Registered Data

[00996] A Decentralized Approach for Dynamic Graph Clustering

  • Session Time & Room : 5D (Aug.25, 15:30-17:10) @G709
  • Type : Contributed Talk
  • Abstract : Interconnected networks characterized by interacting agents can be represented by weighted graphs, with the weight indicating their connection strength. Graph clustering arises naturally in these networks to assist decision making and co-ordination. Among the clustering methods, spectral clustering has emerged as a powerful tool but suffers from slow convergence for large dynamic graphs. Thus, we propose a fast incremental approach for dynamic graphs as an extension of its equivalent decentralized approach based on wave propagation.
  • Classification : 37Mxx, 05Cxx, 65Fxx
  • Format : Talk at Waseda University
  • Author(s) :
    • Hongyu Zhu (Raytheon Technologies Research Center)
    • Hongyu Zhu (Raytheon Technologies Research Center)
    • Tuhin Sahai (SRI International)