Near-optimal Regret Using Policy Optimization in Online MDPs with Aggregate Bandit Feedback.

Tal Lancewicki, Yishay Mansour

Browse the full ICML paper archive.