Network Structure Change Point Detection by Posterior Predictive Discrepancy

Lingbin Bian, Tiangang Cui, Georgy Sofronov, Jonathan Keith

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearch

Abstract

Detecting changes in network structure is important for research into systems as diverse as financial trading networks, social networks and brain connectivity. Here we present novel Bayesian methods for detecting network structure change points. We use the stochastic block model to quantify the likelihood of a network structure and develop a score we call posterior predictive discrepancy based on sliding windows to evaluate the model fitness to the data. The parameter space for this model includes unknown latent label vectors assigning network nodes to interacting communities. Monte Carlo techniques based on Gibbs sampling are used to efficiently sample the posterior distributions over this parameter space.

Original languageEnglish
Title of host publicationMonte Carlo and Quasi-Monte Carlo Methods
Subtitle of host publication13th International Conference on Monte Carlo and Quasi-Monte Carlo Methods in Scientific Computing, MCQMC 2018; Rennes; France; 1 July 2018 through 6 July 2018
EditorsBruno Tuffin, Pierre L’Ecuyer
Place of PublicationSwitzerland
PublisherSpringer-Praxis
Pages107-123
Number of pages17
Volume324
ISBN (Electronic)978-3-030-43465-6
ISBN (Print)9783030434649
DOIs
Publication statusPublished - 2 May 2020
EventInternational Conference on Monte Carlo and Quasi-Monte Carlo Methods in Scientific Computing, 2018 - University of Rennes, Rennes, France
Duration: 1 Jul 20186 Jul 2018
Conference number: 13th
http://mcqmc2018.inria.fr/

Publication series

NameSpringer Proceedings in Mathematics and Statistics
Volume324
ISSN (Print)2194-1009
ISSN (Electronic)2194-1017

Conference

ConferenceInternational Conference on Monte Carlo and Quasi-Monte Carlo Methods in Scientific Computing, 2018
Abbreviated titleMCQMC 2018
CountryFrance
CityRennes
Period1/07/186/07/18
OtherThe MCQMC Conference is a biennial meeting on Monte Carlo and quasi-Monte Carlo methods. It usually attracts 150 to 200 mathematicians, computer scientists, statisticians and researchers in related fields.
Internet address

Keywords

  • Bayesian inference
  • Gibbs sampling
  • Networks
  • Sliding window
  • Stochastic block model

Cite this