The theories of offline and online reinforcement learning, despite having
evolved in parallel, have begun to show signs of the possibility for a
unification, with algorithms and analysis techniques for one setting often
having natural counterparts in the other. However, the notion of d