Efficient Safe Meta-Reinforcement Learning: Provable Near-Optimality and Anytime Safety

Open in new window