Skip to Main content Skip to Navigation
New interface
Reports (Research report)

Distributed Social Graph Embedding

Abstract : Content recommendation is becoming central in the Web 2.0 to leverage the growing information on users available today. In this paper we propose a decentralized gossip-based algorithm called SoCS (Social Coordinate Systems) that achieves efficient distributed social graph embedding for content recommendation purposes. Social graph embedding embeds a graph into a d-dimensional Euclidean coordinate space. SoCS relies on a force-based graph embedding technique to extract communities from a graph. We explore here a distributed algorithm for it (i) scales to large dynamic graph, aggregating the computing power of individual nodes and, (ii) avoids a central entity controlling users sensitive data such as relations and preferences. We evaluate SoCS using two different force-based models and compare them in the context of a generated Kleinberg small-world topology. More specifically, we show that the SoCS graph embedding enables to clearly distinguish between short and long-range links. We also evaluate SoCS against a real DBLP data set, showing that removed links are correctly predicted. Finally, we show that our gossip-based algorithm is extremely resilient to dynamics.
Complete list of metadata

Cited literature [26 references]  Display  Hide  Download
Contributor : Vincent Leroy Connect in order to contact the contributor
Submitted on : Friday, June 25, 2010 - 1:03:58 PM
Last modification on : Thursday, October 27, 2022 - 3:45:38 AM
Long-term archiving on: : Monday, October 22, 2012 - 2:50:36 PM


Files produced by the author(s)


  • HAL Id : inria-00495250, version 1


Anne-Marie Kermarrec, Vincent Leroy, Gilles Tredan. Distributed Social Graph Embedding. [Research Report] RR-7327, INRIA. 2010. ⟨inria-00495250⟩



Record views


Files downloads