“…Many extensions and variations of classical bandit problems have been proposed, including: bandits with a varying finite or infinite numbers of arms (Whittle [78] and Banks and Sundaram [3]), bandits where an adversary has control over the payoffs (Auer et al [2]), bandits with dependent arms (Pandey, Chakrabarti, and Agarwal [57]), bandits where multiple arms can be chosen at the same time (Whittle [79]), bandits whose arms yield rewards even when they are inactive (Glazebrook, Kirkbride, and Ruiz-Hernandez [23]), and bandits with switching costs (Banks and Sundaram [4]). …”