On Primal and Dual Approaches for Distributed Stochastic Convex Optimization over Networks
We introduce primal and dual stochastic gradient oracle methods for distributed convex optimization problems over networks. We show that the proposed methods are optimal (in terms of communication steps) for primal and dual oracles. Additionally, for a dual stochastic oracle, we propose a new analysis method for the rate of convergence in terms of duality gap and probability of large deviations. This analysis is based on a new technique that allows to bound the distance between the iteration sequence and the optimal point. By the proper choice of batch size, we can guarantee that this distance equals (up to a constant) to the distance between the starting point and the solution.