Model-Based Reinforcement Learning
Model-based reinforcement learning improves sample efficiency by learning a dynamics model of the environment (World Model) and generating experience through "imagination." From the classic Dyna architecture to modern Dreamer and MuZero, MBRL has demonstrated powerful capabilities across tasks ranging from board games to pixel-level control.
Contents:
- Model-Based RL Overview — Model-Free vs. Model-Based, environment model learning, Dyna architecture, MBPO, Dreamer series, MuZero