Exploitation de la structure en communautés pour la réduction de défauts de cache dans la fouille des réseaux sociaux

Abstract : One of social graphs property is the community structure. When executing on graphs, most social network mining algorithms make reference to the neighborhood of the current treated node, to nodes situated to its community. Can we reduce the execution time of social graph algorithm if we take into account the community structure of graph in the data structure used in the runtime? The goal of such a data structure is to reduce cache misses, which will reduce execution time. In this paper, we present a data structure responding to this goal. A comparaison between this data structure with those used in graph analysis DSLs and platforms (Yale representation, adjacency list or bloc) shows that taking into account this property contribute to reduce cache misses (until 20%) and to reduce execution time (until 14%) of social network mining programms.
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01252798
Contributor : Jean-Francois Méhaut <>
Submitted on : Friday, January 8, 2016 - 11:37:34 AM
Last modification on : Wednesday, February 13, 2019 - 4:06:06 PM

Identifiers

  • HAL Id : hal-01252798, version 1

Collections

Citation

Thomas Messi Nguélé, Maurice Tchuenté, Jean-François Mehaut. Exploitation de la structure en communautés pour la réduction de défauts de cache dans la fouille des réseaux sociaux. Conférence de Recherche en Informatique (CRI), Dec 2015, Yaoundé, Cameroun. ⟨hal-01252798⟩

Share

Metrics

Record views

392