/tensorflow-reinforce

Implementations of Reinforcement Learning Models in Tensorflow

Primary LanguagePythonMIT LicenseMIT

Tensorflow-Reinforce

A collection of Tensorflow implementations of reinforcement learning models. Models are evaluated in OpenAI Gym environments. Any contribution/feedback is more than welcome. Disclaimer: These implementations are used for educational purposes only (i.e., to learn deep RL myself). There is no guarantee that the exact models will work on any of your particular RL problems without changes.

Environments

This codebase works in both Python 2.7 and 3.5. The models are implemented in Tensorflow 1.0.

Models

Model Code References
Cross-Entropy Method run_cem_cartpole Cross-entropy method
Tabular Q Learning rl/tabular_q_learner Sutton and Barto, Chapter 8
Deep Q Network rl/neural_q_learner Mnih et al.
Double Deep Q Network rl/neural_q_learner van Hasselt et al.
REINFORCE Policy Gradient rl/pg_reinforce Sutton et al.
Actor-critic Policy Gradient rl/pg_actor_critic Minh et al.
Deep Deterministic Policy Gradient rl/pg_ddpg Lillicrap et al.

License

MIT