University of Minnesota
University Relations

Minnesota Supercomputing Institute

Log out of MyMSI

Research Abstracts Online
January 2009 - March 2010

Main TOC ....... College TOC ....... Next Abstract

University of Minnesota Twin Cities
Institute of Technology
Department of Computer Science and Engineering

PI: Arindam Banerjee

Multi-relational Data Clustering With Probabilistic Mixture Models

This project concerns large-scale multi-relational data clustering using Bayesian graphical models. It aims at discovering the latent relational clusters across multiple related entities under the framework of mixture models. The project includes two parts: additive models and multiplicative models. For the additive models, one assumes the distribution of an observation to be an additive combination of a set of component distributions, while for the multiplicative models, one assumes the distribution of an observation to be a product of component distributions. The relational clusters should be more meaningful and accurate than the standalone clusters from traditional clustering algorithms. This work applies to recommendation systems, text analysis, and bioinformatics, among others. The work in the first stage of project has addressed an important special case of relational clustering, i.e., co-clustering for two related entities. The focus of the next stage will be on multi-relational clustering with multiple (more than two) entities, represented as a tensor or more general multi-relational data.

Group Members

Amrudin Agovic, Graduate Student
Qiang Fu, Graduate Student
Hanhuai Shan, Graduate Student