RAAM : The benefits of robustness in approximating aggregated MDPs in reinforcement learning. (2014)

by M Petrik, D Subramanian
Venue:In Neural Information Processing Systems,