faust-wang/deepreinforcementlearning-pytorch

PyTorch implementation of classical deep reinforcement learning algorithms

DeepReinforcementLearning-PyTorch

PyTorch implementation of classical deep reinforcement learning algorithms

Implemented Core Algorithms:

• Vanilla Deep Q-Learning (DQN)

  • Human Level Control Through Deep Reinforement Learning [Publication]

• Double Deep Q-Learning (Double DQN)

  • Deep Reinforcement Learning with Double Q-learning [Publication]

• Dueling Deep Q-Learning (Dueling DQN)

  • Dueling Network Architectures for Deep Reinforcement Learning [Publication]

• Hindsight Experience Replay

  • Hindsight Experience Replay [Publication]

• Prioritized Experience Replay [Publication]

• Synchronous Deep Q-Learning (SDQN)

• REINFORCE

• Deep Deterministic Policy Gradient (DDPG)

  • Continuous control with deep reinforcement learning [Publication]

• Asynchronous/Synchronous Advantage Actor Critic (A3C, A2C) [Publication]

• TD3 [Publication]

• Soft Actor Critic (SAD) [Publication]

• Stacked DQN/DDPG/SAC

Enhancements:

• Hindsight Experience Replay [Publication]
• Prioritized Experience Replay [Publication]
• Noisy Networks for Exploration [Publication]

Application examples:

• Efficient Navigation of Active Particles in an Unseen Environment via Deep Reinforcement Learning [Publication]

Custom envs

• 1D stablizer, 2D stabilizer, and multi-Dim stabilizer
• maze with static obstacles and stchastic/deterministic agent
• maze with dynamic obstacles and stchastic/deterministic agent
• finanical portfolio engineering env (for hedging and investment)
• colloidal assembly env

Third party envs

• [pybullet]

Cited as

@misc{Yang2019, author = {Yuguang Yang}, title = {DRL-Pytorch}, year = {2019}, publisher = {GitHub}, journal = {GitHub repository}, howpublished = {\url{https://github.com/yangyutu/DeepReinforcementLearning-PyTorch}} }