RL similarity metrics work without discounting
A new game-theoretic framework extends a key reinforcement learning tool to settings where future rewards are not discounted.
A new game-theoretic framework extends a key reinforcement learning tool to settings where future rewards are not discounted.