In multi-task learning and meta-learning, the goal is not just to learn one skill, but to learn a number of skills. In multi-task RL, we assume that we want to learn a fixed set of skills with minimal data, while in meta-RL, we want to use experience..
Relevance: 7.5465975