Searcharxiv⌕ Search

arXiv subjects

Noah Schwartz

Publications and source records attributed to Noah Schwartz.

18 recordsLinked to original sources
arXiv2026

Renormalization-guided inverse blocking for lattice field generation: construction and validation

We propose an algorithm for generating lattice field configurations based on the approximate inversion of a renormalization-group blocking transformation. We optimize the blocking transformation using a ``perfect blocking'' condition so that the blocked lattice distribution is well approximated by a simple coarse action. The blocking is separated into an invertible smoothing transformation followed by decimation. Machine learning, in the form of a conditional normalizing flow, is used to reconstruct the short-distance degrees of freedom removed by the decimation. A short fine-action rethermalization then removes the residual mismatch. Because the coarse ensemble supplies the long-distance modes, the same blocking transformation and conditional flow can be reused recursively on larger lattices, producing a cascade of configurations from an initial small-volume ensemble. We test the method in two-dimensional $ϕ^4$ theory with $λ=1$ at criticality and demonstrate stable cascade upscaling from $16^2$ to $2048^2$ lattices on local computational resources. Controlled rethermalization tests show that short-distance mismatches relax rapidly, whereas a deliberately introduced mismatch in the relevant thermal direction relaxes much more slowly. The construction uses ingredients that admit natural extensions to higher-dimensional systems and, ultimately, to gauge and fermionic degrees of freedom.

hep-lat↗
arXiv2026

Renormalization-guided cascade upscaling for lattice field generation

We introduce a renormalization-group (RG) guided machine-learning algorithm for lattice field generation based on approximate inversion of an RG transformation. A ``perfect blocking'' construction supplies equilibrated long-distance modes, while a conditional normalizing flow reconstructs short-distance details and brief rethermalization removes residual errors. In 2D $ϕ^4$ theory at criticality, a flow trained at $L\le32$ is reused recursively in cascades reaching $L=2048$ with correct long-distance physics.

hep-lat↗
arXiv2026

Full frame denoising for pyramid wavefront sensors

Adaptive optics systems operating under low-flux conditions face significant challenges, as photon and detector noise in particular degrade wavefront measurements and ultimately limit correction performance. While pyramid wavefront sensors (PyWFSs) offer greater sensitivity than conventional wavefront sensors such as the Shack-Hartmann sensor in many operating regimes, obtaining accurate wavefront estimates under photon-starved conditions remains a key challenge. We present a full-frame image denoising strategy for the PyWFS that exploits the nonlocal self-similarity of wavefront sensor image patches through FFT-accelerated patch grouping and global collaborative 3D wavelet filtering, applied directly to the raw PyWFS intensity frame prior to slope computation. Specifically, the method suppresses noise while preserving structural features required for accurate wavefront reconstruction. The approach is evaluated using end-to-end simulations of a VLT-scale SCAO system. The results show improved performance in low signal-to-noise regimes, with typical Strehl ratio gains of up to 12% and an increase in limiting magnitude of approximately 0.5 in median seeing conditions. Modal analysis indicates reduced variance across most controlled modes. The improved PSF quality enables a reduction in the FWHM and an enhanced contrast. These results demonstrate that image-domain denoising can improve the robustness of PyWFS-based AO systems and extend their operational range toward fainter guide stars.

physics.optics↗
arXiv2026

Optimized Piecewise Affine Abstractions of Neural Networks with Learnable Activation Functions

We present a generalized framework for the range verification of neural networks featuring non-linear activation functions. Our approach first constructs an ``optimized piecewise affine abstraction" of the network that replaces each non-linear activation function by a piecewise affine (PWA) function plus a bounded error. Such PWA functions are readily amenable to existing neural network verification techniques using specializations of linear arithmetic SMT solvers and mixed-integer optimization approaches. However, there are infinitely many ways to abstract each node, with a natural tradeoff between the number of pieces used, the global error bound, and the complexity of the resulting verification problem. We propose a dynamic programming (DP) algorithm to systematically compute the optimized PWA abstraction for general activation functions, guaranteeing tighter output bounds. The algorithm combines a local DP approximation at each node with a global error bound, yielding a variant of the knapsack problem for deciding how to allocate a fixed budget on the total number of pieces across units so as to minimize the worst-case error bound between the network and its approximation. Although the knapsack problem is itself NP-hard, we can use pseudo-polynomial DP algorithms as well as approximation schemes to solve it efficiently. Crucially, our approach is broadly applicable to diverse networks consisting of non-linear activations, including standard Multi-Layer Perceptrons (MLPs) and recently proposed architectures such as Kolmogorov-Arnold Networks (KANs). Over a series of KAN benchmarks spanning 20 to 22,000 parameters, our approach yields output bounds that are consistently of smaller width than uniform PWA allocation. The overall time taken is roughly comparable while the overhead for computing the optimized abstraction is subsumed by the time taken to compute output bounds.

cs.LG↗
arXiv2026

DM/WFS mis-registration tracking: Implementation and on-sky validation of SPRINT at LBT

The advent of telescopes with an integrated deformable mirror (DM) presents new challenges for adaptive optics (AO) systems. The alignment between the DM and wavefront sensor (WFS) is expected to regularly evolve during operations due to their large separation. Without tracking and correction, these mis-registrations between the DM and WFS lead to loop instability, preventing diffraction limited performance from being realised. SPRINT\cite{heritier2021} provides an approach to track these mis-registrations during observations. Rotation, shift, and magnification mis-registrations can all be recovered. The Large Binocular Telescope (LBT) currently lacks an operational solution for tracking these mis-registrations, while SPRINT has been selected as the baseline approach for several instruments on the forthcoming Extremely Large Telescope (ELT). We report on the implementation of SPRINT into the LBT real time computer and present experimental results from both daytime and on-sky testing to validate the method.

astro-ph.IM↗
arXiv2026

Exploring generative design AI tools for astronomical instrumentation: a CubeSat chassis case study

Generative design artificial intelligence (AI) tools are currently used in multiple scientific fields, yet their adoption in mechanical engineering computer-aided design (CAD) remains limited due to a lack of disseminated case studies, limited availability of accessible tools, insufficient training in CAD data, the absence of universal editable file formats and more. Mechanical design for astronomical instrumentation faces increasing complexity in thermal, vibrational, and mechanical requirements alongside tight project deadlines. This paper presents a practical evaluation of an AI and FEA based generative design tool applied to chassis design for the Active Deployable Optical Telescope (ADOT) CubeSat mission. Our analysis showcases the workflow steps including the setting of design, manufacturing and objective constraints. This study also shows the clear benefits of these types of tools, especially in the early brainstorming stages of multi-constrained mechanical structures, while also highlighting clear limitations like their black-box nature, the non-manufacturing-ready state of the results, and the time-consuming setup, limiting the tangible gain of these tools to high-value mechanical components.

astro-ph.IM↗
arXiv2024

Spectral GRID invariants and Lagrangian cobordisms

We prove that the filtered GRID invariants of Legendrian links in link Floer homology, and consequently their associated invariants in the spectral sequence, obstruct decomposable Lagrangian cobordisms in the symplectization of the standard contact structure on $\mathbb{R}^3$, strengthening a result by Baldwin, Lidman, and the fifth author.

math.GT↗
arXiv2024

Paved2Paradise: Cost-Effective and Scalable LiDAR Simulation by Factoring the Real World

To achieve strong real world performance, neural networks must be trained on large, diverse datasets; however, obtaining and annotating such datasets is costly and time-consuming, particularly for 3D point clouds. In this paper, we describe Paved2Paradise, a simple, cost-effective approach for generating fully labeled, diverse, and realistic lidar datasets from scratch, all while requiring minimal human annotation. Our key insight is that, by deliberately collecting separate "background" and "object" datasets (i.e., "factoring the real world"), we can intelligently combine them to produce a combinatorially large and diverse training set. The Paved2Paradise pipeline thus consists of four steps: (1) collecting copious background data, (2) recording individuals from the desired object class(es) performing different behaviors in an isolated environment (like a parking lot), (3) bootstrapping labels for the object dataset, and (4) generating samples by placing objects at arbitrary locations in backgrounds. To demonstrate the utility of Paved2Paradise, we generated synthetic datasets for two tasks: (1) human detection in orchards (a task for which no public data exists) and (2) pedestrian detection in urban environments. Qualitatively, we find that a model trained exclusively on Paved2Paradise synthetic data is highly effective at detecting humans in orchards, including when individuals are heavily occluded by tree branches. Quantitatively, a model trained on Paved2Paradise data that sources backgrounds from KITTI performs comparably to a model trained on the actual dataset. These results suggest the Paved2Paradise synthetic data pipeline can help accelerate point cloud model development in sectors where acquiring lidar datasets has previously been cost-prohibitive.

cs.CV↗
arXiv2024

Phasing segmented telescopes via deep learning methods: application to a deployable CubeSat

Capturing high resolution imagery of the Earth's surface often calls for a telescope of considerable size, even from Low Earth Orbits (LEO). A large aperture often requires large and expensive platforms. For instance, achieving a resolution of 1m at visible wavelengths from LEO typically requires an aperture diameter of at least 30cm. Additionally, ensuring high revisit times often prompts the use of multiple satellites. In light of these challenges, a small, segmented, deployable CubeSat telescope was recently proposed creating the additional need of phasing the telescope's mirrors. Phasing methods on compact platforms are constrained by the limited volume and power available, excluding solutions that rely on dedicated hardware or demand substantial computational resources. Neural Network (NN) are known for their computationally efficient inference and reduced on board requirements. Therefore we developed a NN based method to measure co phasing errors inherent to a deployable telescope. The proposed technique demonstrates its ability to detect phasing error at the targeted performance level (typically a wavefront error (WFE) below 15 nm RMS for a visible imager operating at the diffraction limit) using a point source. The robustness of the NN method is verified in presence of high order aberrations or noise and the results are compared against existing state of the art techniques. The developed NN model ensures its feasibility and provides a realistic pathway towards achieving diffraction limited images.

astro-ph.IM↗
arXiv2022

Adapting the pyramid wavefront sensor for pupil fragmentation of the ELT class telescopes

The next generation of Extremely Large Telescope (24 to 39m diameter) will suffer from the so-called "pupil fragmentation" problem. Due to their pupil shape complexity (segmentation, large spiders ...), some differential pistons may appear between some isolated part of the full pupil during the observations. Although classical AO system will be able to correct for turbulence effects, they will be blind to this specific telescope induced perturbations. Hence, such differential piston, a.k.a petal modes, will prevent to reach the diffraction limit of the telescope and ultimately will represent the main limitation of AO-assisted observation with an ELT. In this work we analyse the spatial structure of these petal modes and how it affects the ability of a Pyramid Wavefront sensor to sense them. Then we propose a variation around the classical Pyramid concept for increasing the WFS sensitivity to this particular modes. Nevertheless, We show that one single WFS can not accurately and simultaneously measure turbulence and petal modes. We propose a double path wavefront sensor scheme to solve this problem. We show that such a scheme,associated to a spatial filtering of residual turbulence in the second WFS path dedicated to petal mode sensing, allows to fully measure and correct for both turbulence and fragmentation effects and will eventually restore the full capability and spatial resolution of the future ELT.

astro-ph.IM↗
arXiv2021

Active deployable primary mirrors on CubeSat

The volume available on small satellites restricts the size of optical apertures to a few centimetres, limiting the Ground-Sampling Distance (GSD) in the visible to typically 3 m at 500 km. We present in this paper the latest development of a laboratory demonstrator of a segmented deployable telescope that will triple the achievable ground resolution and improve photometric capability of CubeSat imagers. Each mirror segment is folded for launch and unfolds in space. We demonstrate through laboratory validation very high deployment repeatability of the mirrors <{\pm}5 μm. To enable diffraction-limited imaging, segments are controlled in piston, tip, and tilt. This is achieved by an initial coarse alignment of the mirrors followed by a fine phasing step. Finally, we investigate the impact of the thermal environment on high-order wavefront error and the conceptual design of a deployable secondary fitting inside 1U.

astro-ph.IM↗
arXiv2021

Direct imaging and spectroscopy of exoplanets with the ELT/HARMONI high-contrast module

Combining high-contrast imaging with medium-resolution spectroscopy has been shown to significantly boost the direct detection of exoplanets. HARMONI, one of the first-light instruments to be mounted on ESO's ELT, will be equipped with a single-conjugated adaptive optics system to reach the diffraction limit of the ELT in H and K bands, a high-contrast module dedicated to exoplanet imaging, and a medium-resolution (up to R = 17 000) optical and near-infrared integral field spectrograph. Combined together, these systems will provide unprecedented contrast limits at separations between 50 and 400 mas. In this paper, we estimate the capabilities of the HARMONI high-contrast module for the direct detection of young giant exoplanets. We use an end-to-end model of the instrument to simulate observations based on realistic observing scenarios and conditions. We analyze these data with the so-called "molecule mapping" technique combined to a matched-filter approach, in order to disentangle the companions from the host star and tellurics, and increase the S/N of the planetary signal. We detect planets above 5-sigma at contrasts up to 16 mag and separations down to 75 mas in several spectral configurations of the instrument. We show that molecule mapping allows the detection of companions up to 2.5 mag fainter compared to state-of-the-art high-contrast imaging techniques based on angular differential imaging. We also demonstrate that the performance is not strongly affected by the spectral type of the host star, and that we reach close sensitivities for the best three quartiles of observing conditions at Armazones, which means that HARMONI could be used in near-critical observations during 60 to 70% of telescope time at the ELT. Finally, we simulate planets from population synthesis models to further explore the parameter space that HARMONI and its high-contrast module will soon open.

astro-ph.EP↗
arXiv2020

Pyramid wavefront sensor optical gains compensation using a convolutional model

Extremely Large Telescopes have overwhelmingly opted for the Pyramid wavefront sensor (PyWFS) over the more widely used Shack-Hartmann WaveFront Sensor (SHWFS) to perform their Single Conjugate Adaptive Optics (SCAO) mode. The PyWFS, a sensor based on Fourier filtering, has proven to be highly successful in many astronomy applications. However, it exhibits non-linearity behaviors that lead to a reduction of its sensitivity when working with non-zero residual wavefronts. This so-called Optical Gains (OG) effect, degrades the close loop performance of SCAO systems and prevents accurate correction of Non-Common Path Aberrations (NCPA). In this paper, we aim at computing the OG using a fast and agile strategy in order to control the PyWFS measurements in adaptive optics closed loop systems. Using a novel theoretical description of the PyFWS, which is based on a convolutional model, we are able to analytically predict the behavior of the PyWFS in closed-loop operation. This model enables us to explore the impact of residual wavefront error on particular aspects such as sensitivity and associated OG. The proposed method relies on the knowledge of the residual wavefront statistics and enables automatic estimation of the current OG. End-to-End numerical simulations are used to validate our predictions and test the relevance of our approach. We demonstrate, using on non-invasive strategy, that our method provides an accurate estimation of the OG. The model itself only requires AO telemetry data to derive statistical information on atmospheric turbulence. Furthermore, we show that by only using an estimation of the current Fried parameter r_0 and the basic system-level characteristics, OGs can be estimated with an accuracy of less than 10%. Finally, we highlight the importance of OG estimation in the case of NCPA compensation. The proposed method is applied to the PyWFS.

astro-ph.IM↗
arXiv2020

Design of the HARMONI Pyramid WFS module

Current designs for all three extremely large telescopes show the overwhelming adoption of the pyramid wavefront sensor (P-WFS) as the WFS of choice for adaptive optics (AO) systems sensing on natural guide stars (NGS) or extended objects. The key advantages of the P-WFS over the Shack-Hartmann are known and are mainly provided by the improved sensitivity (fainter NGS) and reduced sensitivity to spatial aliasing. However, robustness and tolerances of the P-WFS for the ELTs are not currently well understood. In this paper, we present simulation results for the single-conjugate AO mode of HARMONI, a visible and near-infrared integral field spectrograph for the European Extremely Large Telescope. We first explore the wavefront sensing issues related to the telescope itself; namely the island effect (i.e. differential piston) and M1 segments phasing errors. We present mitigation strategies to the island effect and their performance. We then focus on some performance optimisation aspects of the AO design to explore the impact of the RTC latency and the optical gain issues, which will in particular affect the high-contrast mode of HARMONI. Finally, we investigate the influence of the quality of glass pyramid prism itself, and of optical aberrations on the final AO performance. By relaxing the tolerances on the fabrication of the prism, we are able to reduce hardware costs and simplify integration. We show the importance of calibration (i.e. updating the control matrix) to capture any displacement of the telescope pupil and rotation of the support structure for M4. We also show the importance of the number of pixels used for wavefront sensing to relax tolerances of the pyramid prism. Finally, we present a detailed optical design of the pyramid prism, central element of the P-WFS.

astro-ph.IM↗
arXiv2020

Adaptive optics for multifocal plane microscopy

Multifocal plane microscopy (MUM) allows three dimensional objects to be imaged in a single camera frame. Our approach uses dual orthogonal diffraction phase gratings with a quadratic distortion of the lines to apply defocus to the first diffraction orders which, when paired with a relay lens, allows for 9 focal planes to be imaged on a single camera chip. This approach requires a strong signal level to ensure sufficient intensity in the diffracted light, but has the advantage of being compact and straightforward to implement. As the microscope begins to focus deeper into the sample, aberrations caused by refractive index mismatch and inhomogeneity in the sample's media have an adverse effect on the signal's quality. In this paper, we investigate the image quality improvement brought by applying adaptive optics (AO) to multifocal plane microscopy. A single correction device (an 8x8 deformable mirror (DM)) is combined with an image-based AO control strategy to perform the correction of optical aberrations. We compare full end-to-end modelling results using an established numerical modelling system adapted for microscopy to laboratory results both on a test sample and on a number of biological samples. Finally, we will demonstrate that combining AO and MUM, we are able to improve the image quality of biological samples and provide a good correction throughout the volume of the biological sample.

physics.optics↗
arXiv2018

Laboratory Demonstration of an Active Optics System for High-Resolution Deployable CubeSat

In this paper we present HighRes: a laboratory demonstration of a 3U CubeSat with a deployable primary mirror that has the potential of achieving high-resolution imaging for Earth Observation. The system is based on a Cassegrain telescope with a segmented primary mirror composed of 4 petals that form an effective aperture of 300 mm. The design provides diffraction limited performance over the entire field-of-view and allows for a panchromatic ground-sampling distance of less than 1 m at an altitude of 350 km. The alignment and co-phasing of the mirror segments is performed by focal plane sharpening and is validated through rigorous numerical simulations. The opto-mechanical design of the prototype and its laboratory demonstration are described and measurements from the on-board metrology sensors are presented. This data verifies that the performance of the mirror deployment and manipulation systems is sufficient for co-phasing. In addition, it is shown that the mirrors can be driven to any target position with an accuracy of 25 nm using closed-loop feedback between the mirror motors and the on-board metrology.

astro-ph.IM↗
arXiv2018

Sensing and control of segmented mirrors with a pyramid wavefront sensor in the presence of spiders

The segmentation of the telescope pupil (by spiders & the segmented M4) create areas of phase isolated by the width of the spiders on the wavefront sensor (WFS), breaking the spatial continuity of the wavefront. The poor sensitivity of the Pyramid WFS (PWFS) to differential piston leads to badly seen and therefore uncontrollable differential pistons. In close loop operation, differential pistons between segments will settle around integer values of the average sensing wavelength. The differential pistons typically range from one to ten times the sensing wavelength and vary rapidly over time, leading to extremely poor performance. In addition, aberrations created by atmospheric turbulence will contain large amounts of differential piston between the segments. Removing piston contribution over each of the DM segments leads to poor performance. In an attempt to reduce the impact of unwanted differential pistons that are injected by the AO correction, we compare three different approaches. We first limit ourselves to only use the information measured by the PWFS, in particular by reducing the modulation. We show that using this information sensibly is important but will not be sufficient. We discuss possible ways of improvement by using prior information. A second approach is based on phase closure of the DM commands and assumes the continuity of the correction wavefront over the entire unsegmented pupil. The last approach is based on the pair-wise slaving of edge actuators and shows the best results. We compare the performance of these methods using realistic end-to-end simulations. We find that pair-wise slaving leads to a small increase of the total wavefront error, only adding between 20-45 nm RMS in quadrature for seeing conditions between 0.45-0.85 arcsec. Finally, we discuss the possibility of combining the different proposed solutions to increase robustness.

astro-ph.IM↗