Thomas Möllenhoff
Research Scientist
Approximate Bayesian Inference Team
RIKEN Center for Advanced Intelligence Project
Tokyo, Japan
thomas.moellenhoff (at) riken (dot) jp
CV – Scholar – github – Twitter
About
I’m a tenured research scientist at the RIKEN Center for Advanced Intelligence Project, affiliated with the Approximate Bayesian Inference team and a core member of the Bayes Duality project. Before that, I was a postdoctoral researcher working with Emtiyaz Khan. I completed my PhD in the Computer Vision Group at TU Munich under the guidance of Daniel Cremers.
My current research focuses on the design and analysis of new algorithms to improve deep learning via Bayesian principles, with the aim to develop methods that are robust, adaptable and more interpretable.
News
- December 4–5, 2024. Invited talk at OCAMI Workshop, Osaka, Japan.
- November 7, 2024. Invited talk at the IBIS 2024 Machine Learning Theory workshop, Tokyo, Japan.
- September 26, 2024. I will be an area chair for ICLR and AISTATS 2025.
- September 25-27, 2024. Invited talk at KAIST in Daejong, Korea.
- September 12, 2024. Invited talks at the Seminar on Advances in Probabilistic Machine Learning (Aalto University) and the Mathematical Optimization Research Seminar (Saarland University).
- July 21-27, 2024. I am at ICML 2024 in Vienna, Austria for a spotlight poster presentation on our recent work.
- July 9-19, 2024. I am visiting the Data Science Group at LMU Munich and gave a talk in the Colloquium of the Department of Statistics.
- July 18, 2024. I am visiting Vincent Fortuin’s group at Helmholtz AI for an invited talk.
- July 8, 2024. I visited the Computer Vision Group at TU Munich and gave a talk about our recent IVON optimizer.
- July 1-7, 2024. Invited talk at the 2024 ISBA World Meeting in Venice, Italy.
- June 12-28, 2024. I am attending the second Bayes-duality workshop in Tokyo, Japan.
- June 5-7, 2024. Poster presentation at the LAMDA-RIKENAIP joint workshop at Nanjing University, China.
- June 4, 2024. Guest talk at SambaNova Systems on model merging.
- May 21, 2024. Invited talk at a joint symposium of RIKEN AIP and the Italian Institute of Technology.
- May 14, 2024. I am giving a talk at the Weierstraß-Instutit in Berlin and visiting Jia-Jie Zhu’s group.
- May 7-11, 2024. Presenting two posters at ICLR 2024.
- May 2, 2024. Our paper Variational Learning is Effective for Large Deep Networks is accepted at ICML 2024.
- Feb 29, 2024. New preprint. We show that large models like GPT-2 can be successfully trained and improved using variational learning
- Jan 17, 2024. Two papers are accepted at ICLR 2024: [Model Merging by Uncertainty-Based Gradient Matching] and [Conformal Prediction via Regression-as-Classification].
See older news.