Community Structure Analysis in a Statistical Relational Learning Environment

May 11, 2017
Versione stampabile

Time: May 11, 2017, h. 02:00 pm
Location: Room A203, Polo scientifico e tecnologico "Fabio Ferrari", Building Povo 1, via Sommarive 5, Povo (Trento)

Speaker

Prof. Manfred Jaeger, Institut for Datalogi, Aalborg University

 Abstract

We consider the problem of community detection in multi-relational social networks by means of probabilistic latent variable models.  Latent variables define an embedding of the network nodes in multi-dimensional latent space, which can provide rich structural information from which we not only can  infer graded membership degrees for a given community, but also identify nodes as hubs with strong associations with several communities, or outliers with overall weak associations.

We propose to use tools and techniques from statistical relational learning (SRL) as a platform to perform community structure analysis.  SRL is naturally suited to model rich networks containing multiple link relations and node attributes. However, most SRL frameworks and tools so far have essentially been limited to categorical variables. A recent extension for numeric input relations of the relational Bayesian network (RBN) framework enables us to experiment with a rich class of latent variable models that are supported by general inference and learning algorithms.

Exploiting RBNs as a versatile representational and computational platform, we demonstrate how  latent variable models can be used to
detect rich community structure information. Furthermore, the probabilistic framework enables us to measure the validity of a community in terms of likelihood gain.

 About the Speaker 

Manfred Jaeger studied mathematics in Freiburg, Germany, where he obtained his diploma in mathematics in 1991. He subsequently went to the Max-Planck-Institute for Computer Science in Saarbrücken, where he obtained a PhD in Computer Science from Saarland University in 1995. From 1996-2003 he continued as research associate at the Max-Planck-Institute for Computer Science, and in that period also spent time as postdoctoral researcher at Stanford University, the University of Helsinki, and Freiburg University. In 2002 he obtained the Habilitation in computer science from Saarland University. Since 2003 he is associate professor at the Computer Science department at Aalborg University, Denmark.

Contact person about this talk: Andrea Passerini - andrea.passerini [at] unitn.it