A GPU-accelerated local search algorithm for the Correlation Clustering problem

Abstract : The solution of the Correlation Clustering (CC) problem can be used as a criterion to measure the amount of balance in signed social networks, where positive (friendly) and negative (antagonistic) interactions take place. Metaheuristics have been used successfully for solving not only this problem, as well as other hard combinatorial optimization problems, since they can provide sub-optimal solutions in a reasonable time. In this work, we present an alternative local search implementation based on GPGPUs, which can be used with existing GRASP and ILS metaheuristics for the CC problem. This new approach outperforms the existing local search procedure in execution time, with similar solution quality, presenting average speedups from x1.8 to x28.
Complete list of metadatas

Cited literature [26 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01449689
Contributor : Mario Levorato <>
Submitted on : Monday, March 6, 2017 - 11:12:40 PM
Last modification on : Monday, July 1, 2019 - 11:38:06 AM
Long-term archiving on : Wednesday, June 7, 2017 - 12:14:11 PM

File

SBPO2015.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01449689, version 1

Collections

Citation

Mario Levorato, Lúcia Drummond, Yuri Frota, Rosa Figueiredo. A GPU-accelerated local search algorithm for the Correlation Clustering problem. Proceedings of the Brazilian Symposium on Operations Research, SOBRAPO - Brazilian Society of Operations Research, Aug 2015, Porto de Galinhas, PE, Brazil. ⟨hal-01449689⟩

Share

Metrics

Record views

77

Files downloads

273