Title:Randomized Block Proximal Methods for Distributed Stochastic Big-Data Optimization

Abstract:In this paper we introduce a class of novel distributed algorithms for solving stochastic big-data convex optimization problems over directed graphs. In the addressed set-up, the dimension of the decision variable can be extremely high and the objective function can be nonsmooth. The general algorithm consists of two main steps: a consensus step and an update on a single block of the optimization variable, which is then broadcast to neighbors. Three special instances of the proposed method, involving particular problem structures, are then presented. In the general case, the convergence of a dynamic consensus algorithm over random row stochastic matrices is shown. Then, the convergence of the proposed algorithm to the optimal cost is proven in expected value. Exact convergence is achieved when using diminishing (local) stepsizes, while approximate convergence is attained when constant stepsizes are employed. In both cases, the convergence rate is shown to be sublinear. Finally, a numerical example involving a distributed classification problem is provided to corroborate the theoretical results.