SYSYSep 21, 2018

A Distributed Algorithm for Solving Linear Algebraic Equations Over Random Networks

arXiv:1809.0795536 citations
Originality Synthesis-oriented
AI Analysis

It addresses the problem of distributed linear equation solving under random communication topologies without requiring B-connectivity, which is relevant for multi-agent systems with unreliable communication.

This paper presents a distributed algorithm for solving linear equations Ax=b over random networks, where each agent knows only a subset of rows. The algorithm converges almost surely and in mean square to a solution, with the limit point determined by a convex optimization problem, but the convergence rate cannot be guaranteed.

In this paper, we consider the problem of solving linear algebraic equations of the form $Ax=b$ among multi agents which seek a solution by using local information in presence of random communication topologies. The equation is solved by $m$ agents where each agent only knows a subset of rows of the partitioned matrix $[A,b]$. We formulate the problem such that this formulation does not need the distribution of random interconnection graphs. Therefore, this framework includes asynchronous updates or unreliable communication protocols without B-connectivity assumption. We apply the random Krasnoselskii-Mann iterative algorithm which converges almost surely and in mean square to a solution of the problem for any matrices $A$ and $b$ and any initial conditions of agents' states. We demonestrate that the limit point to which the agents' states converge is determined by the unique solution of a convex optimization problem regardless of the distribution of random communication graphs. Eventually, we show by two numerical examples that the rate of convergence of the algorithm cannot be guaranteed.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes