pith. sign in

arxiv: 2606.30620 · v1 · pith:HJHYJDAGnew · submitted 2026-06-29 · 🌌 astro-ph.IM

Gaussian processes on ray-guided transformed uniform grids for fast, flexible, and auto-differentiable adaptive source reconstruction in lens modelling

Wolfgang J. R. Enzi , Coleman M. Krawczyk , Tian Li , Thomas E. Collett This is my paper

Pith reviewed 2026-06-30 03:11 UTC · model grok-4.3

classification 🌌 astro-ph.IM
keywords strong gravitational lensingsource reconstructionGaussian processadaptive meshdifferentiable modelingray tracinglens modelingevidence lower bound
0
0 comments X

The pith

Ray-guided transformed uniform grids enable adaptive source reconstruction in gravitational lensing with Gaussian processes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes modelling the lensed source on a uniform grid transformed by the cumulative distribution of rays traced from the image plane. This produces source pixels with more uniform ray counts, making the reconstruction adaptive to the lens model. The source is represented as a Gaussian process, which preserves auto-differentiability and allows any power spectrum for regularization. Computations use the fast Fourier transform for speed. Tests on mock data show similar fit quality with about half as many pixels per dimension and improved evidence lower bounds.

Core claim

Defining the source as a Gaussian process on a ray-guided transformed uniform grid (RTU grid) allows adaptive source reconstruction that is auto-differentiable and flexible in regularization choice. The grid transformation is based on cumulative distributions of rays to the source plane, ensuring more uniform ray counts per pixel. This approach achieves comparable reconstruction quality to standard methods but with fewer source pixels, typically a factor of two fewer per dimension, and higher ELBO values for fixed pixel numbers.

What carries the argument

The ray-guided transformed uniform grid (RTU grid), created by transforming a uniform grid using cumulative ray distributions from the lens model to adapt pixel sizes while maintaining uniformity in ray coverage for the Gaussian process representation.

If this is right

  • Comparable fit quality is achieved with roughly half the number of source pixels per dimension.
  • Higher Evidence Lower Bounds (ELBOs) are obtained for the same number of pixels.
  • The difference in ELBO between models with and without substructures is only mildly affected.
  • A fast and differentiable method supports analysis of large lens samples such as those from the Euclid survey.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This method could facilitate fully differentiable pipelines for simultaneous inference of lens mass and source parameters.
  • It may extend to other imaging inverse problems involving ray tracing or non-uniform sampling.
  • Testing on real observational data would reveal if the uniformity assumption holds under noise and complex structures.

Load-bearing premise

The cumulative-distribution transformation produces source pixels with sufficiently uniform ray counts without introducing artifacts that degrade reconstruction quality or differentiability.

What would settle it

Applying the method to mock data with known complex sources and checking if the reconstructed source shows systematic artifacts or if the ELBO improvement disappears when the transformation introduces discontinuities.

Figures

Figures reproduced from arXiv: 2606.30620 by Coleman M. Krawczyk, Thomas E. Collett, Tian Li, Wolfgang J. R. Enzi.

Figure 1
Figure 1. Figure 1: This figure shows, for four test cases, the non-uniform ray density created by lensing and how the RTU grid provides a more even sampling. The four test cases are shown in the four subfigures for Mock 1 (top left), Mock 2 (top right), Mock 3 (bottom left), and SDSS J0946+1006 (bottom right). Each subfigure contains the following panels: the data with the arc mask shown as a black contour (top left), the ma… view at source ↗
Figure 2
Figure 2. Figure 2: This figure shows the reconstructed sources and normalized residuals as a function of the pixel number per side, 𝑁src, for the RTU grid (top two rows) and the uniform grid (bottom two rows). Residuals are clipped to a range of ±3𝜎. From left to right, number of pixels per side is 𝑁src ∈ {8, 16, 32, 64, 128, 256, 512, 1024}. Pixels outside of the source grids are set to 0. All panels show the results of the… view at source ↗
Figure 3
Figure 3. Figure 3: Same as [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Same as [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Same as [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: From left to right, for the three mocks we show the mock input source, the reconstructed source on a uniform grid, the reconstructed source on an RTU grid after the transformation is applied, and the reconstructed source on the RTU grid before the transformation is applied. For SDSS J0946+1006 no input source is available, so we show only the three recon￾structed sources. From top to bottom we show Mock 1,… view at source ↗
Figure 6
Figure 6. Figure 6: The evolution of 𝜒 2 img within the mask (left panels) and the model log(ELBO) (right panels). We show these diagnostics for Mock 1, Mock 2, Mock 3, and SDSS J0946+1006 (from top to bottom). Dashed lines indicate models with a substructure, while models without one are shown with solid lines. The vertical dot-dashed line marks the rule-of-thumb maximum 𝑁src along each dimension to avoid strong prior interp… view at source ↗
Figure 8
Figure 8. Figure 8: The evolution of Δ 2 src for Mock 1 (left), Mock 2 (middle), and Mock 3 (right). to the presence of the included substructure. With a difference of Δ log(ELBO) ≈ 35, the ELBO comparison suggests a preference for the model with an NFW substructure for SDSS J0946+1006 when the RTU grid is employed. Finally, we test whether the reconstructed source itself is accurate. To this end, we consider the difference w… view at source ↗
read the original abstract

Strong gravitational lensing constrains cosmology and dark matter, but robust inference requires accurate source reconstruction. The achievable source resolution is highly position-dependent. Adaptive meshes can place resolution where needed, but typically rely on discontinuous operations, such as Delaunay tessellations or Voronoi binning, which can restrict regularization choices and break differentiability. In this paper, we present a novel approach for modelling the source on a ray-guided transformed uniform grid (RTU grid), that is adaptive to the lens mass model, auto-differentiable and flexible with respect to the regularization by allowing for an arbitrary choice of power spectrum. We achieve this by defining the source as a Gaussian process on a uniform grid, which is then transformed based on the cumulative distributions of rays traced back to the source plane. This approach ensures that source pixels contain a more uniform number of rays. The approach is fast by leveraging the fast Fourier transform to describe the Gaussian process in Fourier space. We apply this new approach to mock data and show that it achieves comparable fit quality with fewer source pixels, typically corresponding to about a factor of two fewer pixels per dimension, and increases Evidence Lower Bounds (ELBOs) for the same number of pixels. Using the RTU grid only mildly affects the difference in ELBO for models with and without substructures within lens galaxies. A fast, flexible, and auto-differentiable source reconstruction can greatly benefit the analysis of large samples of lens systems, e.g. those found within the Euclid survey

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces a ray-guided transformed uniform (RTU) grid for source reconstruction in strong gravitational lens modeling. The source is modeled as a Gaussian process on a uniform grid that is warped via cumulative distributions of rays traced from the image plane, producing source pixels with more uniform ray counts. This enables adaptivity to the lens mass model, full auto-differentiability, compatibility with arbitrary GP power spectra, and FFT-based evaluation for speed. On mock data the method reportedly yields comparable fit quality using roughly half as many pixels per dimension and higher ELBO values, while only mildly affecting ELBO differences between models with and without substructure.

Significance. If the differentiability and artifact-free properties of the RTU transform hold, the approach would provide a useful middle ground between fixed grids and discontinuous adaptive meshes, supporting scalable, differentiable inference on large lens samples such as those expected from Euclid. The reuse of standard GP/FFT machinery together with the ray-guided warp is a clear engineering strength.

major comments (2)
  1. [Abstract] Abstract: the central performance claim ('achieves comparable fit quality with fewer source pixels, typically corresponding to about a factor of two fewer pixels per dimension, and increases Evidence Lower Bounds (ELBOs) for the same number of pixels') is stated without any numerical values, baseline comparisons, error estimates, or statistical significance tests, which is load-bearing for evaluating whether the reported gains are robust.
  2. [Method (ray-guided transform)] Method description of the cumulative-distribution transform: no explicit verification is given that the Jacobian of the ray-guided mapping remains well-conditioned under realistic lens-model gradients or that residual ray-count variation stays below a few percent, which is required to substantiate the claims of preserved auto-differentiability and absence of high-frequency artifacts in the FFT-based GP evaluation.
minor comments (1)
  1. [Abstract] The abstract would be clearer if it briefly stated the specific form of the GP power spectrum employed in the mock-data experiments.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments. We address each major comment point-by-point below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central performance claim ('achieves comparable fit quality with fewer source pixels, typically corresponding to about a factor of two fewer pixels per dimension, and increases Evidence Lower Bounds (ELBOs) for the same number of pixels') is stated without any numerical values, baseline comparisons, error estimates, or statistical significance tests, which is load-bearing for evaluating whether the reported gains are robust.

    Authors: We agree that the abstract would benefit from greater specificity. The body of the manuscript reports detailed mock-data comparisons that quantify the pixel reduction (typically a factor of ~2 per dimension) and ELBO gains relative to uniform grids. We will revise the abstract to incorporate representative numerical values, baseline references, and a brief note on consistency across the tested models. revision: yes

  2. Referee: [Method (ray-guided transform)] Method description of the cumulative-distribution transform: no explicit verification is given that the Jacobian of the ray-guided mapping remains well-conditioned under realistic lens-model gradients or that residual ray-count variation stays below a few percent, which is required to substantiate the claims of preserved auto-differentiability and absence of high-frequency artifacts in the FFT-based GP evaluation.

    Authors: We acknowledge that explicit numerical checks on Jacobian conditioning and residual ray-count uniformity are not provided in the submitted manuscript. While the cumulative-distribution construction is intended to guarantee these properties, we will add a dedicated verification subsection (or appendix) that quantifies the Jacobian eigenvalues under realistic lens gradients and demonstrates that ray-count variation remains below a few percent, thereby supporting the differentiability and FFT-artifact claims. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained from standard components.

full rationale

The paper defines the RTU grid explicitly via cumulative distributions of rays traced from the lens model to the source plane, places a Gaussian process on the resulting transformed uniform grid, and evaluates it in Fourier space via FFT. These steps are constructed from ray-tracing, cumulative distribution functions, and standard GP/FFT operations without any equation reducing a reported performance gain (e.g., ELBO improvement or pixel reduction) to a quantity defined by the same fitted parameters. No self-citation is invoked as a uniqueness theorem or load-bearing premise, no ansatz is smuggled via prior work, and the mock-data results are presented as empirical outcomes rather than tautological predictions. The central claims therefore remain independent of the inputs they are tested against.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The approach rests on standard mathematical properties of Gaussian processes and Fourier transforms together with domain-standard ray-tracing assumptions; the main addition is the specific grid transformation.

axioms (2)
  • standard math Gaussian processes admit an efficient Fourier-space representation that can be evaluated with the fast Fourier transform.
    Invoked to achieve fast computation of the source model.
  • domain assumption Ray tracing from the image plane to the source plane yields cumulative distributions that can be used to define a differentiable transformation producing more uniform ray sampling.
    Central to the definition of the RTU grid.
invented entities (1)
  • Ray-guided transformed uniform (RTU) grid no independent evidence
    purpose: Provide an adaptive yet differentiable source pixelization that maintains roughly uniform ray counts per pixel.
    Newly defined transformation introduced in the paper; no independent evidence supplied.

pith-pipeline@v0.9.1-grok · 5822 in / 1404 out tokens · 74899 ms · 2026-06-30T03:11:16.815932+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

72 extracted references · 67 canonical work pages · 23 internal anchors

  1. [1]

    arXiv e-prints , keywords =

    Lopsidedness in Early-Type Galaxies: the role of the m=1 multipole in Isophote Fitting and Strong Lens Modelling. arXiv e-prints , keywords =. doi:10.48550/arXiv.2407.12983 , archivePrefix =. 2407.12983 , primaryClass =

    work page doi:10.48550/arxiv.2407.12983

  2. [2]

    2007 , publisher=

    An Introduction to Copulas , author=. 2007 , publisher=

    2007

  3. [3]

    R\'enyi Divergence Variational Inference

    R \'e nyi Divergence Variational Inference. arXiv e-prints , keywords =. doi:10.48550/arXiv.1602.02311 , archivePrefix =. 1602.02311 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1602.02311

  4. [4]

    arXiv e-prints , keywords =

    Mind the Information Gap: Unveiling Detailed Morphologies of z 0.5-1.0 Galaxies with SLACS Strong Lenses and Data-Driven Analysis. arXiv e-prints , keywords =. doi:10.48550/arXiv.2511.19595 , archivePrefix =. 2511.19595 , primaryClass =

    work page doi:10.48550/arxiv.2511.19595

  5. [5]

    , keywords =

    Gravitational imaging through a triple source plane lens: revisiting the CDM-defying dark subhalo in SDSSJ0946+1006. , keywords =. doi:10.1093/mnras/stae514 , archivePrefix =. 2309.04535 , primaryClass =

    work page doi:10.1093/mnras/stae514

  6. [6]

    Analytic models of plausible gravitational lens potentials

    Analytic models of plausible gravitational lens potentials. , keywords =. doi:10.1088/1475-7516/2009/01/015 , archivePrefix =. 0705.0682 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/1475-7516/2009/01/015 2009

  7. [7]

    Chen and Martin Jankowiak and Fritz Obermeyer and Neeraj Pradhan and Theofanis Karaletsos and Rohit Singh and Paul A

    Eli Bingham and Jonathan P. Chen and Martin Jankowiak and Fritz Obermeyer and Neeraj Pradhan and Theofanis Karaletsos and Rohit Singh and Paul A. Szerlip and Paul Horsfall and Noah D. Goodman , title =. J. Mach. Learn. Res. , volume =. 2019 , url =

    2019

  8. [8]

    Lenstronomy: multi-purpose gravitational lens modelling software package

    lenstronomy: Multi-purpose gravitational lens modelling software package. Physics of the Dark Universe , keywords =. doi:10.1016/j.dark.2018.11.002 , archivePrefix =. 1803.09746 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1016/j.dark.2018.11.002 2018

  9. [9]

    , keywords =

    Unveiling lens light complexity with a novel multi-Gaussian expansion approach for strong gravitational lensing. , keywords =. doi:10.1093/mnras/stae1577 , archivePrefix =. 2403.16253 , primaryClass =

    work page doi:10.1093/mnras/stae1577

  10. [10]

    James Bradbury and Roy Frostig and Peter Hawkins and Matthew James Johnson and Chris Leary and Dougal Maclaurin and George Necula and Adam Paszke and Jake Vander

  11. [11]

    Multipole Gravitational Lensing and High-order Perturbations on the Quadrupole Lens

    Multipole Gravitational Lensing and High-order Perturbations on the Quadrupole Lens. , keywords =. doi:10.1088/0004-637X/765/2/134 , archivePrefix =. 1302.5482 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/765/2/134

  12. [12]

    , keywords =

    General multipoles and their implications for dark matter inference. , keywords =. doi:10.1093/mnras/stae1228 , archivePrefix =. 2403.08895 , primaryClass =

    work page doi:10.1093/mnras/stae1228

  13. [13]

    2006, MNRAS, 372, 1699, doi: 10.1111/j.1365-2966.2006.10972.x

    A Bayesian analysis of regularized source inversions in gravitational lensing. , keywords =. doi:10.1111/j.1365-2966.2006.10733.x , archivePrefix =. astro-ph/0601493 , primaryClass =

    work page doi:10.1111/j.1365-2966.2006.10733.x 2006

  14. [14]

    doi:10.5281/zenodo.12167630 , url =

    Coleman Krawczyk , title =. doi:10.5281/zenodo.12167630 , url =

    work page doi:10.5281/zenodo.12167630

  15. [15]

    Cosmological Constraints from the double source plane lens SDSSJ0946+1006

    Cosmological constraints from the double source plane lens SDSSJ0946+1006. , keywords =. doi:10.1093/mnras/stu1190 , archivePrefix =. 1403.5278 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stu1190

  16. [16]

    21cmfast: a fast, seminumerical simulation of the high-redshift 21-cm signal: 21cmfast , volume=

    Detection of a dark substructure through gravitational imaging. , keywords =. doi:10.1111/j.1365-2966.2010.16865.x , archivePrefix =. 0910.0760 , primaryClass =

    work page doi:10.1111/j.1365-2966.2010.16865.x 2010

  17. [17]

    , keywords =

    Bayesian strong gravitational-lens modelling on adaptive grids: objective detection of mass substructure in Galaxies. , keywords =. doi:10.1111/j.1365-2966.2008.14005.x , archivePrefix =. 0805.0201 , primaryClass =

    work page doi:10.1111/j.1365-2966.2008.14005.x 2008

  18. [18]

    , keywords =

    Detecting low-mass haloes with strong gravitational lensing I: the effect of data quality and lensing configuration. , keywords =. doi:10.1093/mnras/stab3537 , archivePrefix =. 2111.08718 , primaryClass =

    work page doi:10.1093/mnras/stab3537

  19. [19]

    arXiv e-prints , keywords =

    Detecting low-mass haloes with strong gravitational lensing II: constraints on the density profiles of two detected subhaloes. arXiv e-prints , keywords =. doi:10.48550/arXiv.2407.12910 , archivePrefix =. 2407.12910 , primaryClass =

    work page doi:10.48550/arxiv.2407.12910

  20. [20]

    , keywords =

    The overconcentrated dark halo in the strong lens SDSS J0946 + 1006 is a subhalo: evidence for self-interacting dark matter?. , keywords =. doi:10.1093/mnras/staf697 , archivePrefix =. 2411.08565 , primaryClass =

    work page doi:10.1093/mnras/staf697

  21. [21]

    Entropy , keywords =

    Geometric Variational Inference. Entropy , keywords =. doi:10.3390/e23070853 , archivePrefix =. 2105.10470 , primaryClass =

    work page doi:10.3390/e23070853

  22. [22]

    2013 , publisher=

    Bayesian Data Analysis , author=. 2013 , publisher=

    2013

  23. [23]

    , keywords =

    Time-Delay Cosmography: Measuring the Hubble Constant and Other Cosmological Parameters with Strong Gravitational Lensing. , keywords =. doi:10.1007/s11214-024-01079-w , archivePrefix =. 2210.10833 , primaryClass =

    work page doi:10.1007/s11214-024-01079-w

  24. [24]

    , keywords =

    Using wavelets to capture deviations from smoothness in galaxy-scale strong lenses. , keywords =. doi:10.1051/0004-6361/202244464 , archivePrefix =. 2207.05763 , primaryClass =

    work page doi:10.1051/0004-6361/202244464

  25. [25]

    arXiv e-prints , keywords =

    Exploiting the diversity of modeling methods to probe systematic biases in strong lensing analyses. arXiv e-prints , keywords =. doi:10.48550/arXiv.2406.08484 , archivePrefix =. 2406.08484 , primaryClass =

    work page doi:10.48550/arxiv.2406.08484

  26. [26]

    The Sloan Lens ACS Survey. VI. Discovery and Analysis of a Double Einstein Ring. , keywords =. doi:10.1086/529541 , archivePrefix =. 0801.1555 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1086/529541

  27. [27]

    Measuring the value of the Hubble constant "\`a la Refsdal"

    Measuring the Value of the Hubble Constant \`a la Refsdal. , keywords =. doi:10.3847/1538-4357/aac2c9 , archivePrefix =. 1802.01584 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4357/aac2c9

  28. [28]

    The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo

    The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo. arXiv e-prints , keywords =. doi:10.48550/arXiv.1111.4246 , archivePrefix =. 1111.4246 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1111.4246

  29. [29]

    A Catalog of Mass Models for Gravitational Lensing

    A Catalog of Mass Models for Gravitational Lensing. arXiv e-prints , keywords =. doi:10.48550/arXiv.astro-ph/0102341 , archivePrefix =. astro-ph/0102341 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.astro-ph/0102341

  30. [30]

    2012 , publisher=

    Interpolation of Spatial Data: Some Theory for Kriging , author=. 2012 , publisher=

    2012

  31. [31]

    Slice Sampling

    Slice Sampling. arXiv e-prints , keywords =. doi:10.48550/arXiv.physics/0009028 , archivePrefix =. physics/0009028 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.physics/0009028

  32. [32]

    Handbook of Markov Chain Monte Carlo , year = 2011, pages =

    MCMC Using Hamiltonian Dynamics. Handbook of Markov Chain Monte Carlo , year = 2011, pages =. doi:10.1201/b10905 , adsurl =

    work page doi:10.1201/b10905 2011

  33. [33]

    , keywords =

    Scanning for dark matter subhaloes in Hubble Space Telescope imaging of 54 strong lenses. , keywords =. doi:10.1093/mnras/stad3694 , adsurl =

    work page doi:10.1093/mnras/stad3694

  34. [34]

    Adaptive Semi-linear Inversion of Strong Gravitational Lens Imaging

    Adaptive semi-linear inversion of strong gravitational lens imaging. , keywords =. doi:10.1093/mnras/stv1455 , archivePrefix =. 1412.7436 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stv1455

  35. [35]

    The Journal of Open Source Software , keywords =

    PyAutoLens: Open-Source Strong Gravitational Lensing. The Journal of Open Source Software , keywords =. doi:10.21105/joss.02825 , archivePrefix =. 2106.01384 , primaryClass =

    work page doi:10.21105/joss.02825

  36. [36]

    V., Poznanski D., Wang X., Ganeshalingam M., Mannucci F., 2011, @doi [ ] 10.1111/j.1365-2966.2011.18162.x , https://ui.adsabs.harvard.edu/abs/2011MNRAS.412.1473L 412, 1473

    Detailed cluster lensing profiles at large radii and the impact on cluster weak lensing studies. , keywords =. doi:10.1111/j.1365-2966.2011.18481.x , archivePrefix =. 1101.0650 , primaryClass =

    work page doi:10.1111/j.1365-2966.2011.18481.x 2011

  37. [37]

    , keywords =

    Angular complexity in strong lens substructure detection. , keywords =. doi:10.1093/mnras/stae153 , archivePrefix =. 2310.10714 , primaryClass =

    work page doi:10.1093/mnras/stae153

  38. [38]

    Composable Effects for Flexible and Accelerated Probabilistic Programming in NumPyro

    Composable Effects for Flexible and Accelerated Probabilistic Programming in NumPyro , author=. arXiv preprint arXiv:1912.11554 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 1912

  39. [39]

    Planck 2018 results. VI. Cosmological parameters. , keywords =. doi:10.1051/0004-6361/201833910 , archivePrefix =. 1807.06209 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201833910 2018

  40. [40]

    , keywords =

    A lensed radio jet at milliarcsecond resolution I: Bayesian comparison of parametric lens models. , keywords =. doi:10.1093/mnras/stac2350 , archivePrefix =. 2207.03375 , primaryClass =

    work page doi:10.1093/mnras/stac2350

  41. [41]

    A charming Bayesian strong lensing reconstruction framework

    Introducing LensCharm. A charming Bayesian strong lensing reconstruction framework. , keywords =. doi:10.1051/0004-6361/202348256 , adsurl =

    work page doi:10.1051/0004-6361/202348256

  42. [42]

    Understanding the Core-Halo Relation of Quantum Wave Dark Matter, $\psi$DM, from 3D Simulations

    Understanding the Core-Halo Relation of Quantum Wave Dark Matter from 3D Simulations. , keywords =. doi:10.1103/PhysRevLett.113.261302 , archivePrefix =. 1407.7762 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physrevlett.113.261302

  43. [43]

    Source-position transformation -- an approximate invariance in strong gravitational lensing

    Source-position transformation: an approximate invariance in strong gravitational lensing. , keywords =. doi:10.1051/0004-6361/201322106 , archivePrefix =. 1306.4675 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201322106

  44. [44]

    Unified lensing and kinematic analysis for any elliptical mass profile

    Unified lensing and kinematic analysis for any elliptical mass profile. , keywords =. doi:10.1093/mnras/stz1796 , archivePrefix =. 1906.08263 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stz1796 1906

  45. [45]

    , keywords =

    Quantifying the structure of strong gravitational lens potentials with uncertainty-aware deep neural networks. , keywords =. doi:10.1093/mnras/staa3201 , archivePrefix =. 2010.07314 , primaryClass =

    work page doi:10.1093/mnras/staa3201 2010

  46. [46]

    , keywords =

    The very knotty lenser: Exploring the role of regularization in source and potential reconstructions using Gaussian process regression. , keywords =. doi:10.1093/mnras/stac1924 , archivePrefix =. 2202.09378 , primaryClass =

    work page doi:10.1093/mnras/stac1924

  47. [47]

    arXiv e-prints , keywords =

    High detection significance of the dark substructure in gravitational lens SDSSJ0946+1006 is revealed by image pixel supersampling. arXiv e-prints , keywords =. doi:10.48550/arXiv.2408.11090 , archivePrefix =. 2408.11090 , primaryClass =

    work page doi:10.48550/arxiv.2408.11090

  48. [48]

    , keywords =

    An unexpected high concentration for the dark substructure in the gravitational lens SDSSJ0946+1006. , keywords =. doi:10.1093/mnras/stab2247 , archivePrefix =. 2011.10627 , primaryClass =

    work page doi:10.1093/mnras/stab2247 2011

  49. [49]

    , keywords =

    Strong Gravitational Lensing as a Probe of Dark Matter. , keywords =. doi:10.1007/s11214-024-01087-w , archivePrefix =. 2306.11781 , primaryClass =

    work page doi:10.1007/s11214-024-01087-w

  50. [50]

    Semi-linear gravitational lens inversion

    Semilinear Gravitational Lens Inversion. , keywords =. doi:10.1086/375132 , archivePrefix =. astro-ph/0302587 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1086/375132

  51. [51]

    Automated Variational Inference in Probabilistic Programming

    Automated Variational Inference in Probabilistic Programming. arXiv e-prints , keywords =. doi:10.48550/arXiv.1301.1299 , archivePrefix =. 1301.1299 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1301.1299

  52. [52]

    arXiv e-prints , keywords =

    AdaBelief Optimizer: Adapting Stepsizes by the Belief in Observed Gradients. arXiv e-prints , keywords =. doi:10.48550/arXiv.2010.07468 , archivePrefix =. 2010.07468 , primaryClass =

    work page doi:10.48550/arxiv.2010.07468 2010

  53. [53]

    , year = 1964, month = jan, volume =

    On the possibility of determining Hubble's parameter and the masses of galaxies from the gravitational lens effect. , year = 1964, month = jan, volume =. doi:10.1093/mnras/128.4.307 , adsurl =

    work page doi:10.1093/mnras/128.4.307 1964

  54. [54]

    Time Delay Cosmography

    Time delay cosmography. , keywords =. doi:10.1007/s00159-016-0096-8 , archivePrefix =. 1605.05333 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1007/s00159-016-0096-8

  55. [55]

    , keywords =

    Strong lensing time-delay cosmography in the 2020s. , keywords =. doi:10.1007/s00159-022-00145-y , archivePrefix =. 2210.15794 , primaryClass =

    work page doi:10.1007/s00159-022-00145-y

  56. [56]

    , keywords =

    TDCOSMO 2025: Cosmological constraints from strong lensing time delays. , keywords =. doi:10.1051/0004-6361/202555801 , archivePrefix =. 2506.03023 , primaryClass =

    work page doi:10.1051/0004-6361/202555801 2025

  57. [57]

    , keywords =

    Strong Gravitational Lensing and Microlensing of Supernovae. , keywords =. doi:10.1007/s11214-024-01044-7 , archivePrefix =. 2301.07729 , primaryClass =

    work page doi:10.1007/s11214-024-01044-7

  58. [58]

    A precise extragalactic test of General Relativity

    A precise extragalactic test of General Relativity. Science , keywords =. doi:10.1126/science.aao2469 , archivePrefix =. 1806.08300 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1126/science.aao2469

  59. [59]

    , keywords =

    A Novel Test for MOND: Gravitational Lensing by Disk Galaxies. , keywords =. doi:10.3847/1538-4357/ae11ad , archivePrefix =. 2411.17888 , primaryClass =

    work page doi:10.3847/1538-4357/ae11ad

  60. [60]

    Nature Astronomy , keywords =

    A possible challenge for cold and warm dark matter. Nature Astronomy , keywords =. doi:10.1038/s41550-025-02746-w , archivePrefix =. 2601.02466 , primaryClass =

    work page doi:10.1038/s41550-025-02746-w

  61. [61]

    The dense, low-mass perturbers in B1938+666 and J0946+1006: implications for cold and self-interacting dark matter

    SHARP─IX. The dense, low-mass perturbers in B1938+666 and J0946+1006: implications for cold and self-interacting dark matter. , keywords =. doi:10.1093/mnras/staf1357 , archivePrefix =. 2505.07944 , primaryClass =

    work page doi:10.1093/mnras/staf1357

  62. [62]

    , keywords =

    High Significance Detection of the Dark Substructure in Gravitational Lens SDSS J0946+1006 by Image Pixel Supersampling. , keywords =. doi:10.3847/1538-4357/adb1b6 , archivePrefix =. 2408.11090 , primaryClass =

    work page doi:10.3847/1538-4357/adb1b6

  63. [63]

    Probing dark matter structure down to $10^7$ solar masses: flux ratio statistics in gravitational lenses with line of sight halos

    Probing dark matter structure down to 10 ^ 7 solar masses: flux ratio statistics in gravitational lenses with line-of-sight haloes. , keywords =. doi:10.1093/mnras/stz1593 , archivePrefix =. 1901.11031 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stz1593 1901

  64. [64]

    New constraints on the dark matter free-streaming properties and substructure abundance from gravitationally lensed quasars

    SHARP - VII. New constraints on the dark matter free-streaming properties and substructure abundance from gravitationally lensed quasars. , keywords =. doi:10.1093/mnras/stz3177 , archivePrefix =. 1905.04182 , primaryClass =

    work page doi:10.1093/mnras/stz3177 1905

  65. [65]

    JWST lensed quasar dark matter survey IV: Stringent warm dark matter constraints from the joint reconstruction of extended lensed arcs and quasar flux ratios

    JWST lensed quasar dark matter survey IV: Stringent warm dark matter constraints from the joint reconstruction of extended lensed arcs and quasar flux ratios. arXiv e-prints , keywords =. doi:10.48550/arXiv.2511.07513 , archivePrefix =. 2511.07513 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2511.07513

  66. [66]

    , keywords =

    Strong-lensing source reconstruction with variationally optimized Gaussian processes. , keywords =. doi:10.1093/mnras/stac311 , archivePrefix =. 2105.09465 , primaryClass =

    work page doi:10.1093/mnras/stac311

  67. [67]

    arXiv e-prints , keywords =

    JWST Lensed Quasar Dark Matter Survey III: Dark Matter Sensitive Flux Ratios and Warm Dark Matter Constraint from the Full Sample. arXiv e-prints , keywords =. doi:10.48550/arXiv.2511.07765 , archivePrefix =. 2511.07765 , primaryClass =

    work page doi:10.48550/arxiv.2511.07765

  68. [68]

    , keywords =

    The ``Little Dark Dot'': Evidence for Self-interacting Dark Matter in the Strong Lens SDSS J0946+1006?. , keywords =. doi:10.3847/1538-4357/ae1462 , archivePrefix =. 2504.11800 , primaryClass =

    work page doi:10.3847/1538-4357/ae1462

  69. [69]

    , keywords =

    Not So Dark, Not So Dense: An Alternative Explanation for the Lensing Subhalo in SDSS J0946+1006. , keywords =. doi:10.3847/2041-8213/ae072d , archivePrefix =. 2506.07978 , primaryClass =

    work page doi:10.3847/2041-8213/ae072d 2041

  70. [70]

    A Universal Density Profile from Hierarchical Clustering

    A Universal Density Profile from Hierarchical Clustering. , keywords =. doi:10.1086/304888 , archivePrefix =. astro-ph/9611107 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1086/304888

  71. [71]

    , keywords =

    Exploiting the diversity of modeling methods to probe systematic biases in strong lensing analyses. , keywords =. doi:10.1051/0004-6361/202451095 , archivePrefix =. 2406.08484 , primaryClass =

    work page doi:10.1051/0004-6361/202451095

  72. [72]

    The Strong Lensing Discovery Engine C: Finding lenses with machine learning

    Euclid Quick Data Release (Q1). The Strong Lensing Discovery Engine C: Finding lenses with machine learning. arXiv e-prints , keywords =. doi:10.48550/arXiv.2503.15326 , archivePrefix =. 2503.15326 , primaryClass =

    work page doi:10.48550/arxiv.2503.15326