TY - GEN A1 - Buchholz, Annika A1 - Khebouri, Imene A1 - Vu, Thi Huong A1 - Koch, Thorsten A1 - Kunt, Tim A1 - Peters-Kottig, Wolfgang A1 - Stompor, Tomasz A1 - Zittel, Janina T1 - Detecting and classifying publications based on their abstracts with LLM embeddings and multi-label classifiers T3 - ZIB-Report - 26-04 Y1 - 2026 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:0297-zib-103273 SN - 1438-0064 ER - TY - THES A1 - Amiranashvili, Tamaz T1 - Universal and Expressive Statistical Shape Models for Anatomical Structures N2 - Form and function of anatomical structures are intimately linked. Pathological changes in form can be associated with the loss of function. For example, diseases often cause characteristic shape changes, making shape a sensitive structural biomarker for medical diagnosis. If the link between form and function is causal, correcting a pathological shape can even restore the healthy function of an organ. Accurate shape reconstruction is then crucial for effective, patient-specific treatment planning. This demonstrates the importance of shape in clinical interventions and its potential to improve overall patient outcomes. Statistical shape models are computational methods that capture shape variations in a given population and enable precise shape analysis and generation. We focus on two key properties of a good statistical shape model. First, it should be easy to construct, and second, it should accurately represent the underlying shape distribution. Established existing approaches can only be constructed from surfaces with pre-defined dense correspondence. Such correspondence is tedious to obtain, can introduce undesired biases, and prevents training on partial or sparse observations. While correspondence-free methods exist, they struggle to accurately capture shape distributions with intricate details and large variations. In this thesis, we develop shape models that simplify training and improve accuracy over state-of-the-art. To achieve these goals, we build on approximately diffeomorphic neural deformations and implicit neural representations. First, our proposed methods are trainable on correspondence-free surfaces and even partial segmentations with large slice distances. This makes them universal since they can be trained on heterogeneous data, enabling scalability to large datasets and avoiding potential biases of pre-defined correspondence. Second, our methods are highly expressive, accurately capturing intricate shape details in complex distributions. We evaluate effectiveness of our models on multiple anatomical structures, outperforming established baselines in both generative and discriminative settings. Y1 - 2025 UR - https://mediatum.ub.tum.de/doc/1776778/34rd9v23igjb2ahatohz1rwdx.phd_thesis_Amiranashvili.pdf ER - TY - JOUR A1 - Schaible, Jonas A1 - Özdemir, Asena Karolin A1 - Debus, Charlotte A1 - Burger, Sven A1 - Streit, Achim A1 - Becker, Christiane A1 - Jäger, Klaus A1 - Götz, Markus T1 - Inverse Design of Optical Multilayer Thin Films using Robust Masked Diffusion Models JF - ArXiV Y1 - 2026 U6 - https://doi.org/10.48550/arXiv.2604.01106 SP - arXiv:2604.01106 ER - TY - GEN A1 - Andrés Arcones, Daniel A1 - Paul, Aeneas A1 - Weiser, Martin A1 - Sanio, David A1 - Mark, Peter A1 - Unger, Jörg F. T1 - Bayesian Tendon Break Localization under Model Uncertainty Using Distributed Fiber Optic Sensors Authors/Creators N2 - Implementation, data and results of the paper "Bayesian Tendon Break Localization under Model Uncertainty Using Distributed Fiber Optic Sensors" Y1 - 2006 U6 - https://doi.org/10.5281/zenodo.18713386 ER - TY - JOUR A1 - Andrés Arcones, Daniel A1 - Paul, Aeneas A1 - Weiser, Martin A1 - Sanio, David A1 - Mark, Peter A1 - Unger, Jörg F. T1 - Bayesian Tendon Breakage Localization under Model Uncertainty Using Distributed Fiber Optic Sensors N2 - This study develops a Bayesian, uncertainty-aware framework for tendon breakage localization in pre-stressed concrete members using high-resolution data from distributed fiber-optic sensors (DFOS). DFOS enable full-field monitoring of strain changes on the surface of pre-stressed concrete members due to such failure. A finite element model (FEM) of an experimental tendon-breakage test is constructed, and model parameters are calibrated probabilistically against DFOS measurements. To capture model-form uncertainty (MFU), stochastic perturbations are embedded directly into material parameters, enabling the joint inference of physical properties and MFU within a unified probabilistic framework. Gaussian Process surrogates are employed to efficiently emulate the nonlinear FEM response, supporting computationally tractable Bayesian inference. A ϕ-divergence-based influence analysis identifies the DFOS measurements that most strongly shape the posterior distributions, providing interpretable diagnostics of sensor informativeness and model adequacy. The calibrated parameters and embedded uncertainties are then transferred to a FEM of a full-scale structural configuration, enabling prediction of tendon breakage localization under realistic conditions. A separability analysis of the predictive strain distributions quantifies the identifiability of tendon breakage at varying depths, assessing the confidence with which different damage scenarios can be distinguished given the propagated uncertainties. Results demonstrate that the framework achieves robust parameter calibration, interpretable diagnostics, and uncertainty-informed damage detection, integrating experimental data, embedded MFU, and probabilistic modeling. By systematically propagating both experimental and model uncertainties, the approach supports reliable tendon breakage localization and optimal DFOS placement. Y1 - 2026 ER - TY - JOUR A1 - Kang, Chongjie A1 - Andrés Arcones, Daniel A1 - Becks, Henrik A1 - Beetz, Jakob A1 - Blankenbach, Jörg A1 - Claßen, Martin A1 - Degener, Sebastian A1 - Eisermann, Cedric A1 - Göbels, Anne A1 - Hegger, Josef A1 - Herrmann, Ralf A1 - Kähler, Philipp A1 - Peralta, Patricia A1 - Petryna, Yuri A1 - Schnellenbach-Held, Martina A1 - Schulz, Oliver A1 - Smarsly, Kay A1 - Fatih Sönmez, Mehmet A1 - Sprenger, Bjarne A1 - Unger, Jörg F. A1 - Vassilev, Hristo A1 - Weiser, Martin A1 - Marx, Steffen T1 - Intelligente digitale Methoden zur Verlängerung der Nutzungsdauer der Nibelungenbrücke JF - Beton- und Stahlbetonbau N2 - Um die Lebensdauer von Bauwerken unter Wahrung derer Standsicherheit und Funktionsfähigkeit zu verlängern, bedarf es effektiver Monitorings- sowie Instandhaltungskonzepte. Im Rahmen des von der Deutschen Forschungsgemeinschaft (DFG) geförderten Schwerpunktprogramms 2388 „Hundert plus – Verlängerung der Lebensdauer komplexer Baustrukturen durch intelligente Digitalisierung“ (kurz: SPP 100+) werden hierfür innovative, interdisziplinäre Methoden entwickelt und an der Nibelungenbrücke in Worms (NBW) validiert. Der vorliegende Beitrag stellt einige dieser neuentwickelten digitalen Methoden vor. Unter anderem umfasst dies zwei Systeme des Structural Health Monitoring (SHM) und deren zielorientierte Verknüpfung von mehreren Beschleunigungsmessdaten zur umfassenden Zustandsbewertung. Ergänzend werden innovative datenbasierte Simulationsmethoden zur Bestimmung des Temperaturfelds des Brückenüberbaus vorgestellt sowie mehrere Finite-Elemente-Modelle unterschiedlicher Detailtiefe präsentiert und miteinander verglichen. Abschließend werden innovative Methoden zum Verwalten des Bestandswissens von Brückenbauwerken diskutiert. Die Methoden wurden überwiegend unabhängig voneinander entwickelt und an der NBW validiert. Im nächsten Schritt werden die Methoden integriert, um die Instandhaltung der NBW zu unterstützen. Y1 - 2026 U6 - https://doi.org/10.1002/best.70070 VL - 121 IS - 4 SP - 303 EP - 320 ER - TY - GEN A1 - Kuroiwa, Ryo A1 - Shinano, Yuji A1 - Beck, J. Christopher T1 - Massively Parallel and Distributed Solvers for Domain-Independent Dynamic Programming N2 - In this paper, we develop distributed and parallel general-purpose solvers for combinatorial optimization through the framework of domain-independent dynamic programming (DIDP), a model-based paradigm based on dynamic programming. In particular, we parallelize heuristic state space search algorithms to develop such solvers. Benefiting from the general-purpose nature of DIDP, we apply our solvers to four problem classes: the traveling salesperson problem with time windows (TSPTW), the type1 simple assembly line balancing problem (SALBP-1), the one-to-one multi-commodity pickup and delivery traveling salesperson problem (m-PDTSP), and the type2 assembly line balancing problem with sequence-dependent setup times (SUALBP-2). We demonstrate the scalability of our solvers using up to 192 TB of RAM and 49,152 CPU cores. Using the developed solvers, we close 14 open instances of TSPTW, 49 of m-PDTSP, and 152 of SUALBP-2. T3 - ZIB-Report - 26-03 KW - Dynamic Programming, Combinatorial Optimization, Massively Parallel and Distributed Algorithms Y1 - 2026 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:0297-zib-103202 SN - 1438-0064 ER - TY - CHAP A1 - Kempke, Nils-Christian A1 - Koch, Thorsten T1 - A GPU accelerated variant of Schroeppel-Shamir’s algorithm for solving the market split problem T2 - Operations Research Proceedings 2025 N2 - The market split problem (MSP), introduced by Cornu´ejols and Dawande (1998), is a challenging binary optimization problem on which state-of-the-art linear programming-based branch-and-cut solvers perform poorly. We present a novel algorithm for solving the feasibility version of this problem, derived from Schroeppel–Shamir’s algorithm for the one-dimensional subset sum problem. Our approach is based on exhaustively enumerating one-dimensional solutions of MSP and utilizing GPUs to evaluate candidate solutions across the entire problem. The resulting hybrid CPU-GPU implementation significantly outperforms a parallel CPU-only variant, efficiently solving instances with up to 10 constraints and 90 variables. We demonstrate the algorithm’s performance on benchmark problems, solving instances of size (9, 80) in less than fifteen minutes and (10, 90) in up to one day. Given our results, sorting based algorithms can be considered competitive for solving the MSP on modern hardware. Y1 - 2026 ER - TY - JOUR A1 - Kempke, Nils-Christian A1 - Koch, Thorsten T1 - Fix-and-Propagate Heuristics Using Low-Precision First-Order LP Solutions for Large-Scale Mixed-Integer Linear Optimization JF - Mathematical Programming Computation N2 - We investigate the use of low-precision first-order methods (FOMs) within a fix-and-propagate (FP) framework for solving mixed-integer programming problems (MIPs). We employ GPU-accelerated PDLP, a variant of the Primal-Dual Hybrid Gradient (PDHG) method specialized to LP problems, to solve the LP-relaxation of our MIPs to low accuracy. This solution is used to motivate fixings within our FP framework. We evaluate the performance of our heuristic on MIPLIB 2017, demonstrating that low-accuracy LP solutions do not lead to a loss in the quality of the FP heuristic solutions. Further, we use our FP framework to produce high-accuracy solutions for large-scale (up to 243 million nonzeros and 8 million decision variables) unit commitment-based dispatch and expansion planning problems created with the modeling framework REMix. For the largest problems, we can generate solutions with a primal-dual gap of under 2% in less than 4 hours, whereas state-of-the-art commercial solvers cannot produce feasible solutions within 2 days of runtime. KW - Integer programming KW - Large scale optimization KW - Linear Programming KW - OR in energy KW - Primal heuristics Y1 - 2026 U6 - https://doi.org/10.1007/s12532-026-00312-7 ER - TY - JOUR A1 - Sikorski, Alexander A1 - Donati, Luca A1 - Weber, Marcus A1 - Schütte, Christof T1 - Effective Dynamics and Transition Pathways from Koopman-Inspired Neural Learning of Collective Variables N2 - The ISOKANN (Invariant Subspaces of Koopman Operators Learned by Artificial Neural Networks) framework provides a data-driven route to extract collective variables (CVs) and effective dynamics from complex molecular systems. In this work, we integrate the theoretical foundation of Koopman operators with Krylov-like subspace algorithms, and reduced dynamical modeling to build a coherent picture of how to describe metastable transitions in high-dimensional systems based on CVs. Starting from the identification of CVs based on dominant invariant subspaces, we derive the corresponding effective dynamics on the latent space and connect these to transition rates and times, committor functions, and transition pathways. The combination of Koopman-based learning and reduced-dimensional effective dynamics yields a principled framework for computing transition rates and pathways from simulation data. Numerical experiments on one-, two-, and three-dimensional benchmark potentials illustrate the ability of ISOKANN to reconstruct the coarse-grained kinetics and reproduce transition times across enthalpic and entropic barriers. Y1 - 2026 ER -