Implements of Reinforcement Learning Algorithms

This repo is implements of Reinforcement Learning Algorithms, implementing as learning, some of them are even another version of some tutorial. Any contributions are welcomed.

Content

• Deep Deterministic Policy Gradient (DDPG)
  Implement of DDPG.

  arXiv:1509.02971: Continuous control with deep reinforcement learning

• Asynchronous Advantage Actor-Critic Model (A3C)
  Implement of A3C.

  arXiv:1602.01783: Asynchronous Methods for Deep Reinforcement Learning

• Double-DQN
  Implement of Double-DQN.

  arXiv:1509.06461: Deep Reinforcement Learning with Double Q-learning

• Dueling-DQN
  Implement of Dueling-DQN.

  arXiv:1511.06581: Dueling Network Architectures for Deep Reinforcement Learning

• Deep Q-Network (DQN)
  Implement of DQN.

  arXiv:1312.5602: Playing Atari with Deep Reinforcement Learning

• Actor-Critic Model
  Implement of Actor-Critic Model.

  arXiv:1607.07086: An Actor-Critic Algorithm for Sequence Prediction

• Policy Gradient (PG)
  Implement of Policy Gradient.

  NIPS. Vol. 99. 1999: Policy gradient methods for reinforcement learning with function approximation

• Q-Learning
  Implements of Q-Learning.

  Paper: Convergence of Q-learning: a simple proof

• Sarsa
  Implement of Sarsa.

  Paper: Online Q-Learning using Connectionist Systems

Requirements

• Python3.5
• TensorFlow1.4
• gym
• numpy
• matplotlib
• pandas (option)

How to Run

All algorithms are implemented with TensorFlow, the default environment are games provided by gym. You can just clone this project, and run the each algorithm by:

python3.5 algorithms/algo_name.py

TODO

• More implements of Deep Reinforcement Learning Paper and Methods.