Modern Implementation of PGPortfolio

[CaioSBC]

Hello everyone!

Since this repository is no longer maintained, I have been working on reproducing its implementations by creating a new Python package that utilizes modern versions of Gymnasium and PyTorch. Initially, I was using this code for my personal research, but I realized it could be beneficial to make it public. By doing so, the community can contribute, review, and enhance its reliability.

So I would like to invite all of you to collaborate on RLPortfolio so that we can develop a library that facilitates portfolio optimization research using reinforcement learning and promotes reproducibility in the field.

The library is still under development, with several features yet to be implemented. However, the training algorithm, market simulation, and a few deep learning architectures are already available for use.

[MuslemRahimi]

I like the idea. I am working fulltime on my project Stocknear and I am planning to apply more AI feature to my platform which your repo would be a good match. Let me know if you are interested!

[CaioSBC]

Just like PGPortfolio, RLPortfolio can be applied to any kind of financial portfolio whose assets are able to be bought and sold freely, all you need is the historical price data. Just keep in mind that the formulation considers that there is no slippage and the agent does not affect the market, so it is advisable to apply it in a portfolio that contains high-volume assets.

So, yes, I do believe you can use the repo.