Downloads: 204

Files in This Item:
File Description SizeFormat 
LIPIcs.APPROX-RANDOM.2015.361.pdf545 kBAdobe PDFView/Open
Title: A Tight Approximation Bound for the Stable Marriage Problem with Restricted Ties
Authors: Huang, Chien-Chung
Iwama, Kazuo
Miyazaki, Shuichi  KAKEN_id  orcid https://orcid.org/0000-0003-0369-1970 (unconfirmed)
Yanagisawa, Hiroki
Author's alias: 岩間, 一雄
宮崎, 修一
柳澤, 弘揮
Keywords: stable marriage with ties and incomplete lists
approximation algorithm
integer program
linear program relaxation
integrality gap
Issue Date: 2015
Publisher: Schloss Dagstuhl--Leibniz-Zentrum fuer Informatik
Journal title: Approximation, Randomization, and Combinatorial Optimization. Algorithms and Techniques (APPROX/RANDOM 2015)
Volume: 40
Start page: 361
End page: 380
Abstract: The problem of finding a maximum cardinality stable matching in the presence of ties and unacceptable partners, called MAX SMTI, is a well-studied NP-hard problem. The MAX SMTI is NP-hard even for highly restricted instances where (i) ties appear only in women's preference lists and (ii) each tie appears at the end of each woman's preference list. The current best lower bounds on the approximation ratio for this variant are 1.1052 unless P=NP and 1.25 under the unique games conjecture, while the current best upper bound is 1.4616. In this paper, we improve the upper bound to 1.25, which matches the lower bound under the unique games conjecture. Note that this is the first special case of the MAX SMTI where the tight approximation bound is obtained. The improved ratio is achieved via a new analysis technique, which avoids the complicated case-by-case analysis used in earlier studies. As a by-product of our analysis, we show that the integrality gap of natural IP and LP formulations for this variant is 1.25. We also show that the unrestricted MAX SMTI cannot be approximated with less than 1.5 unless the approximation ratio of a certain special case of the minimum maximal matching problem can be improved.
Rights: Since 2013, documents stored on DROPS are typically published under a CC-BY Creative Commons Attribution 3.0 Unported license.
The authoritative license is always the one given in the PDF.
URI: http://hdl.handle.net/2433/200620
Appears in Collections:Journal Articles

Show full item record

Export to RefWorks


Export Format: 


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.