub page], providing a minimal working example and actions to reproduce the problem.

, Quoting

If you are using the 'ReinforcementLearning' package.jl` in a scientific publication, then you can refer to it in CITATION.bib.