Mattias Villani

About me

I am Professor of Statistics at Stockholm University and Linköping University.

My research focuses on developing computationally efficient Bayesian methods for inference, prediction and decision making using flexible probabilistic models.
My current application areas are machine learning, transportation, neuroimaging and economics.

I currently teach the Master's level courses Bayesian Learning, 7.5 credits and Machine Learning, 7.5 credits. I also give the PhD level course Advanced Bayesian Learning, 7.5 credits every second year.

Selected publications

• The block-Poisson estimator for optimally tuned exact subsampling MCMC, Journal of Computational and Graphical Statistics, 2021 (with Matias Quiroz, Minh Ngoc Tran, Robert Kohn, and Khue-Dung Dang).
• Hamiltonian Monte Carlo with Energy Conserving Subsampling, Journal of Machine Learning Research, 2019 (with Khue-Dung Dang, Matias Quiroz, Robert Kohn and Minh Ngoc Tran)
• Speeding Up MCMC by Efficient Data Subsampling, Journal of the American Statistical Association, 2018 (with Matias Quiroz, Robert Kohn and Minh Ngoc Tran)
• Sparse Partially Collapsed MCMC for Parallel Inference in Topic Models, Journal of Computational and Graphical Statistics, 2018 (with Måns Magnusson, Leif Jonsson and David Broman)
• Fast Bayesian Whole-Brain fMRI Analysis with Spatial 3D Priors, NeuroImage, 2017 (with Per Sidén, Anders Eklund and David Bolin).
• Generalized Smooth Finite Mixtures, Journal of Econometrics, 2012 (with Robert Kohn and David Nott).
• Efficient Bayesian Multivariate Surface Regression, Scandinavian Journal of Statistics, 2013 (with Feng Li).
• Steady State Priors for Vector Autoregressions, Journal of Applied Econometrics, 2009.
• Bayesian Estimation of an Open Economy DSGE Model with Incomplete Pass-Through, Journal of International Economics, 2007 (with Malin Adolfson, Stefan Laséen and Jesper Lindé).
• Bayesian Point Estimation of the Cointegration Space, Journal of Econometrics, 2006.
• Bayesian Reference Analysis of Cointegration, Econometric Theory, 2005.

Research projects

Ongoing

Completed

Spectral Subsampling for Large-Scale Bayesian Inference - Theory and Applications

Technological developments have made it possible to collect large datasets over time and space at low cost. The aim of this research project is to develop statistical models and fast estimation methods for such large-scale spatiotemporal data.