Tue Herlau

About

Aim

Statistical methods for modeling complex networks have received much attention in the machine learning community in recent years. Models and inference tools have matured considerably and are increasingly applied in network science. Contributions include; expressive generative models, a wide range of methods for efficient approximate inference including MCMC and variational approaches, and principled ways to predict unobserved data, model uncertainty and validate structure. Applications have been pursued ranging from web scale social science to brain networks

This workshop aims at introducing machine learning techniques to a wider network science audience. The workshop will feature leading experts in machine learning in network data who will present their views and present the most important challenges modeling complex networks including: What is best practice for validating network models? What are the principal modeling challenges? Which models and techniques admit analysis of large scale networks?

Topics of interests include but are not limited to

Statistical models and relational learning
(generative models for networks, latent variable models, Bayesian non-parametrics).
Multiscale, multiplex and time-evolving models for networks.
Large scale network modeling and efficient inference procedures
(distributed computing, MCMC, maximum likelihood, variational inference, online learning).
Model assessment
(link-prediction and cross-validation, comparing inferred structures to ground truth, and posterior predictive checking).

Program

Speakers

9:00-9:05: Opening remarks
9:05-9:50: Mapping change in large integrated systems; Martin Rosvall; Umeå University
9:50-10:35: A single new algorithm for many new applications; Charles Elkan; University of California, San Diego
10:35-10:50: Coffee Break
10:50-11:35: Modeling networks with node attributes; Jure Leskovec; Stanford University
11:35-12:20: Dynamic probabilistic models for latent feature propagation in social networks; Creighton Heaukulani; University of Cambridge
12:20-13:30: Lunch
13:30-14:15: Statistical inference for the discovery of hidden interactions in complex networks; Roger Guimerà; Universitat Rovira i Virgili
14:15-15:00: Generative Models for Complex Network Structure; Aaron Clauset; University of Colorado at Boulder and Santa Fe Institute
15:00-15:20: Coffee Break
15:20-16:05: Bayesian nonparametric network models: latent space and latent attribute approaches; James Lloyd; University of Cambridge
16:05-16:50: Scalable Model Selection for Networks using Belief Propagation; Xiaoran Yan; Santa Fe Institute
: Contributed Talks:
16:50-17:10: Minimum curvilinearity to enhance topological prediction of protein interactions by network embedding; Carlo Vittorio Cannistraci; University of California San Diego (UCSD), USA & KAUST Saudi Arabia
17:10-17:30: Newer bounds in the Switching Algorithm: fixed point and time convergence; Andrea Gobbi; University of Trento

Practical

Registration

If you plan to participate at the workshop, please pre-register at http://www.eventbrite.com/event/6288389743. Registration is free of charge but we kindly ask participants to register for organizational reasons.

Location

The workshop will be located at the campus of the Technical University of Denmark, building 101, seminar room S12.

Organizers

Morten Mørup, Mikkel N. Schmidt, Tue Herlau, and Lars Kai Hansen

NetSci 2013 Satellite Symposium, June 4th