Bayesian inverse problems have gained significant attention in recent years due to their strong mathematical foundations, which allow for thorough theoretical analysis. Moreover, the incorporation of measurement errors and noise is of utmost importance to obtain reliable results, for instance, in the context of computational imaging problems, such as computed tomography (CT) and magnetic resonance imaging (MRI).
Energy-based methods constitute a particular way to model probability distributions via Gibbs densities containing a suitable energy functional. These approaches can be used to model prior distributions in Bayesian inverse problems. The energy functional is parametrized, for instance, via fields of expert models and learned from image data using modern machine learning architectures. Thanks to its rich mathematical foundations, the Bayesian framework allows for rigorous theoretical guarantees regarding the obtained image reconstructions. A review paper on the topic of energy-based models for inverse imaging problems is available in [HHPZ2025].
A crucial challenge in learning the components of the prior from data lies in sampling from the prior distribution. Two methods, based on subgradient steps and the unadjusted Langevin algorithm, were developed in [HHP2024] and can also be applied to certain non-smooth potentials.
@inbook{holler2025ebms,
ISBN = {9780443417894},
ISSN = {1570-8659},
url = {http://dx.doi.org/10.1016/bs.hna.2025.09.001},
DOI = {10.1016/bs.hna.2025.09.001},
booktitle = {Machine Learning Solutions for Inverse Problems: Part A},
publisher = {Elsevier},
year = {2025},
pages = {1–76}
}
@article{habring2024subgradient,
title = {Subgradient Langevin Methods for Sampling from Nonsmooth Potentials},
volume = {6},
ISSN = {2577-0187},
url = {http://dx.doi.org/10.1137/23m1591451},
DOI = {10.1137/23m1591451},
number = {4},
journal = {SIAM Journal on Mathematics of Data Science},
publisher = {Society for Industrial & Applied Mathematics (SIAM)},
author = {Habring, Andreas and Holler, Martin and Pock, Thomas},
year = {2024},
month = {Oct},
pages = {897–925}
}
@article{holler23uq_mh,
title = {Posterior-Variance–Based Error Quantification for Inverse Problems in Imaging},
volume = {17},
ISSN = {1936-4954},
url = {http://dx.doi.org/10.1137/23M1546129},
DOI = {10.1137/23m1546129},
number = {1},
journal = {SIAM Journal on Imaging Sciences},
publisher = {Society for Industrial & Applied Mathematics (SIAM)},
author = {Narnhofer, Dominik and Habring, Andreas and Holler, Martin and Pock, Thomas},
year = {2024},
month = {Feb},
pages = {301–333}
}
To address this, a well-established strategy is to acquire only a reduced amount of data during an MR measurement and to compensate for the missing information using either hand-crafted or learned imaging models that incorporate prior knowledge about the structure of typical images.
Our research group is actively working in this field, particularly in the context of dynamic MRI. In our work, we have for example developed novel image reconstruction methods that reduce MRI data acquisition times by a factor of eight or more, without compromising the quality or diagnostic value of the reconstructed images.
@article{holler17ictgvmri,
title = {Infimal convolution of total generalized variation functionals for dynamic MRI},
volume = {78},
ISSN = {1522-2594},
url = {http://dx.doi.org/10.1002/mrm.26352},
DOI = {10.1002/mrm.26352},
number = {1},
journal = {Magnetic Resonance in Medicine},
publisher = {Wiley},
author = {Schloegl, Matthias and Holler, Martin and Schwarzl, Andreas and Bredies, Kristian and Stollberger, Rudolf},
year = {2016},
month = {Aug},
pages = {142–155}
}
@article{holler18b1_mh,
title = {Ultrafast 3D Bloch–Siegert B‐mapping using variational modeling},
volume = {81},
ISSN = {1522-2594},
url = {http://dx.doi.org/10.1002/mrm.27434},
DOI = {10.1002/mrm.27434},
number = {2},
journal = {Magnetic Resonance in Medicine},
publisher = {Wiley},
author = {Lesch, Andreas and Schlöegl, Matthias and Holler, Martin and Bredies, Kristian and Stollberger, Rudolf},
year = {2018},
month = {Oct},
pages = {881–892}
}
@article{holler16mrpet,
title = {Joint MR-PET Reconstruction Using a Multi-Channel Image Regularizer},
volume = {36},
ISSN = {1558-254X},
url = {http://dx.doi.org/10.1109/TMI.2016.2564989},
DOI = {10.1109/tmi.2016.2564989},
number = {1},
journal = {IEEE Transactions on Medical Imaging},
publisher = {Institute of Electrical and Electronics Engineers (IEEE)},
author = {Knoll, Florian and Holler, Martin and Koesters, Thomas and Otazo, Ricardo and Bredies, Kristian and Sodickson, Daniel K},
year = {2017},
month = {Jan},
pages = {1–16}
}
@inbook{holler22gans_mri_mh,
title = {Latent-Space Disentanglement with Untrained Generator Networks for the Isolation of Different Motion Types in Video Data},
ISBN = {9783031319754},
ISSN = {1611-3349},
url = {http://dx.doi.org/10.1007/978-3-031-31975-4_25},
DOI = {10.1007/978-3-031-31975-4_25},
booktitle = {Scale Space and Variational Methods in Computer Vision},
publisher = {Springer International Publishing},
author = {Abdullah, Abdullah and Holler, Martin and Kunisch, Karl and Landman, Malena Sabate},
year = {2023},
pages = {326–338}
}
@inproceedings{knoll2016petmritgv,
title = {Simultaneous PET-MRI reconstruction with vectorial second order total generalized variation},
url = {http://dx.doi.org/10.1109/NSSMIC.2015.7582009},
DOI = {10.1109/nssmic.2015.7582009},
booktitle = {2015 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)},
publisher = {IEEE},
author = {Knoll, Florian and Holler, Martin and Koesters, Thomas and Bredies, Kristian and Sodickson, Daniel K},
year = {2015},
month = {Oct},
pages = {1–4}
}
Optimal control problems play an important role in several areas of applied mathematics. The governing dynamical systems frequently involve parameters, and the resulting optimal control problem needs to be solved quickly for many different values of the parameters – for instance in a real-time or many-query context. Solving the exact optimal control problem is usually already costly for a single parameter. Hence, doing so for many values of the parameter is prohibitively expensive and infeasible in most applications.
We are interested in applying model order reduction combined with machine learning approaches. Such a combination has proven to achieve significant speedups while maintaining theoretical guarantees such as a posteriori error estimates. In particular, the certifications available for model order reduction methods, such as the reduced basis method, can be transferred to the machine learning prediction as well. This results in certified machine learning methods that are orders of magnitude faster than classical methods [KLM2025], [KR2025]. A purely data-driven method with certification using the high-fidelity model was developed in [KKLOO2022] to improve the results of enhanced oil recovery.
The aforementioned machine learning methods can be integrated in adaptive model hierarchies [HKOSW2023] consisting of a full order model, a reduced model and a machine learning surrogate. It is possible to apply different machine learning approaches [WHKOS2024] while maintaining the certification via the reduced basis approach. Such adaptive model hierarchies are automatically tailored to the parameters of interest and adjust dynamically depending on the performance of the individual components. In the context of parametrized optimal control problems, adaptive model hierarchies have for instance been applied in [K2024].
Most of the methods developed in this research topic are presented extensively in [K2025].
The methods developed in this research area contribute to Work Group 2 (ML for CT) of the COST Action InterCoML, in which we actively participate.
@misc{kleikamp2026data,
doi = {10.48550/ARXIV.2608.00082},
url = {https://arxiv.org/abs/2608.00082},
author = {Kleikamp, Hendrik and Mlinarić, Petar and Rave, Stephan and Schindler, Felix},
title = {Data-Driven Model Order Reduction with pyMOR},
publisher = {arXiv},
year = {2026}
}
@misc{kleikamp2026overcoming,
doi = {10.48550/ARXIV.2607.13226},
url = {https://arxiv.org/abs/2607.13226},
author = {Kleikamp, Hendrik and Lazar, Martin and Muñoz, Juan Ricardo},
title = {Overcoming slow Kolmogorov width decay in parametric optimal control via neural network surrogates},
publisher = {arXiv},
year = {2026}
}
@phdthesis{kleikamp2025parametrized,
author = {H. Kleikamp},
title = {Parametrized optimal control and transport-dominated problems reduced basis methods, nonlinear reduction strategies and data-driven surrogates},
school = {Universität Münster},
year = {2024}
}
@article{kleikamp2025twostage,
title = {Two-Stage Model Reduction Approaches for the Efficient and Certified Solution of Parametrized Optimal Control Problems},
volume = {104},
ISSN = {1573-7691},
url = {http://dx.doi.org/10.1007/s10915-025-02988-w},
DOI = {10.1007/s10915-025-02988-w},
number = {3},
journal = {Journal of Scientific Computing},
publisher = {Springer Science and Business Media LLC},
author = {Kleikamp, Hendrik and Renelt, Lukas},
year = {2025},
month = {July}
}
@article{kleikamp2025greedy,
title = {Be greedy and learn: efficient and certified algorithms for parametrized optimal control problems},
volume = {59},
ISSN = {2804-7214},
url = {http://dx.doi.org/10.1051/m2an/2024074},
DOI = {10.1051/m2an/2024074},
number = {1},
journal = {ESAIM: Mathematical Modelling and Numerical Analysis},
publisher = {EDP Sciences},
author = {Kleikamp, Hendrik and Lazar, Martin and Molinari, Cesare},
year = {2025},
month = {Jan},
pages = {291–330}
}
@article{haasdonk2023certified,
title = {A New Certified Hierarchical and Adaptive RB-ML-ROM Surrogate Model for Parametrized PDEs},
volume = {45},
ISSN = {1095-7197},
url = {http://dx.doi.org/10.1137/22M1493318},
DOI = {10.1137/22m1493318},
number = {3},
journal = {SIAM Journal on Scientific Computing},
publisher = {Society for Industrial & Applied Mathematics (SIAM)},
author = {Haasdonk, Bernard and Kleikamp, Hendrik and Ohlberger, Mario and Schindler, Felix and Wenzel, Tizian},
year = {2023},
month = {May},
pages = {A1039–A1065}
}
@inbook{wenzel2024application,
title = {Application of Deep Kernel Models for Certified and Adaptive RB-ML-ROM Surrogate Modeling},
ISBN = {9783031562082},
ISSN = {1611-3349},
url = {http://dx.doi.org/10.1007/978-3-031-56208-2_11},
DOI = {10.1007/978-3-031-56208-2_11},
booktitle = {Large-Scale Scientific Computations},
publisher = {Springer Nature Switzerland},
author = {Wenzel, Tizian and Haasdonk, Bernard and Kleikamp, Hendrik and Ohlberger, Mario and Schindler, Felix},
year = {2024},
pages = {117–125}
}
@inproceedings{kleikamp2024application,
author = {H. Kleikamp},
title = {Application of an adaptive model hierarchy to parametrized optimal control problems},
year = {2024},
url = {https://arxiv.org/abs/2402.10708}
}
@article{keil2022adaptive,
title = {Adaptive machine learning-based surrogate modeling to accelerate PDE-constrained optimization in enhanced oil recovery},
volume = {48},
ISSN = {1572-9044},
url = {http://dx.doi.org/10.1007/s10444-022-09981-z},
DOI = {10.1007/s10444-022-09981-z},
number = {6},
journal = {Advances in Computational Mathematics},
publisher = {Springer Science and Business Media LLC},
author = {Keil, Tim and Kleikamp, Hendrik and Lorentzen, Rolf J. and Oguntola, Micheal B. and Ohlberger, Mario},
year = {2022},
month = {Nov}
}