arXiv subjects

Lisa Kusch

Publications and source records attributed to Lisa Kusch.

10 recordsLinked to original sources

arXiv2026

Direct Optimization of a 3D Finite-Source Reflector via Neural-Network Parameterization

We present a direct optimization method for three-dimensional freeform reflectors that transform the light of a finite-\'etendue source into a prescribed far-field angular intensity distribution. The reflector profile is represented by a small neural network (a multilayer perceptron), which is trained end-to-end through a differentiable ray-tracing objective. We furthermore parameterize the emission directions in gnomonic coordinates, and show how we use this to ensure that every emitted ray intersects the reflector. At each iteration, the network is converted to a bicubic spline representation for ray-tracing efficiency, and intersections with this smooth surface are solved by a damped Newton solve, with gradients computed via the implicit function theorem. The traced output distribution is compared with the desired target on a 'soft' histogram, under an $H^{-1}$-type spectral weighting that emphasizes long-range transport of flux to improve convergence. Optimization is performed using a BFGS method with self-scaled Broyden updates and a plateau-perturbation rule to prevent stalling. The method converges reliably within seconds on a single GPU for all examples tested.

physics.optics↗

arXiv2026

One-Shot Optimization with Additional Inequality Constraints

The one-shot approach is a powerful simultaneous optimization framework for design tasks governed by computationally expensive steady-state systems. While previous formulations mainly focused on additional equality constraints, this work extends the one-shot framework to optimization problems with inequality constraints using slack variables embedded into a doubly augmented Lagrangian formulation. After elimination of the slack variables, the resulting formulation contains nonsmooth active-set dependent terms while preserving the characteristic coupled one-shot matrix structure. The resulting generalized gradient system admits a representation of the form $ \nabla \DLag = -Ms, $ analogous to the equality-constrained one-shot framework. To analyze the nonsmooth active-set transitions, generalized second-order subdifferentials in the sense of Clarke and Rockafellar--Wets are employed. We derive explicit positivity conditions ensuring positive definiteness of all admissible generalized Hessians and hence strict local optimality. Furthermore, we give an outlook on the construction of preconditioners based on generalized Hessian approximations.

math.OC↗

arXiv2026

Neural-network methods for two-dimensional finite-source reflector design

We address the inverse problem of designing two-dimensional reflectors that transform light from a finite, extended source into a prescribed far-field distribution. The reflector height is represented by a neural network and optimized with two objective functions: a direct change-of-variables loss based on the closed-form inverse ray map, and a mesh-based loss that maps target cells back to the source and remains usable for discontinuous sources. Gradients are computed by automatic differentiation and minimized with a robust quasi-Newton method. As a baseline, we adapt a deconvolution pipeline built on a simplified finite-source approximation: a one-dimensional monotone map is recovered from flux balance, converted to a reflector by an integrating-factor ODE solve, and embedded in a modified Van Cittert iteration with nonnegativity clipping and ray-traced feedback. Across four benchmarks, covering continuous and discontinuous sources and minimum-height constraints, accuracy is measured by ray-traced normalized mean absolute error. On the two main benchmarks, the neural method reaches errors of about 2e-5 and 5e-5 within a few seconds on one NVIDIA RTX 4090 GPU, compared with 4e-3 and 5e-2 for the deconvolution baseline after several hundred seconds. The results show that the neural formulation is both more accurate and substantially faster, while still supporting practical height constraints. We also discuss extensions to rotationally symmetric and full three-dimensional reflector design through iterative correction schemes.

cs.LG↗

arXiv2025

An inverse design method for generalized zero-\'etendue sources and two targets

We present an inverse method to compute freeform optical surfaces that transform a light distribution, parameterized by two source planes, into two separate target distributions. The surfaces can be reflectors or lenses, and control both the spatial and directional source and target coordinates of light rays. From energy conservation we derive Jacobian equations for optical mappings, and the optical path length provides generating functions for the optical surfaces. A three-stage least-squares algorithm numerically solves the resulting equations. We present examples with complex source and target distributions.

physics.optics↗

arXiv2025

Design of a three-dimensional parallel-to-point imaging system based on inverse methods

We present an inverse method for designing a three-dimensional imaging system comprising of freeform optical surfaces. We impose an imaging condition on the optical map and combine it with the law of conservation of energy to conclude that the ratio of energy distributions at the source and target of an imaging system must be constant. A mathematical model for the design of a parallel-to-point system consisting of two freeform reflectors is presented. A Schwarzschild telescope, a classical design known for maximum correction of third-order aberrations, is utilized to specify the design parameters in the mathematical model, enabling us to compute an inverse freeform imaging system. The performance of both designs is compared by ray tracing various parallel beams of light and determining the corresponding spot sizes of the image. We demonstrate that our inverse freeform design is superior to the classical design.

physics.optics↗

arXiv2025

An Inverse Method for the Design of Freeform Double-Reflector Imaging Systems

We propose an inverse method to design two-dimensional freeform imaging systems. We present the mathematical model to design a parallel-to-point double-reflector imaging system using inverse methods from nonimaging optics. We impose an imaging condition on the energy distributions at the source and target of the optical system. Our freeform design is compared to the classical Schwarzschild telescope, which is well-known for minimizing third-order aberrations. A raytracer using quasi-interpolation is employed to test the performance of both designs by comparing the spot sizes corresponding to on-axis and off-axis light rays. We show that the inverse freeform design outperforms the classical design.

physics.optics↗

arXiv2024

A neural network approach for solving the Monge-Amp\`ere equation with transport boundary condition

This paper introduces a novel neural network-based approach to solving the Monge-Amp\`ere equation with the transport boundary condition, specifically targeted towards optical design applications. We leverage multilayer perceptron networks to learn approximate solutions by minimizing a loss function that encompasses the equation's residual, boundary conditions, and convexity constraints. Our main results demonstrate the efficacy of this method, optimized using L-BFGS, through a series of test cases encompassing symmetric and asymmetric circle-to-circle, square-to-circle, and circle-to-flower reflector mapping problems. Comparative analysis with a conventional least-squares finite-difference solver reveals the competitive, and often superior, performance of our neural network approach on the test cases examined here. A comprehensive hyperparameter study further illuminates the impact of factors such as sampling density, network architecture, and optimization algorithm. While promising, further investigation is needed to verify the method's robustness for more complicated problems and to ensure consistent convergence. Nonetheless, the simplicity and adaptability of this neural network-based approach position it as a compelling alternative to specialized partial differential equation solvers.

cs.LG↗

arXiv2023

Data-driven aerodynamic shape design with distributionally robust optimization approaches

We formulate and solve data-driven aerodynamic shape design problems with distributionally robust optimization (DRO) approaches. Building on the findings of the work \cite{gotoh2018robust}, we study the connections between a class of DRO and the Taguchi method in the context of robust design optimization. Our preliminary computational experiments on aerodynamic shape optimization in transonic turbulent flow show promising design results.

cs.CE↗

arXiv2023

Progress in End-to-End Optimization of Detectors for Fundamental Physics with Differentiable Programming

In this article we examine recent developments in the research area concerning the creation of end-to-end models for the complete optimization of measuring instruments. The models we consider rely on differentiable programming methods and on the specification of a software pipeline including all factors impacting performance -- from the data-generating processes to their reconstruction and the extraction of inference on the parameters of interest of a measuring instrument -- along with the careful specification of a utility function well aligned with the end goals of the experiment. Building on previous studies originated within the MODE Collaboration, we focus specifically on applications involving instruments for particle physics experimentation, as well as industrial and medical applications that share the detection of radiation as their data-generating mechanism.

physics.ins-det↗

arXiv2022

Exploration of Differentiability in a Proton Computed Tomography Simulation Framework

Objective. Algorithmic differentiation (AD) can be a useful technique to numerically optimize design and algorithmic parameters by, and quantify uncertainties in, computer simulations. However, the effectiveness of AD depends on how "well-linearizable" the software is. In this study, we assess how promising derivative information of a typical proton computed tomography (pCT) scan computer simulation is for the aforementioned applications. Approach. This study is mainly based on numerical experiments, in which we repeatedly evaluate three representative computational steps with perturbed input values. We support our observations with a review of the algorithmic steps and arithmetic operations performed by the software, using debugging techniques. Main results. The model-based iterative reconstruction (MBIR) subprocedure (at the end of the software pipeline) and the Monte Carlo (MC) simulation (at the beginning) were piecewise differentiable. Jumps in the MBIR function arose from the discrete computation of the set of voxels intersected by a proton path. Jumps in the MC function likely arose from changes in the control flow that affect the amount of consumed random numbers. The tracking algorithm solves an inherently non-differentiable problem. Significance. The MC and MBIR codes are ready for the integration of AD, and further research on surrogate models for the tracking subprocedure is necessary.

physics.med-ph↗