pith. sign in

arxiv: 2603.06339 · v2 · submitted 2026-03-06 · 🌌 astro-ph.CO

Enhancing Gravitational Lens Study with Deep Learning: A Study on Effects of Dropout Regularization

Juan J. Ancona-Flores , A. Hern\'andez-Almada , V. Motta This is my paper

Pith reviewed 2026-05-15 15:06 UTC · model grok-4.3

classification 🌌 astro-ph.CO
keywords strong gravitational lensingdeep learningdropout regularizationconvolutional neural networkssingular isothermal ellipsoidparameter inferencesynthetic images
0
0 comments X

The pith

Dropout in CNNs cuts errors in strong-lens SIE parameter inference by 60-76 percent.

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

The paper shows that adding dropout layers to a convolutional neural network based on AlexNet markedly improves the recovery of Singular Isothermal Ellipsoid parameters from simulated galaxy-galaxy strong-lensing images. Training on 76,396 synthetic images derived from a telescope catalog and testing three dropout settings, the authors find that dropout raises coefficients of determination to approximately 0.96 while cutting relative errors by 60-76 percent relative to the no-dropout case. A sympathetic reader would care because strong lensing is a direct probe of galaxy mass distributions and dark matter, yet traditional modeling is too slow for the data volumes expected from next-generation surveys. The work therefore tests whether a standard regularization technique can make deep learning reliable enough for routine scientific use in this domain.

Core claim

The central claim is that dropout regularization is critical for enhancing the precision and robustness of estimated SIE parameters from synthetic galaxy-galaxy lens systems, as demonstrated by 4-fold cross-validation: dropout configurations yield R² values up to approximately 0.96 for most parameters, mean Peak Signal-to-Noise Ratios up to approximately 37 dB, and relative errors reduced by 60-76 percent to at most approximately 9 percent at the 90 percent confidence level for the majority of parameters.

What carries the argument

Modified AlexNet convolutional neural network with three distinct dropout configurations, trained to regress four SIE parameters (Einstein radius, axis ratio, and two ellipticity components) from synthetic strong-lensing images.

If this is right

  • Dropout is necessary to achieve high precision and robustness in the inferred SIE parameters.
  • Deep learning with dropout enables scalable, computationally efficient modeling of large numbers of strong-lensing systems.
  • Parameter errors remain at most approximately 9 percent at 90 percent for most parameters when dropout is used.
  • The approach supports high-precision inference of galaxy mass distributions from lensing data.

Where Pith is reading between the lines

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

  • If the synthetic training distribution matches real data, the same architecture could supply rapid initial parameter estimates that speed up full modeling of observed lenses.
  • Higher parameter accuracy could tighten statistical constraints on dark-matter halo properties derived from lens samples.
  • Applying the model to real survey data would reveal how domain shift between simulations and observations limits performance.

Load-bearing premise

The synthetic images generated from the China Space Station Telescope catalog and the Singular Isothermal Ellipsoid profile faithfully represent the statistical properties of real observed strong-lensing systems.

What would settle it

Running the trained model on a set of real observed strong-lensing systems and comparing the inferred SIE parameters against independent measurements obtained from traditional lens-modeling software would test whether the reported error reductions hold outside the simulated domain.

Figures

Figures reproduced from arXiv: 2603.06339 by A. Hern\'andez-Almada, Juan J. Ancona-Flores, V. Motta.

Figure 1
Figure 1. Figure 1: FIG. 1: Sketch not drawn to scale, illustrates the parameters involved in a typical [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2: Illustrative architecture of a CNN model. The design incorporates convolution [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3: Marginal distributions of the training parameters of the SIE lens model. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4: Random examples of synthetic images of galaxy-galaxy lens systems used in [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5: The CNN that we use consists of a series of convolution groups interspersed with [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6: The performance metrics of the three models that were analyzed during the [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7: PSNR values for all the images used in test where the red dotted line indicates the [PITH_FULL_IMAGE:figures/full_fig_p016_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8: Random example of a galaxy-galaxy lens systems. Left column shows the original [PITH_FULL_IMAGE:figures/full_fig_p017_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9: Predicted vs. true parameters for the lens SIE model (Einstein radius, axis ratio, [PITH_FULL_IMAGE:figures/full_fig_p018_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10: Relative errors of the SIE parameters for the three CNN models. Median values [PITH_FULL_IMAGE:figures/full_fig_p019_10.png] view at source ↗
read the original abstract

Strong gravitational lensing provides valuable insights into the mass distribution of galaxies and the nature of dark matter. However, its modeling is computationally demanding due to the large volume of strong lensing observations. In this work, we explore the application of Convolutional Neural Networks to infer physical parameters from simulated galaxy-galaxy lens systems, described by the Singular Isothermal Ellipsoid (SIE) profile for the galaxy lens. We construct a dataset of 76,396 synthetic lensing images derived from the China Space Station Telescope catalog and employ it to train a modified CNN model, based on AlexNet architecture, to predict four key SIE parameters, Einstein radius, axis ratio and ellipticity components. We analyze the network performance under three distinct dropout configurations to quantify their influence on generalization and parameter inference accuracy. The results indicate that the incorporation of dropout is critical for enhancing the precision and robustness of the estimated parameters, as demonstrated using a 4-fold cross-validation procedure. When dropout tools are included we obtain yields coefficients of determination up to $R^2 \sim 0.96$ for most SIE parameters and mean Peak Signal-to-Noise Ratios of up to $\sim 37$ dB. Relative to the configuration without dropout, the use of dropout reduces the relative errors in the inferred SIE parameters by approximately $60-76\%$, resulting in errors of at most $\sim 9\%$ at the $90\%$ confidence level for the majority of parameters. These findings highlight the potential of deep learning approaches to enable scalable, computationally efficient, and high-precision modeling of strong gravitational lensing systems.

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

3 major / 2 minor

Summary. The manuscript describes the use of a modified AlexNet convolutional neural network to predict four Singular Isothermal Ellipsoid (SIE) lens parameters (Einstein radius, axis ratio, and ellipticity components) from 76,396 simulated galaxy-galaxy strong lensing images generated from the China Space Station Telescope catalog. It evaluates three dropout configurations via 4-fold cross-validation on held-out synthetic data, claiming that dropout yields R² up to ~0.96, mean PSNR up to ~37 dB, and reduces relative parameter errors by 60-76% (to ≤9% at 90% CL) relative to the no-dropout baseline.

Significance. If the results hold under the stated assumptions, the work provides quantitative evidence that dropout regularization improves generalization and accuracy in CNN-based inference of SIE parameters on simulated strong-lensing data. This could support scalable analysis pipelines for large upcoming surveys such as CSST, where traditional modeling is computationally intensive. The reported use of 4-fold CV and standard metrics (R², relative error, PSNR) on independent validation folds strengthens the internal validity of the dropout-benefit claim within the simulation framework.

major comments (3)
  1. The central performance claims (R² ∼ 0.96, 60-76% error reduction) rest entirely on synthetic images generated under the SIE profile assumption from the CSST catalog; no transfer tests to real observed lenses or to non-SIE profiles (e.g., NFW or composite models) are reported, which is load-bearing for the broader claim of enhancing gravitational lens studies.
  2. Insufficient detail is provided on the exact modifications to the AlexNet architecture, the specific dropout probabilities and placement in the three configurations, and the full data-generation pipeline (noise model, source properties, and image rendering steps), preventing independent reproduction and verification of the reported gains.
  3. No baseline comparisons to traditional lens-modeling codes or to other machine-learning approaches are included, making it difficult to quantify the practical advantage of the dropout-enhanced CNN over existing methods.
minor comments (2)
  1. Clarify the phrase 'dropout tools' in the abstract to 'dropout layers' or 'dropout regularization' for precision.
  2. Ensure that all reported confidence intervals (e.g., 90% CL) are accompanied by explicit descriptions of how they were computed from the 4-fold CV folds.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for the constructive comments, which have helped us identify areas to strengthen the manuscript. We address each major comment below and outline the revisions we will make.

read point-by-point responses

  1. Referee: The central performance claims (R² ∼ 0.96, 60-76% error reduction) rest entirely on synthetic images generated under the SIE profile assumption from the CSST catalog; no transfer tests to real observed lenses or to non-SIE profiles (e.g., NFW or composite models) are reported, which is load-bearing for the broader claim of enhancing gravitational lens studies.

    Authors: We agree that the study is restricted to simulated SIE-profile images and does not include transfer tests on real lenses or alternative profiles such as NFW. This is a genuine limitation of the current work, which was designed to isolate the effect of dropout within a controlled simulation framework. In the revised manuscript we will add an explicit Limitations and Future Work subsection that states the scope of the claims, notes the absence of real-data validation, and outlines the need for subsequent studies on observed lenses and more complex mass models. revision: yes

  2. Referee: Insufficient detail is provided on the exact modifications to the AlexNet architecture, the specific dropout probabilities and placement in the three configurations, and the full data-generation pipeline (noise model, source properties, and image rendering steps), preventing independent reproduction and verification of the reported gains.

    Authors: We acknowledge that the original manuscript lacks sufficient technical detail for full reproducibility. The revised version will expand the Methods section to specify: (i) the precise architectural modifications to AlexNet, (ii) the dropout probabilities and layer placements for each of the three configurations, and (iii) the complete data-generation pipeline, including the noise model, source galaxy properties, and image rendering steps. revision: yes

  3. Referee: No baseline comparisons to traditional lens-modeling codes or to other machine-learning approaches are included, making it difficult to quantify the practical advantage of the dropout-enhanced CNN over existing methods.

    Authors: The manuscript’s primary objective is to quantify the isolated contribution of dropout regularization rather than to perform a comprehensive benchmark. We will nevertheless add a concise discussion section that places our R² and error-reduction figures in the context of previously published results from both traditional lens-modeling codes and other CNN-based approaches, while explicitly noting that a dedicated comparative study lies outside the present scope. revision: partial

standing simulated objections not resolved
  • We cannot supply transfer tests on real observed lenses or non-SIE profiles within the current study, as all experiments were performed exclusively on the simulated SIE dataset described in the manuscript.

Circularity Check

0 steps flagged

No significant circularity; empirical ML metrics on held-out synthetic data

full rationale

The paper trains a modified AlexNet CNN on 76,396 synthetic SIE lens images generated from the CSST catalog and reports R², PSNR, and relative error metrics obtained via 4-fold cross-validation on held-out folds. These quantities are computed directly from the network's predictions versus ground-truth SIE parameters on validation data; they are not defined in terms of the fitted weights or inputs by construction, nor do they reduce to any self-referential loop. No equations, uniqueness theorems, or ansatzes are invoked that would make the reported performance gains tautological. The only potential concern is the external validity of the synthetic ensemble (noted in the skeptic headline), but that is a question of generalization, not circularity in the derivation chain. The central claims remain independent empirical measurements.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that the SIE profile and the catalog-derived simulations capture the relevant statistics of real lenses; dropout rates are treated as tunable hyperparameters rather than derived quantities.

free parameters (1)
  • dropout probabilities
    Three distinct dropout configurations are tested; their specific rates are chosen to optimize generalization on the simulated set.
axioms (1)
  • domain assumption The Singular Isothermal Ellipsoid profile is an adequate description of the mass distribution in the simulated galaxy-galaxy lens systems.
    All parameter inference targets the four SIE quantities, presupposing that real lenses are well approximated by this model.

pith-pipeline@v0.9.0 · 5605 in / 1322 out tokens · 53993 ms · 2026-05-15T15:06:31.196380+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

84 extracted references · 84 canonical work pages · 5 internal anchors

  1. [1]

    The SIE model profile is fundamentally defined by the line-sight velocity dispersion of the lens galaxy,σ v

    Under this prescription, the projected density profile and the corresponding surface mass density (Σ(⃗ξ)) become ρ(⃗ r) = σ2 v 2πGr2 ,(6) Σ(⃗ξ) = σ2 v 2G √fp ξ2 1 +f 2ξ2 2 .(7) In this context,frefers to the axis ratio of the ellipses, specified within the interval 0< f≤1. The SIE model profile is fundamentally defined by the line-sight velocity dispersio...

    work page 2012

  2. [2]

    Regarding image reconstruction performance, Figure 7 reports the PSNR distributions, with median values of 36.9 dB, 36.8 dB, and 29.2 dB for models 1, 2, and 3, respectively. When dropout is incorporated into the models, these values place the reconstructed im- 16 0 25 50 75 100 pixels 0 25 50 75 100pixels True 0 25 50 75 100 pixels 0 25 50 75 100pixels P...

    work page

  3. [3]

    Schneider, C

    P. Schneider, C. Kochanek, and J. Wambsganss,Gravitational lensing: strong, weak and micro: Saas-Fee advanced course 33, Vol. 33 (Springer Science & Business Media, 2006)

    work page 2006

  4. [4]

    Schneider,Extragalactic Astronomy and Cosmology: An Introduction(Springer Berlin Hei- delberg, 2015)

    P. Schneider,Extragalactic Astronomy and Cosmology: An Introduction(Springer Berlin Hei- delberg, 2015)

    work page 2015

  5. [5]

    Y. D. Hezaveh, N. Dalal, D. P. Marrone, Y.-Y. Mao, W. Morningstar, D. Wen, R. D. Bland- ford, J. E. Carlstrom, C. D. Fassnacht, G. P. Holder, A. Kemball, P. J. Marshall, N. Murray, L. P. Levasseur, J. D. Vieira, and R. H. Wechsler, The Astrophysical Journal823, 37 (2016)

    work page 2016

  6. [6]

    Meneghetti,Introduction to gravitational lensing: with Python examples, Vol

    M. Meneghetti,Introduction to gravitational lensing: with Python examples, Vol. 956 (Springer Nature, 2021)

    work page 2021

  7. [7]

    A. J. Shajib, G. Vernardos, T. E. Collett, V. Motta, D. Sluse, L. L. R. Williams, P. Saha, S. Birrer, C. Spiniello, and T. Treu, Space Science Reviews220, 87 (2024)

    work page 2024

  8. [8]

    Treu, Annual Review of Astronomy and Astrophysics48, 87125 (2010)

    T. Treu, Annual Review of Astronomy and Astrophysics48, 87125 (2010). 20

    work page 2010

  9. [9]

    Natarajan, L

    P. Natarajan, L. L. R. Williams, M. Bradaˇ c, C. Grillo, A. Ghosh, K. Sharon, and J. Wagner, Space Science Reviews220, 19 (2024)

    work page 2024

  10. [10]

    Grillo, M

    C. Grillo, M. Lombardi, and G. Bertin, Astronomy & Astrophysics477, 397 (2008)

    work page 2008

  11. [11]

    Chae, Monthly Notices of the Royal Astronomical Society346, 746 (2003)

    K.-H. Chae, Monthly Notices of the Royal Astronomical Society346, 746 (2003)

    work page 2003

  12. [12]

    Birrer, M

    S. Birrer, M. Millon, D. Sluse, A. J. Shajib, F. Courbin, S. Erickson, L. V. E. Koopmans, S. H. Suyu, and T. Treu, Space Science Reviews220, 48 (2024)

    work page 2024

  13. [13]

    Collaboration, Y

    E. Collaboration, Y. Mellier, and et. al., Astronomy and Astrophysics697, A1 (2025)

    work page 2025

  14. [14]

    Euclid Definition Study Report

    R. a. a. Laureijs, “Euclid Definition Study Report,” (2011), arXiv:1110.3193 [astro-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  15. [15]

    Enard, Journal of Optics22, 33 (1991)

    D. Enard, Journal of Optics22, 33 (1991)

    work page 1991

  16. [16]

    M. W. e. a. McElwain, Publications of the Astronomical Society of the Pacific135, 058001 (2023), arXiv:2301.01779 [astro-ph]

    work page arXiv 2023

  17. [17]

    J. P. Gardner, J. C. Mather, M. Clampin, R. Doyon, M. A. Greenhouse, H. B. Hammel, J. B. Hutchings, P. Jakobsen, S. J. Lilly, K. S. Long, J. I. Lunine, M. J. Mccaughrean, M. Mountain, J. Nella, G. H. Rieke, M. J. Rieke, H.-W. Rix, E. P. Smith, G. Sonneborn, M. Stiavelli, H. S. Stockman, R. A. Windhorst, and G. S. Wright, Space Science Reviews123, 485 (2006)

    work page 2006

  18. [18]

    F. B. e. a. Bianco, The Astrophysical Journal Supplement Series258, 1 (2021)

    work page 2021

  19. [19]

    Hook, inScience with the VLT in the ELT Era, edited by A

    I. Hook, inScience with the VLT in the ELT Era, edited by A. Moorwood (Springer Nether- lands, Dordrecht, 2009) pp. 225–232

    work page 2009

  20. [20]

    E. e. a. Palle, Experimental Astronomy59, 29 (2025)

    work page 2025

  21. [21]

    Gong and et

    Y. Gong and et. al. (CSST Collaboration), Science China Physics, Mechanics & Astronomy 69, 239501 (2026)

    work page 2026

  22. [22]

    T. E. Collett, The Astrophysical Journal811, 20 (2015)

    work page 2015

  23. [23]

    S. T. Myers, N. J. Jackson, I. W. A. Browne, A. G. de Bruyn, T. J. Pearson, A. C. S. Readhead, P. N. Wilkinson, A. D. Biggs, R. D. Blandford, C. D. Fassnacht, L. V. E. Koopmans, D. R. Marlow, J. P. McKean, M. A. Norbury, P. M. Phillips, D. Rusin, M. C. Shepherd, and C. M. Sykes, Monthly Notices of the Royal Astronomical Society341, 1 (2003)

    work page 2003

  24. [24]

    I. W. A. Browne, P. N. Wilkinson, N. J. F. Jackson, S. T. Myers, C. D. Fassnacht, L. V. E. Koopmans, D. R. Marlow, M. Norbury, D. Rusin, C. M. Sykes, A. D. Biggs, R. D. Blandford, A. G. de Bruyn, K.-H. Chae, P. Helbig, L. J. King, J. P. McKean, T. J. Pearson, P. M. Phillips, A. C. S. Readhead, E. Xanthopoulos, and T. York, Monthly Notices of the Royal Ast...

    work page 2003

  25. [25]

    Scoville, H

    N. Scoville, H. Aussel, M. Brusa, P. Capak, C. M. Carollo, M. Elvis, M. Giavalisco, L. Guzzo, G. Hasinger, C. Impey, J.-P. Kneib, O. LeFevre, S. J. Lilly, B. Mobasher, A. Renzini, R. M. Rich, D. B. Sanders, E. Schinnerer, D. Schminovich, P. Shopbell, Y. Taniguchi, and N. D. Tyson, The Astrophysical Journal Supplement Series172, 1 (2007)

    work page 2007

  26. [26]

    A. S. Bolton, S. Burles, L. V. E. Koopmans, T. Treu, and L. A. Moustakas, The Astrophysical Journal638, 703 (2006)

    work page 2006

  27. [27]

    A. S. Bolton, S. Burles, L. V. E. Koopmans, T. Treu, R. Gavazzi, L. A. Moustakas, R. Wayth, and D. J. Schlegel, The Astrophysical Journal682, 964 (2008)

    work page 2008

  28. [28]

    Fowlie, W

    A. Fowlie, W. Handley, and L. Su, Monthly Notices of the Royal Astronomical Society497, 5256 (2020)

    work page 2020

  29. [29]

    Zhang, A

    W. Zhang, A. Hasegawa, O. Matoba, K. Itoh, Y. Ichioka, and K. Doi, Applications of Artificial Neural Networks III1709, 257 (1992)

    work page 1992

  30. [30]

    Spectral classification using convolutional neural networks

    P. H´ ala, arXiv e-prints , arXiv:1412.8341 (2014), arXiv:1412.8341 [cs.CV]

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  31. [31]

    C. E. Petrillo, C. Tortora, S. Chatterjee, G. Vernardos, L. V. E. Koopmans, G. Verdoes Kleijn, N. R. Napolitano, G. Covone, P. Schneider, A. Grado, and J. McFarland, Monthly Notices 21 of the Royal Astronomical Society472, 1129 (2017)

    work page 2017

  32. [32]

    Y. D. Hezaveh, L. P. Levasseur, and P. J. Marshall, Nature548, 555 (2017)

    work page 2017

  33. [33]

    Schuldt, S

    S. Schuldt, S. H. Suyu, T. Meinhardt, L. Leal-Taix´ e, R. Ca˜ nameras, S. Taubenberger, and A. Halkola, Astronomy & Astrophysics646, A126 (2021)

    work page 2021

  34. [34]

    W. R. Morningstar, L. P. Levasseur, Y. D. Hezaveh, R. Blandford, P. Marshall, P. Putzky, T. D. Rueter, R. Wechsler, and M. Welling, The Astrophysical Journal883, 14 (2019)

    work page 2019

  35. [35]

    J. W. Park, S. Wagner-Carena, S. Birrer, P. J. Marshall, J. Y.-Y. Lin, A. Roodman, and T. L. D. E. S. Collaboration), The Astrophysical Journal910, 39 (2021)

    work page 2021

  36. [36]

    Pearson, J

    J. Pearson, J. Maresca, N. Li, and S. Dye, Monthly Notices of the Royal Astronomical Society 505, 4362 (2021)

    work page 2021

  37. [37]

    Busillo and et

    V. Busillo and et. al., Astronomy & Astrophysics (2026), 10.1051/0004-6361/202554538, arXiv:2503.15329 [astro-ph.CO]

    work page doi:10.1051/0004-6361/202554538 2026

  38. [38]

    Parlange, J

    R. Parlange, J. C. Cuevas-Tello, O. Valenzuela, O. d. J. Cabrera-Rosas, T. Verdugo, A. More, and A. T. Jaelani, Monthly Notices of the Royal Astronomical Society545, staf1747 (2025), https://academic.oup.com/mnras/article-pdf/545/2/staf1747/64612524/staf1747.pdf

    work page 2025

  39. [39]

    Srivastava, G

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, J. Mach. Learn. Res.15, 1929 (2014)

    work page 1929

  40. [40]

    Hinton, A

    G. Hinton, A. Krizhevsky, I. Sutskever, and Y. Rachmad, Advances in Neural Information Processing Systems , 1097 (2012)

    work page 2012

  41. [41]

    Kormann, P

    R. Kormann, P. Schneider, and M. Bartelmann, Astronomy and Astrophysics284, 285 (1994)

    work page 1994

  42. [42]

    Tessore and R

    N. Tessore and R. B. Metcalf, Astronomy & Astrophysics580, A79 (2015)

    work page 2015

  43. [43]

    Asada, T

    H. Asada, T. Hamana, and M. Kasai, Astronomy & Astrophysics397, 825 (2003)

    work page 2003

  44. [44]

    Etherington, J

    A. Etherington, J. W. Nightingale, R. Massey, X. Cao, A. Robertson, N. C. Amorisco, A. Amvrosiadis, S. Cole, C. S. Frenk, Q. He, R. Li, and S.-I. Tam, Monthly Notices of the Royal Astronomical Society517, 3275 (2022)

    work page 2022

  45. [45]

    J. L. S´ ersic, Boletin de la Asociacion Argentina de Astronomia La Plata Argentina6, 41 (1963)

    work page 1963

  46. [46]

    J. L. Sersic, Cordoba (1968)

    work page 1968

  47. [47]

    Analytical properties of the R^(1/m) luminosity law

    L. Ciotti and G. Bertin, “Analytical properties of ther (1/m) luminosity law,” (1999), arXiv:astro-ph/9911078 [astro-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 1999

  48. [48]

    V. F. Cardone, Astronomy & Astrophysics415, 839 (2004)

    work page 2004

  49. [49]

    Z. Li, F. Liu, W. Yang, S. Peng, and J. Zhou, IEEE Transactions on Neural Networks and Learning Systems33, 6999 (2022)

    work page 2022

  50. [50]

    Alzubaidi, J

    L. Alzubaidi, J. Zhang, A. J. Humaidi, A. Al-Dujaili, Y. Duan, O. Al-Shamma, J. Santamar´ ıa, M. A. Fadhel, M. Al-Amidie, and L. Farhan, Journal of Big Data8, 53 (2021)

    work page 2021

  51. [51]

    Nair and G

    V. Nair and G. E. Hinton, inProceedings of the 27th International Conference on International Conference on Machine Learning, ICML’10 (Omnipress, Madison, WI, USA, 2010) pp. 807– 814

    work page 2010

  52. [52]

    Glorot, A

    X. Glorot, A. Bordes, and Y. Bengio, inProceedings of the Fourteenth International Con- ference on Artificial Intelligence and Statistics, Proceedings of Machine Learning Research, Vol. 15, edited by G. Gordon, D. Dunson, and M. Dudk (PMLR, Fort Lauderdale, FL, USA,

    work page

  53. [53]

    LeCun, B

    Y. LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, and L. D. Jackel, Neural computation1, 541 (1989)

    work page 1989

  54. [54]

    Baldi and P

    P. Baldi and P. J. Sadowski, inAdvances in Neural Information Processing Systems, Vol. 26 (Curran Associates, Inc., 2013). 22

    work page 2013

  55. [55]

    Improving neural networks by preventing co-adaptation of feature detectors

    G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, and R. R. Salakhutdinov, “Improving neural networks by preventing co-adaptation of feature detectors,” (2012), arXiv:1207.0580 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  56. [56]

    Salehin and D.-K

    I. Salehin and D.-K. Kang, Electronics12(2023), 10.3390/electronics12143106

    work page doi:10.3390/electronics12143106 2023

  57. [57]

    X. Cao, R. Li, N. Li, R. Li, Y. Chen, K. Ding, H. Shan, H. Zhan, X. Zhang, W. Du, and S. Cao, Monthly Notices of the Royal Astronomical Society533, 1960 (2024), https://academic.oup.com/mnras/article-pdf/533/2/1960/58913741/stae1865.pdf

    work page 1960

  58. [58]

    Birrer and A

    S. Birrer and A. Amara, Physics of the Dark Universe22, 189 (2018)

    work page 2018

  59. [59]

    Mukherjee, L

    S. Mukherjee, L. V. E. Koopmans, R. B. Metcalf, N. Tessore, C. Tortora, M. Schaller, J. Schaye, R. A. Crain, G. Vernardos, F. Bellagamba, and T. Theuns, Monthly Notices of the Royal Astronomical Society479, 4108 (2018), https://academic.oup.com/mnras/article- pdf/479/3/4108/25170031/sty1741.pdf

    work page 2018

  60. [60]

    Stone, Journal of the Royal Statistical Society: Series B (Methodological)36, 111 (2018), https://academic.oup.com/jrsssb/article-pdf/36/2/111/49096683/jrsssb 36 2 111.pdf

    M. Stone, Journal of the Royal Statistical Society: Series B (Methodological)36, 111 (2018), https://academic.oup.com/jrsssb/article-pdf/36/2/111/49096683/jrsssb 36 2 111.pdf

    work page 2018

  61. [61]

    Tfrecord and tf.train.example,

    TensorFlow Authors, “Tfrecord and tf.train.example,”https://www.tensorflow.org/ tutorials/load_data/tfrecord(2024)

    work page 2024

  62. [62]

    Ballester and R

    P. Ballester and R. Araujo, Proceedings of the AAAI Conference on Artificial Intelligence30 (2016), 10.1609/aaai.v30i1.10171

    work page doi:10.1609/aaai.v30i1.10171 2016

  63. [63]

    Zhang, in2021 2nd International Conference on Big Data & Artificial Intelligence & Software Engineering (ICBASE)(2021) pp

    X. Zhang, in2021 2nd International Conference on Big Data & Artificial Intelligence & Software Engineering (ICBASE)(2021) pp. 414–419

    work page 2021

  64. [64]

    Perreault Levasseur, Y

    L. Perreault Levasseur, Y. D. Hezaveh, and R. H. Wechsler, The Astrophysical Journal Letters 850, L7 (2017)

    work page 2017

  65. [65]

    Dozat, inProceedings of the 4th International Conference on Learning Representations, Workshop Track(San Juan, Puerto Rico, 2016) pp

    T. Dozat, inProceedings of the 4th International Conference on Learning Representations, Workshop Track(San Juan, Puerto Rico, 2016) pp. 1–4

    work page 2016

  66. [66]

    Sutskever, J

    I. Sutskever, J. Martens, G. Dahl, and G. Hinton, inProceedings of the 30th International Conference on Machine Learning(PMLR, 2013) pp. 1139–1147

    work page 2013

  67. [67]

    Cholletet al., “Keras,”https://keras.io(2015)

    F. Cholletet al., “Keras,”https://keras.io(2015)

    work page 2015

  68. [68]

    Goodfellow, Y

    I. Goodfellow, Y. Bengio, and A. Courville,Deep Learning(MIT Press, 2016)http://www. deeplearningbook.org

    work page 2016

  69. [69]

    Caruana, Machine Learning28, 4175 (1997)

    R. Caruana, Machine Learning28, 4175 (1997)

    work page 1997

  70. [70]

    The elements of statistical learning: data mining, inference, and prediction,

    D. Ruppert, “The elements of statistical learning: data mining, inference, and prediction,” (2004)

    work page 2004

  71. [71]

    Botchkarev, Interdisciplinary Journal of Information, Knowledge, and Management14, 045076 (2019)

    A. Botchkarev, Interdisciplinary Journal of Information, Knowledge, and Management14, 045076 (2019)

    work page 2019

  72. [72]

    Chicco, M

    D. Chicco, M. J. Warrens, and G. Jurman, PeerJ Computer Science7, e623 (2021)

    work page 2021

  73. [73]

    R. J. Hyndman and A. B. Koehler, International Journal of Forecasting22, 679 (2006)

    work page 2006

  74. [74]

    Robust statistics,

    P. J. Huber, “Robust statistics,” inInternational Encyclopedia of Statistical Science, edited by M. Lovric (Springer Berlin Heidelberg, Berlin, Heidelberg, 2011) pp. 1248–1251

    work page 2011

  75. [75]

    Sadykova and A

    D. Sadykova and A. P. James, in2017 International Conference on Advances in Computing, Communications and Informatics (ICACCI)(2017) pp. 2366–2369

    work page 2017

  76. [76]

    Petrovic, B

    V. Petrovic, B. Pavlovi, M. Andri, and B. Bondzulic, Electronics Letters52(2016), 10.1049/el.2015.3784

    work page doi:10.1049/el.2015.3784 2016

  77. [77]

    Lupton, M

    R. Lupton, M. R. Blanton, G. Fekete, D. W. Hogg, W. OMullane, A. Szalay, and N. Wherry, Publications of the Astronomical Society of the Pacific116, 133 (2004)

    work page 2004

  78. [78]

    Analyzing interferometric observations of strong gravitational lenses with recurrent and convolutional neural networks

    W. R. Morningstar, Y. D. Hezaveh, L. P. Levasseur, R. D. Blandford, P. J. Marshall, P. Putzky, and R. H. Wechsler, “Analyzing interferometric observations of strong gravitational lenses with 23 recurrent and convolutional neural networks,” (2018), arXiv:1808.00011 [astro-ph.IM]

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  79. [79]

    Gentile, C

    F. Gentile, C. Tortora, G. Covone, L. V. E. Koopmans, R. Li, L. Leuzzi, and N. R. Napolitano, Monthly Notices of the Royal Astronomical Society522, 54425455 (2023)

    work page 2023

  80. [80]

    Binney and S

    J. Binney and S. Tremaine,Galactic Dynamics, Princeton Series in Astrophysics (Princeton University Press, 1987)

    work page 1987

Showing first 80 references.