Distributed approximate maximum matching in the CoNGEst model

Mohamad Ahmadi, Fabian Kuhn, Rotem Oshman

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

We study distributed algorithms for the maximum matching problem in the CONGEST model, where each message must be bounded in size. We give new deterministic upper bounds, and a new lower bound on the problem. We begin by giving a distributed algorithm that computes an exact maximum (unweighted) matching in bipartite graphs, in O(n log n) rounds. Next, we give a distributed algorithm that approximates the fractional weighted maximum matching problem in general graphs. In a graph with maximum degree at most ∆, the algorithm computes a (1−ε)-approximation for the problem in time O log(∆W)/ε 2 , where W is a bound on the ratio between the largest and the smallest edge weight. Next, we show a slightly improved and generalized version of the deterministic rounding algorithm of Fischer [DISC’17]. Given a fractional weighted maximum matching solution of value f for a given graph G, we show that in time O((log 2 (∆) + log n)/ε), the fractional solution can be turned into an integer solution of value at least (1 − ε)f for bipartite graphs and (1 − ε) · g− g 1 · f for general graphs, where g is the length of the shortest odd cycle of G. Together with the above fractional maximum matching algorithm, this implies a deterministic algorithm that computes a (1 − ε) · g− g 1 -approximation for the weighted maximum matching problem in time O log(∆W)/ε 2 + (log 2 (∆) + log n)/ε. On the lower-bound front, we show that even for unweighted fractional maximum matching in bipartite graphs, computing an (1 − O(1/n))-approximate solution requires at least Ω( D + n) rounds in CONGEST. This lower bound requires the introduction of a new 2-party communication problem, for which we prove a tight lower bound.

Original languageEnglish
Title of host publication32nd International Symposium on Distributed Computing, DISC 2018
EditorsUlrich Schmid, Josef Widder
PublisherSchloss Dagstuhl- Leibniz-Zentrum fur Informatik GmbH, Dagstuhl Publishing
ISBN (Electronic)9783959770927
DOIs
StatePublished - 1 Oct 2018
Event32nd International Symposium on Distributed Computing, DISC 2018 - New Orleans, United States
Duration: 15 Oct 201819 Oct 2018

Publication series

NameLeibniz International Proceedings in Informatics, LIPIcs
Volume121
ISSN (Print)1868-8969

Conference

Conference32nd International Symposium on Distributed Computing, DISC 2018
Country/TerritoryUnited States
CityNew Orleans
Period15/10/1819/10/18

Keywords

  • Communication complexity
  • Deterministic rounding
  • Distributed graph algorithms
  • Maximum matching

Fingerprint

Dive into the research topics of 'Distributed approximate maximum matching in the CoNGEst model'. Together they form a unique fingerprint.

Cite this