Joint Online Learning and Decision-making via Dual Mirror Descent

                     Alfonso Lobos 1 Paul Grigas 1 Zheng Wen 2

              Abstract                ensuring that we do not incur too many costs too early, (ii)
    We consider an online revenue maximization         ensuring that enough costs are incurred to meet the lower
    problem over a finite time horizon subject to lower     bound constraints over the full time horizon, and (iii) under-
    and upper bounds on cost. At e ...