Reinforcement Learning and Bandit Algorithms

Algorithms that learn by trial and error from rewards, such as multi-armed bandit methods, best arm identification and KL-UCB. An algorithm belongs here when its cited article describes learning from rewards, an agent choosing actions or a bandit problem. It excludes supervised and unsupervised learning methods. A member is placed here by what its own cited article states about it, not by the name of the algorithm or by the uses it is put to today, so each entry sits in the group its source describes it as belonging to.