Convergence Properties of the Distributed Projected Subgradient Algorithm over General Graphs

TL;DR

This paper analyzes the convergence of a distributed projected subgradient algorithm over general, possibly time-varying graphs, establishing conditions for convergence under various switching patterns and network assumptions.

Contribution

It extends convergence analysis to general row-stochastic weight matrices and provides necessary and sufficient conditions for convergence under different graph switching scenarios.

Findings

Convergence fails for some arbitrary graph sequences.

A necessary and sufficient condition for convergence under uniformly jointly strongly connected graphs.

Convergence guaranteed for periodically and certain quasi-periodically switching graphs.

Abstract

In this paper, we revisit a well-known distributed projected subgradient algorithm which aims to minimize a sum of cost functions with a common set constraint. In contrast to most of existing results, weight matrices of the time-varying multi-agent network are assumed to be more general, i.e., they are only required to be row stochastic instead of doubly stochastic. We focus on analyzing convergence properties of this algorithm under general graphs. We first show that there generally exists a graph sequence such that the algorithm is not convergent when the network switches freely within finitely many general graphs. Then to guarantee the convergence of this algorithm under any uniformly jointly strongly connected general graph sequence, we provide a necessary and sufficient condition, i.e., the intersection of optimal solution sets to all local optimization problems is not empty. Furthermore, we surprisingly find that the algorithm is convergent for any periodically switching general graph sequence, and the converged solution minimizes a weighted sum of local cost functions, where the weights depend on the Perron vectors of some product matrices of the underlying periodically switching graphs. Finally, we consider a slightly broader class of quasi-periodically switching graph sequences, and show that the algorithm is convergent for any quasi-periodic graph sequence if and only if the network switches between only two graphs.