Reinforcement Learning Code with PyTorch

Papers

• Deep Q-Network (DQN)
• Double DQN (DDQN)
• Advantage Actor-Critic (A2C)
• Asynchronous Advantage Actor-Critic (A3C)
• Deep Deterministic Policy Gradient (DDPG)
• Truncated Natural Policy Gradient (TNPG)
• Trust Region Policy Optimization (TRPO)
• Generalized Advantage Estimator (GAE)
• Proximal Policy Optimization (PPO)
• Soft Actor-Critic (SAC)
• Apprenticeship Learning via Inverse Reinforcement Learning (APP)
• Maximum Entropy Inverse Reinforcement Learning (MaxEnt)
• Generative Adversarial Imitation Learning (GAIL)
• Variational Adversarial Imitation Learning (VAIL)

Algorithms

01. Model-Free Reinforcement Learning

Deep Q-Network (DQN)

• CartPole(Classic control)

Double DQN (DDQN)

• CartPole(Classic control)

Advantage Actor-Critic (A2C)

• CartPole(Classic control)

Asynchronous Advantage Actor-Critic (A3C)

• CartPole(Classic control)

Deep Deterministic Policy Gradient (DDPG)

• Pendulum(Classic control)

Truncated Natural Policy Gradient (TNPG)

• Pendulum(Classic control)
• Hopper(MoJoCo)

Trust Region Policy Optimization (TRPO)

• Pendulum(Classic control)

TRPO + Generalized Advantage Estimator (GAE)

• Pendulum(Classic control)
• Hopper(MoJoCo)

Proximal Policy Optimization (PPO)

• Pendulum(Classic control)

PPO + Generalized Advantage Estimator (GAE)

• Pendulum(Classic control)
• Hopper(MoJoCo)

Soft Actor-Critic (SAC)

• Pendulum(Classic control)
• Hopper(MoJoCo)

---

02. Inverse Reinforcement Learning

Apprenticeship Learning via Inverse Reinforcement Learning (APP)

• MountainCar(Classic control)

Maximum Entropy Inverse Reinforcement Learning (MaxEnt)

• MountainCar(Classic control)

Generative Adversarial Imitation Learning (GAIL)

• Hopper(MoJoCo)

Variational Adversarial Imitation Learning (VAIL)

• Hopper(MoJoCo)

---

Learning curve

CartPole

Pendulum

Hopper

---

Reference

• Minimal and Clean Reinforcement Learning Examples in PyTorch
• Pytorch implementation for Policy Gradient algorithms (REINFORCE, NPG, TRPO, PPO)
• Implementation of APP
• Implementation of MaxEnt
• Pytorch implementation of GAIL
• Pytorch implementation of SAC1
• Pytorch implementation of SAC2