This is a preview. Log in through your library . Abstract For an arbitrary convex loss function (including the case of mean absolute deviation), a simple and direct proof of the superiority of ...