This project contains a simple tensorflow implementation of recurrent policy gradients [1]. For an alternative presentation of recurrent policy gradients, see [2].
This implementation is not particularly clear, efficient, well tested, or numerically stable. We advise against using this software for nondidactic purposes.
This software is licensed under the MIT License.
See the examples directory. Remember to add the (outermost) rpg folder to the PYTHONPATH.
[1] Wierstra, D., Förster, A., Peters, J., Schmidhuber, J. Recurrent policy gradients. Logic Journal of IGPL, 2010.
[2] Rauber, P. Notes on Reinforcement Learning. Available at paulorauber.com.