Rl algorithms

Below are the implemented algorithms and their brief descriptions.

• [x] Deep Q-Learning (DQN)
  • dqn.py
    • For discrete action space.
  • dqn_atari.py
    • For playing Atari games. It uses convolutional layers and common atari-based pre-processing techniques.
• [x] Categorical DQN (C51)
  • c51.py
    • For discrete action space.
  • c51_atari.py
    • For playing Atari games. It uses convolutional layers and common atari-based pre-processing techniques.
  • c51_atari_visual.py
    • Adds return and q-values visulization for dqn_atari.py.
• [x] Proximal Policy Gradient (PPO)
  • All of the PPO implementations below are augmented with some code-level optimizations. See https://costa.sh/blog-the-32-implementation-details-of-ppo.html for more details
  • ppo.py
    • For discrete action space.
  • ppo_continuous_action.py
    • For continuous action space. Also implemented Mujoco-specific code-level optimizations
  • ppo_atari.py
    • For playing Atari games. It uses convolutional layers and common atari-based pre-processing techniques.
• [x] Soft Actor Critic (SAC)
  • sac_continuous_action.py
    • For continuous action space.
• [x] Deep Deterministic Policy Gradient (DDPG)
  • ddpg_continuous_action.py
    • For continuous action space.
• [x] Twin Delayed Deep Deterministic Policy Gradient (TD3)
  • td3_continuous_action.py
    • For continuous action space.
• [x] Phasic Policy Gradient (PPG)
  • ppg_procgen.py
    • PPG implementation for Procgen