A Comparison of learning algorithms on the Arcade Learning Environment

Aaron Defazio, Thore Graepel

Reinforcement learning agents have traditionally been evaluated on small toy problems. With advances in computing power and the advent of the Arcade Learning Environment, it is now…