FortsAndMills / Learning-Reinforcement-Learning

Learning Reinforcement Learning:

Algorithms from DQN to Rainbow, PPO, DDPG with unified interface.

• Full text: Modern Deep Reinforcement Learning Algorithms
• Theory overview (slides): Theory Overview - some conspect of articles about implemented algorithms.
• LRL Demo: LRL demo.ipynb Library interface guide.
• LRL Cartpole: LRL Playing with cartpole.ipynb - launching a bunch of algorithms on the simplest environment Cartpole.

Requirements:

• gym
• PyTorch
• matplotlib
• pickle
• JSAnimation

Sources of code and inspiration:

• RL Adventure, RL Adventure pt.2 - beware of bugs!
• DeepRL Tutorials
• and more

Many more algorithms to discuss:

IQN, DQRN, ACKTR, ACER, SAC, TD3, REACTOR, APE-X, R2D2...