Policy Improvement via Imitation of Multiple Oracles

Despite its promise, reinforcement learning's real-world adoption has been hampered by the need for costly exploration to learn a good policy.