Deep Reinforcement Learning

Implementing and showcasing reinforcement learning algorithms with deep networks. Following Open AI spinning up suggestions.

Contents

Deep Q learning and Improvements

1. Deep Q learning
2. Deep Recurrent Q learning
3. Double Deep Q learning
4. Dueling Deep Q learning
5. Dueling Double Deep Q learning
6. Prioritized Experience Replay

Policy Gradient Methods

1. Vanilla Policy Gradient [Reinforce]
2. Value Actor Critic
3. Advantage Actor Crictic
4. Proximal Policy Optimization
5. GAE
6. Soft Actor Critic

Status

All implementations upto Advantage actor critic is done. Working on updating the comments.

Todo

1. Add texts explaining the inner workings of the codes.
2. Solve other environments like cartpole and atari 2600.
3. Use CNNs.
4. Actor critics are barebone. Will need to implement improvement methods for performance.