We consider the problem of adversarial (nonstochastic) online learning with partial-information feedback, in which, at each round, a decision maker selects an action from a finite set of alternatives. We develop a black-box approach for such problems in ...