Mushroom - Reinforcement Learning Python library

[Carlo D'Eramo]

Hi everyone,

few months ago I posted here to let you know about a RL Python library, called Mushroom, I'm working on for my PhD. It has the purpose to simplify the developing of RL experiments and it is highly customizable to let user add and test their own algorithms. It uses Gym and Pytorch (one can also use Tensorflow or other libraries transparently) to exploit already developed powerful libraries and just to focus on RL aspects. Recently, we added several policy gradient and actor-critic algorithms, such as the recent DPG. I'd really appreciate if you give a look to Mushroom and consider it for doing your research, or simply to give me a feedback on it. Here are the GitHub link: https://github.com/AIRLab-POLIMI/mushroom and the doc link: https://mushroomrl.readthedocs.io/en/latest/.

Cheers!

[BartoszKP]

Hi Carlo,

  thanks for sharing this. Just a side note from my side: For about 10 years I'm observing many researchers creating platforms/libraries with "the purpose to simplify the developing of RL experiments and it is highly customizable to let user add and test their own algorithms.", and I've been guilty of that myself :-)

So if you really want to advertise your work it would be interesting to learn what are the key advantages over other frameworks/libraries - and there are a lot of them! Also, most of the time for other researchers it is still more convenient to start something for their own, then learn someone else's code :-)

But of course even if that's the case, it's good that we share our work, and you never know who might benefit from your work, so please don't read my comment as a discouragement:-)

Cheers!