Bi-objective matchings with the triangle inequality

Abstract : This article deals with a bi-objective matching problem. The input is a complete graph and two values on each edge (a weight and a length) which satisfy the triangle inequality. It is unlikely that every instance admits a matching with maximum weight and maximum length at the same time. Therefore, we look for a compromise solution, i.e. a matching that simultaneously approximates the best weight and the best length. For which approximation ratio ρ can we guarantee that any instance admits a ρ-approximate matching? We propose a general method which relies on the existence of an approximate matching in any graph of small size. An algorithm for computing a 1/3-approximate matching in any instance is provided. The algorithm uses an analytical result stating that every instance on at most 6 nodes must admit a 1/2-approximate matching. We extend our analysis with a computer-aided approach for larger graphs, indicating that the general method may produce a 2/5-approximate matching. We conjecture that a 1/2-approximate matching exists in any bi-objective instance satisfying the triangle inequality.
Document type :
Journal articles
Complete list of metadatas

Cited literature [19 references]  Display  Hide  Download

https://hal.sorbonne-universite.fr/hal-01488424
Contributor : Fanny Pascual <>
Submitted on : Monday, March 13, 2017 - 4:08:53 PM
Last modification on : Monday, May 6, 2019 - 11:50:03 AM
Long-term archiving on : Wednesday, June 14, 2017 - 2:05:13 PM

File

bimatchfinal.pdf
Files produced by the author(s)

Identifiers

Citation

Laurent Gourvès, Jérôme Monnot, Fanny Pascual, Daniel Vanderpooten. Bi-objective matchings with the triangle inequality. Theoretical Computer Science, Elsevier, 2017, 670, pp.1 - 10. ⟨10.1016/j.tcs.2017.01.012⟩. ⟨hal-01488424⟩

Share

Metrics

Record views

417

Files downloads

239