We present a randomized primal-dual algorithm that solves the problem to additive error in time , for matrix with larger dimension and nonzero entries. This improves the best known exact gradient methods by a factor of and is faster than fully stochastic gradient methods in the accurate and/or sparse regime . Our results hold for in the simplex (matrix games, linear programming) and for in an ball and in the simplex (perceptron / SVM, minimum enclosing ball). Our algorithm combines Nemirovski's "conceptual prox-method" and a novel reduced-variance gradient estimator based on "sampling from the difference" between the current iterate and a reference point.
View on arXiv