Online mirror descent (OMD) uses the Bregman divergence as the distance measure, and for convex Lipschitz-continuous loss sequences it guarantees an upper bound on cumulative regret of O(√T) or O(log T).