I joined the Department in August 2014. I work in the EPSRC project "A Population Approach to Ubicomp System Design" developing advanced statistical models and inference methods for analyzing populations of software instances. My research interests include nonparametric Bayes, sparsity, Bayesian statistics, latent variable models and multi-view learning.
Prior to joining the Department I worked at the Helsinki Institute for Information Technology HIIT, Department of Information and Computer Science (ICS), Aalto University, with Prof. Sami Kaski, developing novel Bayesian latent variable models for learning dependencies between multiple data sources. The R package CCAGFA implements some of these models using variational Bayes. This Google Scholar page contains an up-to-date list of my publications.
Seppo Virtanen, Mattias Rost, Alistair Morrison, Matthew Chalmers and Mark Girolami. Uncovering smartphone usage patterns with multi-view mixed membership models. To appear in STAT.
Seppo Virtanen, Mattias Rost, Matthew Higgs, Alistair Morrison, Matthew Chalmers, and Mark Girolami. Non-parametric Bayes to infer playing strategies adopted in a population of mobile gamers. STAT, Wiley, 2015.
Arto Klami, Seppo Virtanen, Eemeli Leppa-aho, and Samuel Kaski. Group factor analysis. To appear in IEEE TNNLS, 2015.
Seppo Virtanen, Yangqing Jia, Arto Klami, and Trevor Darrell. Factorized Multi-Modal Topic Model. In Nando de Freitas and Kevin Murphy, editors, Proceedings of the Twenty-Eighth Conference on Uncertainty in Artificial Intelligence, pages 843–851, Corvallis, Oregon, 2012. AUAI Press. PDF
The novel topic model learns topics that are shared between the modalities (views) as well as topics specific to each modality using Hierarchical Dirichlet Process (HDP) formulation. The model was applied for a collection of Wikipedia pages that consist of images and the whole text on the page.