Scientific interests - links to project pages and papers can be found below:
- Machine learning. Novel methods for solving difficult inference problems for example variational methods for deep learning.
- Approximate Bayesian inference with deterministic methods. Manfred Opper and I have contributed to the development of Expectation Propagation (formerly known as Adaptive TAP and also called Expectation Consistent by us). Current interests are perturbation corrections to EP with Ulrich Paquet and Manfred Opper and linear response and cavity methods for improving the Laplace approximation with Aki Vehtari and Aki's group.
- Bioinformatics with specific focus on gene regulation (see gene regulation group homepage). Albin Sandelin, Anders Krogh and I are interested in probabilistic modelling of genomic data such as gene expression micro-array type and sequence tags, chip-on-chip and CAGE in close collaboration with experimental groups at BRIC, the University hospital and Riken, Japan. My part of the group focus primarily on cancer research and use biological, machine learning and systems biology methods to address classification and survival.
- Approximate Bayesian inference with Markov chain Monte Carlo methods. I am interesting in MCMC for providing ground truth for deterministic methods and for making MCMC practical for large problems. This involves improving the mixing properties of Gibbs sampling type algorithms, proposal distributions for non-linear state space models and generalized ensemble methods to get marginal likelihood estimates.
- Some machine learning models that have been and still are very interesting: classification with Gaussian processes, independent component analysis (ICA), low rank matrix factorization for collaborative filtering, non-parametric Bayes, sparse linear latent variable models and non-linear state space models.
- Information retrieval in the medical domain.
- Probabilistic modeling in EEG and fMRI.
Links to projects with separate pages
- Sparse Linear Identifiable Multivariate Modeling (SLIM)
- Propulsion modelling
- Ordinal matrix factorization (collaborative filtering)
- PASS-GP: Predictive Active Set Selection for Gaussian Processes
- FindZebra: a Search Engine for Rare Diseases
- HemaExplorer and BloodSpot - analyse the expression of your favourite gene in the hematopoietic system
Publications: Google scholar
Current students (main supervisor unless otherwise stated):
- Michael Riis Andersen (DTU, co-supervisor)
- Jonas Busk (DTU)
- Martin Bach-Andersen (DTU, Bo Rømer-Odgaard, Siemens Wind Power, co-supervisor)
- Marco Fraccaro (DTU, Tom Minka, Microsoft Cambridge and Ulrich Paquet, Deepmind, co-supervisors)
- Ditte Høvenhoff Hald (DTU)
- Tomas Martin-Bertelsen (KU)
- Lars Maaløe (DTU)
- Rasmus Berg Palm (DTU, Florian Laws, Tradeshift, co-supervisor)
- Konrad Stanak (DTU)
- Dan Svenstrup (DTU)
- Casper Sønderby (KU)
- Søren Sønderby (KU)
- David Kofoed Wind (DTU)
- Trine Julie Abrahamsen (DTU, co-supervisor)
- Morten Nonboe Andersen (DTU)
- Frederik Otzen Bagger (KU)
- Thomas Beierholm (DTU)
- Morten Hansen (DTU)
- Rehannah Borup Helweg-Larsen (KU)
- Ricardo Henao (DTU and KU)
- Jens Vilstrup Johansen (KU)
- Bogumil Kaczkowski (KU)
- Ulla Brasch Mogensen (KU, co-supervisor)
- Lars Rønn Olsen (KU, PhD and postdoc)
- Jóan Petur Petersen (DTU)
- Kare Brandt Petersen (DTU)
- Nicolas Rapin (KU, postdoc)
- Sune Olander Rasmussen (KU, co-supervisor)
- Carsten Stahlhut (DTU, co-supervisor)
- Man-Hung Eric Tang (KU, joint supervision with Albin Sandelin and Anders Krogh)
- Eivind Valen (KU, joint supervision with Albin Sandelin and Anders Krogh)
Master projects are available for DTU and KU students.