Thomas Hamelryck

News

01/09/2025 New grant! Danmarks Frie Forskningsfond | Teknologi og Produktion funds our interdisciplinary research on deep probabilistic models of protein evolution. Main PI is Hiren Joshi, SUND, one postdoc, DKK 6.2 million.
26/05/2025 New arXiv publication! We explain the first version of the Nobel prize winning protein structure prediction method AlphaFold as an application of probability kinematics, a form of Bayesian updating that makes use of soft evidence. Joint publication with Kanti Mardia, Leeds.
06/03/2025 Thomas gives an invited talk at the Dpt. of Statistics, University of Oxford, on the statistical underpinnings of AlphaFold's first breakthrough in probability kinematics.
22/01/2025 Our article "ELBOing Stein: Variational Bayes with Stein Mixture Inference" was accepted by the leading AI conference ICLR2025. SMI bridges the gap between particle based inference and variational inference, offering new directions for Bayesian estimation of deep models. Congratulations to first author and postdoc Ola Rønning!
14-15/10/2024 Thomas is invited speaker at a two-day workshop on Machine Learning and Statistical Methods for Biological Data, Nijmegen, NL, organized by the EU COST programme DYNALIFE. Title of the talk was "Deep probabilistic programming for bioinformatics".
01/02/2024 Thomas is appointed to the position of full professor in High dimensional biological data analysis / Machine learning. The new position continues to be shared between the departments of biology and computer science at KU.
26/01/2024 PhD student Jon Eklöf successfully defends his PhD. Jon studied the role of leadership and multidisciplinary collaboration in successfully integrating artificial intelligence into the complex manufacturing industry, in collaboration with GKN Aerospace.
12/09/2023: Our article "AI Implementation and Capability Development in Manufacturing: An Action Research Case" by PhD student Jon Eklöf was accepted by the Hawaii International Conference on System Sciences (HICSS), 2023.
09/06/2023: PhD student Kübra Altinel (co-supervised by Thomas, main supervisor Anders H. Lund) sucessfully defends her PhD on using a semi-supervised variational autoencoder for characterization of ribosome subtypes.
31/05/2023: Our article "Abstraction, mimesis and the evolution of deep learning" by PhD student Jon Eklöf was accepted by the journal "AI & Society".
17/04/2023: PhD student Ola Rønning successfully defends his PhD. Ola worked on deep generative models of protein structure and Stein variational inference for Bayesian deep models. Funding came from the Independent Research Fund Denmark. Co-supervisor was Assoc. Prof. Christophe Ley, University of Luxemburg
13/03/2023: Thomas becomes Senior Associate Editor of the new journal ACM Transactions on Probabilistic Machine Learning (TOPML).
04/01/2023: Christian Thygesen, Industrial PhD student with Evaxion, successfully defends his PhD. Christian worked on deep generative models of protein structure (published in ICML, 2021), with applications in vaccine design.
08/09/2022: Thomas is awarded a Villum EXPERIMENT grant ("Predicting protein structure, mutations and dynamics with deep generative models", DKK 1,8 million.)
22/04/2022: Lys Sanz Moreta successfully defends her PhD. Lys worked on deep generative models of protein evolution, with applications in ancestral sequence reconstruction (published in ICLR, 2022).