machin 0.1.0

Reinforcement learning library

--- [](https://machin.readthedocs.io/en/latest/)

Machin is a reinforcement library purely based on pytorch. It is designed to be readable, reusable and extendable.

Supported algorithms?

---

Currently Machin has implemented the following algorithms, the list is still growing:

Single agent algorithms:

• Deep Q-Network (DQN)
• Double DQN
• Dueling DQN
• RAINBOW
• Deep Deterministic policy Gradient (DDPG),
• Twin Delayed DDPG (TD3)
• Hystereric DDPG (Modified from Hys-DQN)
• Advantage Actor-Critic (A2C)
• Proximal Policy Optimization (PPO)
• Soft Actor Critic (SAC)

Multi-agent algorithms:

• Multi-agent DDPG (MADDPG)

Massively parallel algorithms:

• Asynchronous A2C (A3C)
• APEX-DQN, APEX-DDPG
• IMPALA

Enhancements:

• Prioritized Experience Replay (PER)
• Generalized Advantage Estimation (GAE)
• Recurrent networks in DQN, etc.

Algorithms to be supported:

• Distributed DDPG (D4PG)
• Generative Adversarial Imitation Learning (GAIL)
• Evolution Strategies
• QMIX (multi agent)
• Model-based methods

TODO: write examples, test and debug TODO: add clip grad norm for DDPG, DQN etc. TODO: add update interval for IMPALA, APEX, A3C TODO: integrate with NNI TODO: add more network structure implementations