Simple implementations of RL algorithms in tensorflow or numpy. Note: Used numpy for ARS because it was quicker. Still some clean up work to do. Also additionally, there is a linear regression model, which I found useful to build an understanding of tensorflow.
The requirements.txt states tensorflow not tensorflow-gpu. So if you want to use tensorflow-gpu please install that.
- Document everything
- Normalisation for ARS.
- Baseline for REINFORCE.
- DQN variants.
- PPO.