google-deepmind/dqn_zoo

DQN Zoo is a collection of reference implementations of reinforcement learning agents developed at DeepMind based on the Deep Q-Network (DQN) agent.

What it solves

DQN Zoo provides a set of reference implementations for various Deep Q-Network (DQN) agents. It solves the problem of reproducing research results from DeepMind's DQN-based reinforcement learning papers, offering a readable and self-contained codebase for researchers to fork and customize.

How it works

The project is implemented using JAX, Haiku, and RLax. It contains separate directories for different DQN variants (such as Double DQN, Prioritized Experience Replay, C51, QR-DQN, Rainbow, and IQN), each replicating the algorithmic details of their respective papers. The agents are configured to run on the standard set of 57 Atari games using the OpenAI Gym interface.

Who it’s for

It is designed for reinforcement learning researchers who need reliable reference implementations of DQN variants to use as a baseline or a starting point for their own research.

Highlights

  • Reference Implementations: Replicates results on 57 Atari games for multiple DQN variants including Rainbow and IQN.
  • Research-Friendly: Prioritizes readability and minimal dependencies over general-purpose library flexibility.
  • JAX-based: Built with JAX, Haiku, and RLax for efficient computation.
  • Self-Contained: Includes a Docker-based quick start to handle dependencies and Atari ROMs.

Related

  • Dispatch
    OpenAI Baselines: DQN release and reinforcement learning best practices

    OpenAI has open-sourced OpenAI Baselines, starting with a high-performance implementation of DQN and three of its variants to improve reproducibility in reinforcement learning.

  • Dispatch
    Deep Q-Learning with Space Invaders

    Hugging Face explains Deep Q-Learning (DQN), a reinforcement learning method that uses neural networks to approximate Q-values for high-dimensional state spaces like Atari games.

  • Project
    rlcode/reinforcement-learning

    A collection of easy-to-read, standalone PyTorch implementations of core reinforcement learning algorithms, from basic Grid World to deep RL for Atari games.

  • Project
    instadeepai/jumanji

    A diverse suite of scalable, hardware-accelerated reinforcement learning environments written in JAX for high-speed RL research and experimentation.

  • Project
    Farama-Foundation/PettingZoo

    A Python library for multi-agent reinforcement learning research that provides a standardized API and a diverse set of environments, similar to a multi-agent version of Gymnasium.