Shared-Memory Parallel Maximal Biclique Enumeration

We present shared memory parallel algorithms for maximal biclique enumeration (MBE), the task of enumerating all complete dense subgraphs (maximal bicliques) from a bipartite graph, which is widely used in the analysis of social, biological, and transactional networks. Since MBE is computationally e...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Proceedings - International Conference on High Performance Computing S. 34 - 43
Hauptverfasser: Das, Apurba, Tirthapura, Srikanta
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 01.12.2019
Schlagworte:
Biology
Bipartite graph
dynamic graph
graph algorithm
High performance computing
maximal biclique enumeration
Memory management
Parallel algorithms
shared memory parallel algorithm
ISSN:2640-0316
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:We present shared memory parallel algorithms for maximal biclique enumeration (MBE), the task of enumerating all complete dense subgraphs (maximal bicliques) from a bipartite graph, which is widely used in the analysis of social, biological, and transactional networks. Since MBE is computationally expensive, it is necessary to use parallel computing to scale to large graphs. Our parallel algorithm ParMBE efficiently uses the power of multiple cores that share memory. From a theoretical view, ParMBE is work-efficient with respect to a state-of-the-art sequential algorithm. Our experimental evaluation shows that ParMBE scales well up to 64 cores, and is significantly faster than current parallel algorithms. Since ParMBE was yielding a super-linear speedup compared to the sequential algorithm on which it was based (MineLMBC), we develop an improved sequential algorithm FMBE, through "sequentializing" ParMBE
ISSN:2640-0316
DOI:10.1109/HiPC.2019.00016