Nicolas Ralph Gauger
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Nicolas Ralph Gauger has a total of 212 co-authors. Here are the top co-authors by number of shared publications:
- Max Aehle
RPTU University Kaiserslautern-Landau, Germany
12 shared publications - Ralf Keidel
RPTU University Kaiserslautern-Landau
11 shared publications - Long Chen
University of Kaiserslautern-Landau (RPTU)
10 shared publications - Tobias Kortus
RPTU University Kaiserslautern-Landau
8 shared publications - Johannes Blühdorn
RPTU University Kaiserslautern-Landau, Germany
7 shared publications - Tommaso Dorigo
INFN Sezione di Padova
7 shared publications
Latest Publications
- Open Access
- 10 November 2026
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 71 (21) , 215023
Experimental feasibility study of proton therapy range verification by a digital tracking calorimeter prototype
Objective. Uncertainties in the proton range remain an important challenge in proton therapy. In-situ range verification (RV) techniques can be used to ensure the Bragg peak depth in the patient matches the treatment plan. Various such methods are under investigation, utilizing the signal from secondary radiation caused by proton interactions with the patient. This study experimentally investigates the feasibility of using a silicon-pixel-based sampling calorimeter for RV through the detection of secondary photons and neutrons. Approach. The proton RV capability of a sampling calorimeter based on monolithic active pixel sensors was tested both theoretically and experimentally. We theoretically investigate the plausibility of the planned sampling calorimeter to pick up a signal from the neutral secondary particles distal to the patient. Then, a smaller prototype pixel telescope was constructed to verify our theoretical findings experimentally by placing it behind a water tank irradiated by a medical proton beam with various energies in the range of 100–200 MeV at the Danish Center for Particle Therapy, Aarhus, Denmark. Main results. We find that the aluminum used as absorbers between sensitive layers in the calorimeter serves as a converter for prompt gammas and neutrons. The combined detection efficiency ranges between 5.8 ⋅ 10 − 5 and 2.7 ⋅ 10 − 3 detector hits per primary proton for beam energies between 50 and 200 MeV. Irradiation of the water phantom leads to a signal in our detector, which is well-correlated with the nominal beam energy. A simulation-based model using only the number of hits per proton can predict the beam energy, acting as a proxy for proton range in a homogeneous target, down to a root mean squared error of 1.34 MeV. Significance. This study marks a promising step towards utilizing a tracking calorimeter designed for proton computed tomography for the secondary application of proton therapy RV without additional hardware costs.
- 10 November 2026
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 71 (21) , 215023
Open AccessExperimental feasibility study of proton therapy range verification by a digital tracking calorimeter prototype
Objective. Uncertainties in the proton range remain an important challenge in proton therapy. In-situ range verification (RV) techniques can be used to ensure the Bragg peak depth in the patient matches the treatment plan. Various such methods are under investigation, utilizing the signal from secondary radiation caused by proton interactions with the patient. This study experimentally investigates the feasibility of using a silicon-pixel-based sampling calorimeter for RV through the detection of secondary photons and neutrons. Approach. The proton RV capability of a sampling calorimeter based on monolithic active pixel sensors was tested both theoretically and experimentally. We theoretically investigate the plausibility of the planned sampling calorimeter to pick up a signal from the neutral secondary particles distal to the patient. Then, a smaller prototype pixel telescope was constructed to verify our theoretical findings experimentally by placing it behind a water tank irradiated by a medical proton beam with various energies in the range of 100–200 MeV at the Danish Center for Particle Therapy, Aarhus, Denmark. Main results. We find that the aluminum used as absorbers between sensitive layers in the calorimeter serves as a converter for prompt gammas and neutrons. The combined detection efficiency ranges between 5.8 ⋅ 10 − 5 and 2.7 ⋅ 10 − 3 detector hits per primary proton for beam energies between 50 and 200 MeV. Irradiation of the water phantom leads to a signal in our detector, which is well-correlated with the nominal beam energy. A simulation-based model using only the number of hits per proton can predict the beam energy, acting as a proxy for proton range in a homogeneous target, down to a root mean squared error of 1.34 MeV. Significance. This study marks a promising step towards utilizing a tracking calorimeter designed for proton computed tomography for the secondary application of proton therapy RV without additional hardware costs.
- Open Access
- 14 October 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Mathematical Software
Backpropagation Through Compiled Code
Algorithmic differentiation (AD) tools provide accurate derivatives of computer-implemented functions, for use in gradient-based optimization tasks like the training of neural networks or the improvement of computational engineering designs. Typically, AD tools require the source code to be available in a restricted set of programming languages, and often necessitate manual modifications because the language standard is only partially supported. In contrast, Derivgrind , a novel AD tool that inserts forward-mode differentiation logic on the machine code level via dynamic binary instrumentation, supports an unprecedented range of programming languages with minimal manual integration efforts. In this work, we extend Derivgrind with tape-recording logic, enabling reverse-mode AD for compiled code.
- 14 October 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Mathematical Software
Open AccessBackpropagation Through Compiled Code
Algorithmic differentiation (AD) tools provide accurate derivatives of computer-implemented functions, for use in gradient-based optimization tasks like the training of neural networks or the improvement of computational engineering designs. Typically, AD tools require the source code to be available in a restricted set of programming languages, and often necessitate manual modifications because the language standard is only partially supported. In contrast, Derivgrind , a novel AD tool that inserts forward-mode differentiation logic on the machine code level via dynamic binary instrumentation, supports an unprecedented range of programming languages with minimal manual integration efforts. In this work, we extend Derivgrind with tape-recording logic, enabling reverse-mode AD for compiled code.
- Open Access
- 1 April 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (2) , 025061
Differentiable surrogate for detector simulation and design with diffusion models
In this work, we present a conditional denoising-diffusion surrogate for electromagnetic calorimeter showers that is trained to generate high-fidelity energy-deposition maps conditioned on key detector and beam parameters. The model employs efficient inference using denoising diffusion implicit model sampling and is pre-trained on GEANT4 simulations before being adapted to a new calorimeter geometry through low-rank adaptation, requiring only a small post-training dataset. We evaluate physically meaningful observables, including total deposited energy, energy-weighted radius, and shower dispersion, obtaining relative root mean square error values below 2 % for representative high-energy cases. This is in line with state-of-the-art calorimeter surrogates which report comparable fidelity on high-level observables. Furthermore, we compare gradients of a reconstruction-based utility function with respect to design parameters between the surrogate and finite-difference references. The diffusion surrogate reproduces the qualitative structure and directional trends of the true utility landscape, providing usable sensitivities for gradient-based optimization. These results show that diffusion-based surrogates can accelerate simulation-driven detector design while enabling differentiable, gradient-informed analysis.
- 1 April 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (2) , 025061
Open AccessDifferentiable surrogate for detector simulation and design with diffusion models
In this work, we present a conditional denoising-diffusion surrogate for electromagnetic calorimeter showers that is trained to generate high-fidelity energy-deposition maps conditioned on key detector and beam parameters. The model employs efficient inference using denoising diffusion implicit model sampling and is pre-trained on GEANT4 simulations before being adapted to a new calorimeter geometry through low-rank adaptation, requiring only a small post-training dataset. We evaluate physically meaningful observables, including total deposited energy, energy-weighted radius, and shower dispersion, obtaining relative root mean square error values below 2 % for representative high-energy cases. This is in line with state-of-the-art calorimeter surrogates which report comparable fidelity on high-level observables. Furthermore, we compare gradients of a reconstruction-based utility function with respect to design parameters between the surrogate and finite-difference references. The diffusion surrogate reproduces the qualitative structure and directional trends of the true utility landscape, providing usable sensitivities for gradient-based optimization. These results show that diffusion-based surrogates can accelerate simulation-driven detector design while enabling differentiable, gradient-informed analysis.
- Open Access
- 1 April 2026
- Published by Elsevier in SSRN Electronic Journal
Generation of efficient adjoint lattice Boltzmann methods with algorithmic differentiation
Adjoint-based optimization is widely used to address large-scale flow control problems with distributed control variables, as the computational cost of gradient evaluation is independent of the dimension of the control space. Within this context, the lattice Boltzmann method (LBM) represents an attractive discretization scheme, as it not only reduces the computational expense within optimization iteration loops but also exposes Jacobian expressions through its explicit operator-split formulation, thereby simplifying adjoint analysis. Despite these advantages, most existing adjoint LBM approaches rely on manual derivation of the adjoint system, lack automation, and are highly case-specific. High-level frameworks based on automatic differentiation (AD) address this limitation by enabling generic gradient computation but often compromise numerical performance. In this work, we present a novel framework implemented in the open-source LBM library OpenLB that enables the automated generation of discrete adjoint LBM collision kernels. AD is applied locally to evaluate the Jacobians of the adjoint system, eliminating manual derivations. Code generation combined with common subexpression elimination (CSE) removes the runtime overhead of AD tapes, allowing efficient adjoint collision kernels to be generated from their primal implementations. We conduct a detailed kernel-level performance analysis that is rarely addressed in the literature, where the arithmetically preferred AD mode for adjoint LBM kernels is identified within the proposed approach, and roofline analysis evaluates performance rather than strong and weak scaling tests, which primarily reflect communication overhead. The impact of CSE is investigated across different hardware architectures and floating-point precisions, demonstrating speedups of up to a factor of seven for adjoint collision kernels for double precision on GPUs.
- 1 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessGeneration of efficient adjoint lattice Boltzmann methods with algorithmic differentiation
Adjoint-based optimization is widely used to address large-scale flow control problems with distributed control variables, as the computational cost of gradient evaluation is independent of the dimension of the control space. Within this context, the lattice Boltzmann method (LBM) represents an attractive discretization scheme, as it not only reduces the computational expense within optimization iteration loops but also exposes Jacobian expressions through its explicit operator-split formulation, thereby simplifying adjoint analysis. Despite these advantages, most existing adjoint LBM approaches rely on manual derivation of the adjoint system, lack automation, and are highly case-specific. High-level frameworks based on automatic differentiation (AD) address this limitation by enabling generic gradient computation but often compromise numerical performance. In this work, we present a novel framework implemented in the open-source LBM library OpenLB that enables the automated generation of discrete adjoint LBM collision kernels. AD is applied locally to evaluate the Jacobians of the adjoint system, eliminating manual derivations. Code generation combined with common subexpression elimination (CSE) removes the runtime overhead of AD tapes, allowing efficient adjoint collision kernels to be generated from their primal implementations. We conduct a detailed kernel-level performance analysis that is rarely addressed in the literature, where the arithmetically preferred AD mode for adjoint LBM kernels is identified within the proposed approach, and roofline analysis evaluates performance rather than strong and weak scaling tests, which primarily reflect communication overhead. The impact of CSE is investigated across different hardware architectures and floating-point precisions, demonstrating speedups of up to a factor of seven for adjoint collision kernels for double precision on GPUs.
- 25 March 2026
- Published by Elsevier in Computer Methods in Applied Mechanics and Engineering
- Vol. 456
Refined stability estimates for mixed problems by exploiting semi norm arguments
- 25 March 2026
- Published by Elsevier in Computer Methods in Applied Mechanics and Engineering
- Vol. 456
Refined stability estimates for mixed problems by exploiting semi norm arguments
- 4 March 2026
- Published by Taylor & Francis in Optimization Methods and Software
- Vol. 41 (2) , 277-278
Preface for the special issue honouring Andreas Griewank
This special issue is dedicated to the memory of Andreas Griewank (1950–2021), an extraordinary mathematician whose work has profoundly shaped modern optimization and scientific computing. With his...
- 4 March 2026
- Published by Taylor & Francis in Optimization Methods and Software
- Vol. 41 (2) , 277-278
Preface for the special issue honouring Andreas Griewank
This special issue is dedicated to the memory of Andreas Griewank (1950–2021), an extraordinary mathematician whose work has profoundly shaped modern optimization and scientific computing. With his...
- 24 February 2026
- Published by Springer Nature
- p. 295-304
Parametric Aerodynamic Shape Optimization with Latent Diffusion
- 24 February 2026
- Published by Springer Nature
- p. 295-304
Parametric Aerodynamic Shape Optimization with Latent Diffusion
- 24 February 2026
- Published by Springer Nature
- p. 348-357
Sobolev Learning for Bayesian Neural Network Assisted Aerodynamic Shape Optimization
- 24 February 2026
- Published by Springer Nature
- p. 348-357
Sobolev Learning for Bayesian Neural Network Assisted Aerodynamic Shape Optimization
- Open Access
- 7 January 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (1)
Constrained collaborative optimization of charged particle tracking with multi-agent reinforcement learning
Reinforcement learning demonstrated immense success in modeling complex physics-driven systems, providing end-to-end trainable solutions by interacting with a simulated or real environment, maximizing a scalar reward signal. In this work, we propose, building upon previous work, an end-to-end multi-agent reinforcement learning approach with assignment constraints for reconstructing particle tracks in pixelated particle detectors. Our approach optimizes collaboratively a parametrized policy, functioning as a heuristic to a multidimensional assignment problem, by jointly minimizing the total amount of particle scattering over the reconstructed tracks in a readout frame. To satisfy constraints, guaranteeing a unique assignment of particle hits, we propose a safety layer solving a linear assignment problem for every joint action. Further, to enforce cost margins, increasing the distance of the local policies predictions to the decision boundaries of the optimizer mappings, we recommend the use of an additional component in the blackbox gradient estimation, forcing the policy to solutions with lower total assignment costs. We empirically show on simulated data, generated for a particle detector developed for proton imaging, the effectiveness of our approach, compared to multiple single-and multi-agent baselines. We further demonstrate the effectiveness of constraints with cost margins for both optimization and generalization, introduced by wider regions with high reconstruction performance as well as reduced predictive instabilities. Our results form the basis for further developments in RL-based tracking, offering both enhanced performance with constrained policies and greater flexibility in optimizing tracking algorithms through the option for individual and team rewards.
- 7 January 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (1)
Open AccessConstrained collaborative optimization of charged particle tracking with multi-agent reinforcement learning
Reinforcement learning demonstrated immense success in modeling complex physics-driven systems, providing end-to-end trainable solutions by interacting with a simulated or real environment, maximizing a scalar reward signal. In this work, we propose, building upon previous work, an end-to-end multi-agent reinforcement learning approach with assignment constraints for reconstructing particle tracks in pixelated particle detectors. Our approach optimizes collaboratively a parametrized policy, functioning as a heuristic to a multidimensional assignment problem, by jointly minimizing the total amount of particle scattering over the reconstructed tracks in a readout frame. To satisfy constraints, guaranteeing a unique assignment of particle hits, we propose a safety layer solving a linear assignment problem for every joint action. Further, to enforce cost margins, increasing the distance of the local policies predictions to the decision boundaries of the optimizer mappings, we recommend the use of an additional component in the blackbox gradient estimation, forcing the policy to solutions with lower total assignment costs. We empirically show on simulated data, generated for a particle detector developed for proton imaging, the effectiveness of our approach, compared to multiple single-and multi-agent baselines. We further demonstrate the effectiveness of constraints with cost margins for both optimization and generalization, introduced by wider regions with high reconstruction performance as well as reduced predictive instabilities. Our results form the basis for further developments in RL-based tracking, offering both enhanced performance with constrained policies and greater flexibility in optimizing tracking algorithms through the option for individual and team rewards.
- Open Access
- 17 December 2025
- Published by Springer Nature in Journal of Optimization Theory and Applications
- Vol. 208 (1) , 63
Stochastic Dynamical Low-Rank Approximation in the Context of Machine Learning
The central challenges of today’s neural network architectures are the prohibitive memory footprint and training costs associated with determining optimal weights and biases. A large portion of research in machine learning is therefore dedicated to constructing memory-efficient training methods. One promising approach is dynamical low-rank training (DLRT), which represents and trains parameters as a low-rank factorization. While DLRT is equipped with several beneficial properties, analytic results are currently limited to deterministic gradient flows. In this work, we show that dynamical low-rank training, in combination with stochastic gradient and momentum methods, fulfills certain robustness and descent guarantees. Moreover, we prove convergence to stationary points given the assumption that basis updates are omitted after a certain number of iterations. This agrees with the main numerical behavior observed in applications of this method.
- 17 December 2025
- Published by Springer Nature in Journal of Optimization Theory and Applications
- Vol. 208 (1) , 63
Open AccessStochastic Dynamical Low-Rank Approximation in the Context of Machine Learning
The central challenges of today’s neural network architectures are the prohibitive memory footprint and training costs associated with determining optimal weights and biases. A large portion of research in machine learning is therefore dedicated to constructing memory-efficient training methods. One promising approach is dynamical low-rank training (DLRT), which represents and trains parameters as a low-rank factorization. While DLRT is equipped with several beneficial properties, analytic results are currently limited to deterministic gradient flows. In this work, we show that dynamical low-rank training, in combination with stochastic gradient and momentum methods, fulfills certain robustness and descent guarantees. Moreover, we prove convergence to stationary points given the assumption that basis updates are omitted after a certain number of iterations. This agrees with the main numerical behavior observed in applications of this method.