Bandits with many optimal arms

We consider a stochastic bandit problem with a possibly infinite number of arms. We write for the proportion of optimal arms and for the minimal mean-gap between optimal and sub-optimal arms. We characterize the optimal learning rates both in the cumulative regret setting, and in the best-arm identification setting in terms of the problem parameters (the budget), and . For the objective of minimizing the cumulative regret, we provide a lower bound of order and a UCB-style algorithm with matching upper bound up to a factor of . Our algorithm needs to calibrate its parameters, and we prove that this knowledge is necessary, since adapting to in this setting is impossible. For best-arm identification we also provide a lower bound of order on the probability of outputting a sub-optimal arm where is an absolute constant. We also provide an elimination algorithm with an upper bound matching the lower bound up to a factor of order in the exponential, and that does not need or as parameter.
View on arXiv