In this folder, you will find 5 repositories:
- statrl: This defines several Reinforcement Learning settings, including environments, learing agents and experimenation helpers. This repository replaces and aggregates environments, learners, experiments older repositories.
- environments (Deprecated, Python 3.11): This defines several Reinforcement Learning environments, especially discrete (tabular) Markov Decision Processes and Multi-armed Bandits.
- learners (Deprecated, Python 3.11): This defines several Reinforcement Learning learning agents, including classical UCB algorithm for bandits or UCRL2 for Markov Decision Processes.
- experiments (Deprecated, Python 3.11): This defines useful tools to run and compare regret of multiple algorithms in the same environment, producing regret plots and log files directly usable in a research article.
- articles: A list of articles using this library, with companion code to fully reproduce experiments from these articles (using the version of the lib available at that time)
Make sure to use Python 3.14
pip install statrl