RLkit: A simple Reinforcement Learning library

This project is still a work in progress. More algorithms and detailed documentation coming soon :)

Currently supported agents-

1. Random agent
2. REINFORCE (Policy Gradients)
3. DQN
4. DQN with baseline
5. Actor-Critic

See examples for details on how to use the library.

Installation:

Stable:

$ pip install -U RLkit

Dev:

To get the project's source code, clone the github repository:

$ git clone https://github.com/shubhamjha97/rlkit.git
$ cd RLkit

Install VirtualEnv using the following (optional):

$ [sudo] pip install virtualenv

Create and activate your virtual environment (optional):

$ virtualenv venv
$ source venv/bin/activate

Install all the required packages:

$ pip install -r requirements.txt

Install the package by running the following command from the root directory of the repository:

$ python setup.py install	

Requirements-

tensorflow==1.11.0
gym==0.10.8
numpy==1.15.4

New in v0.2

• Added DQN and DQN with baseline agents
• Added ActorCritic agent
• Added support for various activation functions

Upcoming

• Duelling DQN
• Support for logging and plotting
• Support for adding seeds
• Support for custom environments

Compatibility

The package has been tested with python 3.5.2

References

• Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D., & Riedmiller, M. (2013). Playing atari with deep reinforcement learning. arXiv preprint arXiv:1312.5602. [paper]

• Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness, J., Bellemare, M. G., ... & Petersen, S. (2015). Human-level control through deep reinforcement learning. Nature, 518(7540), 529. [paper]

• Williams, R. J. (1992). Simple statistical gradient-following algorithms for connectionist reinforcement learning. Machine learning, 8(3-4), 229-256. [paper]

• Konda, V. R., & Tsitsiklis, J. N. (2000). Actor-critic algorithms. In Advances in neural information processing systems (pp. 1008-1014). [paper]