Contrastive Example-Based Control

Kyle Hatch, Benjamin Eysenbach, Rafael Rafailov +4

While many real-world problems that might benefit from reinforcement learning, these problems rarely fit into the MDP mold: interacting with the environment is often expensive and…