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


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


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

Computer Science > Social and Information Networks

Title: Overlapping and nonoverlapping models

Authors: Huan Qing
Abstract: Consider a directed network with $K_{r}$ row communities and $K_{c}$ column communities. Previous works found that modeling directed networks in which all nodes have overlapping property requires $K_{r}=K_{c}$ for identifiability. In this paper, we propose an overlapping and nonoverlapping model to study directed networks in which row nodes have overlapping property while column nodes do not. The proposed model is identifiable when $K_{r}\leq K_{c}$. Meanwhile, we provide one identifiable model as extension of ONM to model directed networks with variation in node degree. Two spectral algorithms with theoretical guarantee on consistent estimations are designed to fit the models. A small scale of numerical studies are used to illustrate the algorithms.
Subjects: Social and Information Networks (cs.SI); Machine Learning (cs.LG); Physics and Society (physics.soc-ph)
Cite as: arXiv:2111.01392 [cs.SI]
  (or arXiv:2111.01392v1 [cs.SI] for this version)

Submission history

From: Huan Qing [view email]
[v1] Tue, 2 Nov 2021 06:36:52 GMT (151kb)

Link back to: arXiv, form interface, contact.