Langevin DQN

Vikranth Dwaracherla, Benjamin Van Roy

Algorithms that tackle deep exploration -- an important challenge in reinforcement learning -- have relied on epistemic uncertainty representation through ensembles or other hyperm…