The True Sample Complexity of Identifying Good Arms

We consider two multi-armed bandit problems with arms: (i) given an , identify an arm with mean that is within of the largest mean and (ii) given a threshold and integer , identify arms with means larger than . Existing lower bounds and algorithms for the PAC framework suggest that both of these problems require samples. However, we argue that these definitions not only conflict with how these algorithms are used in practice, but also that these results disagree with intuition that says (i) requires only samples where and (ii) requires samples where . We provide definitions that formalize these intuitions, obtain lower bounds that match the above sample complexities, and develop explicit, practical algorithms that achieve nearly matching upper bounds.
View on arXiv