MuZero is a computer program developed by artificial intelligence research company DeepMind to master games without knowing their rules. Its release in 2019 included benchmarks of its performance in go, chess, shogi, and a standard suite of Atari games. The algorithm uses an approach similar to AlphaZero. It matched AlphaZero's performance in chess and shogi, improved on its performance in Go (setting a new world record), and improved on the state of the art in mastering a suite of 57 Atari games (the Arcade Learning Environment), a visually-complex domain.
MuZero was trained via self-play and play against AlphaZero, with no access to rules, opening books, or endgame tables. The trained algorithm used the same convolutional and residual algorithms as AlphaZero, but with 20% fewer computation steps per node in the search tree.
MuZero really is discovering for itself how to build a model and understand it just from first principles.— David Silver, DeepMind, Wired
On November 19, 2019, the DeepMind team released a preprint introducing MuZero.
Derivation from AlphaZero
MuZero (MZ) is a combination of the high-performance planning of the AlphaZero (AZ) algorithm with approaches to model-free reinforcement learning. The combination allows for more efficient training in classical planning regimes, such as Go, while also handling domains with much more complex inputs at each stage, such as visual video games.
- AZ's planning process uses a simulator. The simulator knows the rules of the game. It has to be explicitly programmed. A neural network then predicts the policy and value of a future position. Perfect knowledge of game rules is used in modeling state transitions in the search tree, actions available at each node, and termination of a branch of the tree. MZ does not have access to the rules, and instead learns one with neural networks.
- AZ has a single model for the game (from board state to predictions); MZ has separate models for representation of the current state (from board state into its internal embedding), dynamics of states (how actions change representations of board states), and prediction of policy and value of a future position (given a state's representation).
- MZ's hidden model may be complex, and it may turn out it can host computation; exploring the details of the hidden model in a trained instance of MZ is a topic for future exploration.
- MZ does not expect a two-player game where winners take all. It works with standard reinforcement-learning scenarios, including single-agent environments with continuous intermediate rewards, possibly of arbitrary magnitude and with time discounting. AZ was designed for two-player games that could be won, drawn, or lost.
Comparison with R2D2
The previous state of the art technique for learning to play the suite of Atari games was R2D2, the Recurrent Replay Distributed DQN.
MuZero surpassed both R2D2's mean and median performance across the suite of games, though it did not do better in every game.
Training and results
MuZero used 16 third-generation tensor processing units (TPUs) for training, and 1000 TPUs for selfplay (for board games, with 800 simulations per step) and 8 TPUs for training and 32 TPUs for selfplay (for Atari games, with 50 simulations per step).
AlphaZero used 64 first-generation TPUs for training, and 5000 second-generation TPUs for selfplay. As TPU design has improved (third-generation chips are 2x as powerful individually as second-generation chips, with further advances in bandwidth and networking across chips in a pod), these are comparable training setups.
R2D2 was trained for 5 days through 2M training steps.
MuZero matched AlphaZero's performance in chess and Shogi after roughly 1 million training steps. It matched AZ's performance in Go after 500 thousand training steps, and surpassed it by 1 million steps. It matched R2D2's mean and median performance across the Atari game suite after 500 thousand training steps, and surpassed it by 1 million steps; though it never performed well on 6 games in the suite.
MuZero was viewed as a significant advancement over AlphaZero, and a generalizable step forward in unsupervised learning techniques. The work was seen as advancing understanding of how to compose systems from smaller components, a systems-level development more than a pure machine-learning development.
While only pseudocode was released by the development team, Werner Duvaud produced an open source implementation based on that.
MuZero has been used as a reference implementation in other work, for instance as a way to generate model-based behavior.
- Wiggers, Kyle. "DeepMind's MuZero teaches itself how to win at Atari, chess, shogi, and Go". VentureBeat. Retrieved 22 July 2020.
- Friedel, Frederic. "MuZero figures out chess, rules and all". ChessBase GmbH. Retrieved 22 July 2020.
- Rodriguez, Jesus. "DeepMind Unveils MuZero, a New Agent that Mastered Chess, Shogi, Atari and Go Without Knowing the Rules". KDnuggets. Retrieved 22 July 2020.
- Schrittwieser, Julian; Antonoglou, Ioannis; Hubert, Thomas; Simonyan, Karen; Sifre, Laurent; Schmitt, Simon; Guez, Arthur; Lockhart, Edward; Hassabis, Demis; Graepel, Thore; Lillicrap, Timothy (2020). "Mastering Atari, Go, chess and shogi by planning with a learned model". Nature. 588 (7839): 604–609. arXiv:1911.08265. Bibcode:2020Natur.588..604S. doi:10.1038/s41586-020-03051-4. PMID 33361790. S2CID 208158225.
- "What AlphaGo Can Teach Us About How People Learn". Wired. ISSN 1059-1028. Retrieved 2020-12-25.
- Silver, David; Hubert, Thomas; Schrittwieser, Julian; Antonoglou, Ioannis; Lai, Matthew; Guez, Arthur; Lanctot, Marc; Sifre, Laurent; Kumaran, Dharshan; Graepel, Thore; Lillicrap, Timothy; Simonyan, Karen; Hassabis, Demis (5 December 2017). "Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm". arXiv:1712.01815 [cs.AI].
- Kapturowski, Steven; Ostrovski, Georg; Quan, John; Munos, Remi; Dabney, Will. RECURRENT EXPERIENCE REPLAY IN DISTRIBUTED REINFORCEMENT LEARNING. ICLR 2019 – via Open Review.
- Shorten, Connor (2020-01-18). "The Evolution of AlphaGo to MuZero". Medium. Retrieved 2020-06-07.
- "[AN #75]: Solving Atari and Go with learned game models, and thoughts from a MIRI employee - LessWrong 2.0". www.lesswrong.com. Retrieved 2020-06-07.
- Wu, Jun. "Reinforcement Learning, Deep Learning's Partner". Forbes. Retrieved 2020-07-15.
- "Machine Learning & Robotics: My (biased) 2019 State of the Field". cachestocaches.com. Retrieved 2020-07-15.
- Duvaud, Werner (2020-07-15), werner-duvaud/muzero-general, retrieved 2020-07-15
- van Seijen, Harm; Nekoei, Hadi; Racah, Evan; Chandar, Sarath (2020-07-06). "The LoCA Regret: A Consistent Metric to Evaluate Model-Based Behavior in Reinforcement Learning". arXiv:2007.03158 [cs.stat].