We consider a supervised learning problem in which data are revealed sequentially and the goal is to determine what will next be revealed. In the context of this problem, algorithms based on associa
