Parallel heuristics for scalable community detection
Abstract
Community detection has become a fundamental operation in numerous graph-theoretic applications. Despite its potential for application, there is only limited support for community detection on large-scale parallel computers, largely owing to the irregular and inherently sequential nature of the underlying heuristics. In this paper, we present parallelization heuristics for fast community detection using the Louvain method as the serial template. The Louvain method is an iterative heuristic for modularity optimization. Originally developed in 2008, the method has become increasingly popular owing to its ability to detect high modularity community partitions in a fast and memory-efficient manner. However, the method is also inherently sequential, thereby limiting its scalability. Here, we observe certain key properties of this method that present challenges for its parallelization, and consequently propose heuristics that are designed to break the sequential barrier. For evaluation purposes, we implemented our heuristics using OpenMP multithreading, and tested them over real world graphs derived from multiple application domains. Compared to the serial Louvain implementation, our parallel implementation is able to produce community outputs with a higher modularity for most of the inputs tested, in comparable number or fewer iterations, while providing real speedups of up to 16x using 32 threads.
- Authors:
- Publication Date:
- Research Org.:
- Pacific Northwest National Laboratory (PNNL), Richland, WA (United States)
- Sponsoring Org.:
- USDOE
- OSTI Identifier:
- 1198720
- Alternate Identifier(s):
- OSTI ID: 1208748
- Report Number(s):
- PNNL-SA-108735
Journal ID: ISSN 0167-8191; S0167819115000472; PII: S0167819115000472
- Grant/Contract Number:
- AC05-76RL01830
- Resource Type:
- Journal Article: Published Article
- Journal Name:
- Parallel Computing
- Additional Journal Information:
- Journal Name: Parallel Computing Journal Volume: 47 Journal Issue: C; Journal ID: ISSN 0167-8191
- Publisher:
- Elsevier
- Country of Publication:
- Netherlands
- Language:
- English
- Subject:
- 97 MATHEMATICS AND COMPUTING; community detection; Parallel graph heuristics; Graph coloring; Graph clustering
Citation Formats
Lu, Hao, Halappanavar, Mahantesh, and Kalyanaraman, Ananth. Parallel heuristics for scalable community detection. Netherlands: N. p., 2015.
Web. doi:10.1016/j.parco.2015.03.003.
Lu, Hao, Halappanavar, Mahantesh, & Kalyanaraman, Ananth. Parallel heuristics for scalable community detection. Netherlands. https://doi.org/10.1016/j.parco.2015.03.003
Lu, Hao, Halappanavar, Mahantesh, and Kalyanaraman, Ananth. 2015.
"Parallel heuristics for scalable community detection". Netherlands. https://doi.org/10.1016/j.parco.2015.03.003.
@article{osti_1198720,
title = {Parallel heuristics for scalable community detection},
author = {Lu, Hao and Halappanavar, Mahantesh and Kalyanaraman, Ananth},
abstractNote = {Community detection has become a fundamental operation in numerous graph-theoretic applications. Despite its potential for application, there is only limited support for community detection on large-scale parallel computers, largely owing to the irregular and inherently sequential nature of the underlying heuristics. In this paper, we present parallelization heuristics for fast community detection using the Louvain method as the serial template. The Louvain method is an iterative heuristic for modularity optimization. Originally developed in 2008, the method has become increasingly popular owing to its ability to detect high modularity community partitions in a fast and memory-efficient manner. However, the method is also inherently sequential, thereby limiting its scalability. Here, we observe certain key properties of this method that present challenges for its parallelization, and consequently propose heuristics that are designed to break the sequential barrier. For evaluation purposes, we implemented our heuristics using OpenMP multithreading, and tested them over real world graphs derived from multiple application domains. Compared to the serial Louvain implementation, our parallel implementation is able to produce community outputs with a higher modularity for most of the inputs tested, in comparable number or fewer iterations, while providing real speedups of up to 16x using 32 threads.},
doi = {10.1016/j.parco.2015.03.003},
url = {https://www.osti.gov/biblio/1198720},
journal = {Parallel Computing},
issn = {0167-8191},
number = C,
volume = 47,
place = {Netherlands},
year = {Sat Aug 01 00:00:00 EDT 2015},
month = {Sat Aug 01 00:00:00 EDT 2015}
}
Web of Science