Community Detection on Weighted Networks: A Variational Bayesian Method

Authors:
Qixia Jiang;Yan Zhang;Maosong Sun
Affiliations:
State Key Laboratory on Intelligent Technology and Systems, Tsinghua National Laboratory for Information Science and Technology, Department of Computer Science and Technology, Tsinghua University, ...;State Key Laboratory on Intelligent Technology and Systems, Tsinghua National Laboratory for Information Science and Technology, Department of Computer Science and Technology, Tsinghua University, ...;State Key Laboratory on Intelligent Technology and Systems, Tsinghua National Laboratory for Information Science and Technology, Department of Computer Science and Technology, Tsinghua University, ...
Venue:
ACML '09 Proceedings of the 1st Asian Conference on Machine Learning: Advances in Machine Learning
Year:
2009

Citing 4
Cited 0

An Introduction to Variational Methods for Graphical Models

Machine Learning
Latent dirichlet allocation

The Journal of Machine Learning Research
A Bayesian Hierarchical Model for Learning Natural Scene Categories

CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 2 - Volume 02
Mixed Membership Stochastic Blockmodels

The Journal of Machine Learning Research

Quantified Score

Hi-index	0.00

Visualization

Abstract

Massive real-world data are network-structured, such as social network, relationship between proteins and power grid. Discovering the latent communities is a useful way for better understanding the property of a network. In this paper, we present a fast, effective and robust method for community detection. We extend the constrained Stochastic Block Model (conSBM) on weighted networks and use a Bayesian method for both parameter estimation and community number identification. We show how our method utilizes the weight information within the weighted networks, reduces the computation complexity to handle large-scale weighted networks, measure the estimation confidence and automatically identify the community number. We develop a variational Bayesian method for inference and parameter estimation. We demonstrate our method on a synthetic data and three real-world networks. The results illustrate that our method is more effective, robust and much faster.