Publication
NeurIPS 2009
Conference paper
Online submodular minimization
Abstract
We consider an online decision problem over a discrete space in which the loss function is submodular. We give algorithms which are computationally efficient and are Hannan-consistent in both the full information and bandit settings.