We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: Link Prediction in the Stochastic Block Model with Outliers

Authors: Solenne Gaucher (LMO), Olga Klopp (CREST), Geneviève Robin (ENPC, MATHERIALS)
Abstract: The Stochastic Block Model is a popular model for network analysis in the presence of community structure. However, in numerous examples, the assumptions underlying this classical model are put in default by the behaviour of a small number of outlier nodes such as hubs, nodes with mixed membership profiles, or corrupted nodes. In addition, real-life networks are likely to be incomplete, due to non-response or machine failures. We introduce a new algorithm to estimate the connection probabilities in a network, which is robust to both outlier nodes and missing observations. Under fairly general assumptions, this method detects the outliers, and achieves the best known error for the estimation of connection probabilities with polynomial computation cost. In addition, we prove sub-linear convergence of our algorithm. We provide a simulation study which demonstrates the good behaviour of the method in terms of outliers selection and prediction of the missing links.
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Social and Information Networks (cs.SI)
Cite as: arXiv:1911.13122 [stat.ML]
  (or arXiv:1911.13122v1 [stat.ML] for this version)

Submission history

From: Solenne Gaucher [view email]
[v1] Fri, 29 Nov 2019 14:33:00 GMT (78kb,D)

Link back to: arXiv, form interface, contact.