Solving the graph edit distance problem with variable partitioning local search

Abstract : In the world of graph matching, the Graph Edit Distance (GED) problem is a well-known distance measure between graphs. It has been proven to be a NP-hard minimization problem. This paper presents an adapted version of Variable Partitioning Local Search (VPLS) matheuristic for solving the GED problem. The main idea in VPLS is to perform local searches in the solution space of a Mixed Integer Linear Program (MILP). A local earch is done in a small neighborhood defined based on a set of special variables. Those special variables are selected based on a procedure that extracts useful characteristics from the instance at hand. This actually ensures that the neighborhood contains high quality solutions. Finally, the experimentation results have shown that VPLS has outperformed existing heuristics in terms of solution quality on CMUHOUSE database.
Document type :
Conference papers
Complete list of metadatas

Cited literature [19 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01579501
Contributor : Mostafa Darwiche <>
Submitted on : Friday, September 13, 2019 - 12:12:16 PM
Last modification on : Saturday, September 14, 2019 - 1:43:38 AM

File

2019GbRGEDVPLS.pdf
Files produced by the author(s)

Identifiers

Citation

Mostafa Darwiche, Donatello Conte, Romain Raveaux, Vincent t'Kindt. Solving the graph edit distance problem with variable partitioning local search. 12th International workshop on Graph-Based Representation in Pattern Recognition (GbR19), Jun 2019, Tours, France. pp.67-77, ⟨10.1007/978-3-030-20081-7_7⟩. ⟨hal-01579501⟩

Share

Metrics

Record views

154

Files downloads

86