A Comparison of learning algorithms on the Arcade Learning Environment
This provides a benchmark for comparing RL algorithms on complex tasks, though it is incremental as it applies existing methods to new data.
The paper compared model-free linear reinforcement learning algorithms on the Arcade Learning Environment, finding that this framework enables evaluation on diverse, difficult problems beyond traditional toy environments.
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 possible to evaluate algorithms on diverse and difficult problems within a consistent framework. We discuss some challenges posed by the arcade learning environment which do not manifest in simpler environments. We then provide a comparison of model-free, linear learning algorithms on this challenging problem set.