LowerBound

Then, we consider sufficient assumptions under which learning good policies requires polynomial number of episodes.