Mattes Mollenhauer




I am a machine learning researcher at Merantix Momentum. Currently, my work is focused on combining ideas from high-dimensional probability, nonparametric statistics and stochastic processes for inference in inverse problems and time series data. Over the last years, I have been particularly interested in models with downstream applications in biomedical research, causality and statistical hypothesis testing as well as dimensionality reduction.

Academic research

In 2022, I obtained my PhD under the supervision of Christof Schütte in the Biocomputing group at the Department of Mathematics and Computer Science of Freie Universität Berlin. My dissertation investigates theoretical aspects of nonparametric models for Markov processeswith relevant applications in fields like molecular dynamics (such as computational drug design) and fluid mechanics (such as complex combustion processes). My recent work together with Claudia Schillings addresses problems in uncertainty quantification and the data-driven numerical analysis of stochastic differential equations.

Industry experience

During my PhD program, I spent some time doing R&D in the aircraft engine development division of Rolls-Royce Deutschland, where I worked on security critical machine learning software for system identification and control as well as pattern recognition in sensor data. In this context, my main goal was to contribute to robust models aimed at the detection of rare events (such as specific failures of structural components), possibly including synthetic data obtained from simulations and rare event sampling.

After that, I stayed with the people at dida for the past five yearsmainly working on design and devops of large-scale deep learning systems for natural language processing and computer vision. My projects tackled a variety of problems arising in weather forecasting, semantic search, remote sensing and image segmentation. In particular, we made use of ideas from domain adaptation, few-shot learning and various pretraining and fine-tuning techniques, aiming to transform recent deep learning research into production software.

Publications and preprints

Reports and preprints

Dimitri Meunier, Zikai Shen, Mattes Mollenhauer, Arthur Gretton and Zhu Li.
"Optimal Rates for Vector-Valued Spectral Regularization Learning Algorithms".
Preprint, 2024. link

Zhu Li, Dimitri Meunier, Mattes Mollenhauer and Arthur Gretton.
"Towards Optimal Sobolev Norm Rates for the Vector-Valued Regularized Least-Squares Algorithm".
Preprint, 2023. link

Mattes Mollenhauer and Claudia Schillings.
"On the concentration of subgaussian vectors and positive quadratic forms in Hilbert spaces".
Preprint, 2023. link

Mattes Mollenhauer, Nicole Mücke and Tim Sullivan.
"Learning linear operators: Infinite-dimensional regression as a well-behaved non-compact inverse problem".
Preprint, 2022. link

Mattes Mollenhauer and Péter Koltai.
"Nonparametric approximation of conditional expectation operators."
Preprint, 2020. link


Andreas Bittracher, Mattes Mollenhauer, Péter Koltai and Christof Schütte.
“Optimal Reaction Coordinates: Variational Characterization and Sparse Computation”.
SIAM Multiscale Modeling & Simulation, 21:449-488, 2023. link

Zhu Li, Dimitri Meunier, Mattes Mollenhauer and Arthur Gretton.
“Optimal Rates for Regularized Conditional Mean Embedding Learning”.
Advances in Neural Information Processing Systems (NeurIPS). Vol. 36, 2022. link

Mattes Mollenhauer, Stefan Klus, Christof Schütte and Péter Koltai.
“Kernel autocovariance operators of stationary processes: Estimation and convergence”.
Journal of Machine Learning Research, 23(327):1−34, 2022. link

Mattes Mollenhauer, Ingmar Schuster, Stefan Klus and Christof Schütte.
“Singular Value Decomposition of Operators on Reproducing Kernel Hilbert Spaces”.
Advances in Dynamics, Optimization and Computation, 109-130, Springer, 2020. link

Ingmar Schuster, Mattes Mollenhauer, Stefan Klus and Krikamol Muandet.
“Kernel Conditional Density Operators”.
23rd International Conference on Artificial Intelligence and Statistics (AISTATS), 2020. link

Stefan Klus, Brooke E. Husic, Mattes Mollenhauer and Frank Noé.
“Kernel methods for detecting coherent structures in dynamical data”.
Chaos: An Interdisciplinary Journal of Nonlinear Science 29.12, 2019. link


"On the Statistical Approximation of Conditional Expectation Operators".
PhD thesis, Freie Universität Berlin. 2022. link

Design courtesy of Vasilios Mavroudis: Plain Academic