Mastering Atari, Go, chess and shogi by planning with a learned model

https://doi.org/10.1038/s41586-020-03051-4

MuZero: learned representation, dynamics, and prediction networks planned over with MCTS, without being given the rules; preprint November 2019.

Self-play planning with learned models (AlphaZero, MuZero, Dreamer)