I have very broad interests in machine learning, including supervised, unsupervised and reinforcement learning. I'm particularly interested in the probabilistic (Bayesian) approach to represent the reliability of knowlegde, without which it seems very difficult to envisage intelligent behaviour from a learning system.

Gaussian Processes

I have worked extensively on Gaussian Process (GP) models for regression and classification. GPs are flexible probabilistic kernels machines. I have co-authored a book with Chris Williams entitled Gaussian Processes for Machine Learning published by the MIT Press. I also work on mixture models based on the Dirichlet Process (infinite mixture models).

Inference

Exact inference in complex probabilistic models is frequently intractible, so one has to resort to approximation techniques, such as variational techinques and Markov Chain Monte Carlo. I work on the development and assessment of these techniques for machine learning problems. I'm interested in computationally challenging probabilistic inference in statistical models.

Teaching

This semester I am (co-)teaching Unsupervised Learning at the Computer Science department of the Univeristy of Tübingen.

Last semester I (co-)taught Learning in Computer Vision II at the Computer Science department of the University of Tübingen.