A central location for my reinforcement learning experiments
For minimalist, concept-based code:
There are several notebooks here. Some of them are under active development.
First I solved Cartpole as a toy problem (I know that pg is overkill)
Then I used the same algorithm to solve something trickier - Atari Pong!
Code for both of these was derived in part from karpathy's excellent deep rl post.
- Mostly Python 2.7. You will need:
- Numpy
- Matplotlib
- TensorFlow
- OpenAI Gym