pith. sign in

arxiv: 2505.08089 · v2 · submitted 2025-05-12 · 🌀 gr-qc · astro-ph.IM

Assessment of normalizing flows for parameter estimation on time-frequency representations of gravitational-wave data

Daniel Lanchares , Osvaldo G. Freitas , Lysiane Mornas , Jos\'e A. Font , Joaqu\'in Gonz\'alez-Nuevo , Luigi Toffolatti , Pietro Vischia This is my paper

Pith reviewed 2026-05-22 15:13 UTC · model grok-4.3

classification 🌀 gr-qc astro-ph.IM
keywords gravitational wavesparameter estimationnormalizing flowsbinary black holesspectrogramsdeep learningLIGOVirgo
0
0 comments X

The pith

A normalizing flow model on RGB images of BBH spectrograms recovers LVK parameter estimates for most quantities.

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

The paper presents GP15, a method that converts time-frequency spectrograms of binary black hole gravitational-wave signals from LIGO and Virgo into RGB images and processes them with residual networks combined with normalizing flows. Trained exclusively on IMRPhenomXPHM simulations, the model is tested on all three-detector events from the GWTC-3 and GWTC-2.1 catalogs. It produces posterior samples in roughly one second and shows good agreement with the official LVK results over most parameters. A sympathetic reader would care because rapid sampling could complement slower traditional methods as the number of detected events grows.

Core claim

GP15 maps BBH spectrograms from the Advanced LIGO and Advanced Virgo detectors to color channels in an RGB image and uses residual networks with normalizing flows to estimate binary black hole parameters. The model is trained on IMRPhenomXPHM waveform simulations and tested on real three-detector events reported in the GWTC-3 and GWTC-2.1 catalogs. It yields good agreement with LVK results over most parameters and can generate large numbers of posterior samples in about one second.

What carries the argument

GP15, which converts BBH spectrograms to RGB images and applies residual networks together with normalizing flows to produce parameter posteriors

If this is right

  • Large numbers of posterior samples can be generated in roughly one second.
  • The approach complements existing normalizing flow methods that operate directly on time or frequency representations.
  • Shortcomings such as the lack of detector PSD noise conditioning limit current performance.
  • Splitting the parameter space into intrinsic and extrinsic subspaces offers one route for future refinement.

Where Pith is reading between the lines

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

  • The image-based format might allow the same architecture to incorporate additional data channels such as strain amplitude alongside the spectrogram.
  • Retraining on a broader set of waveform approximants could reduce systematic differences when the model is applied to events whose signals deviate from IMRPhenomXPHM.
  • Integration with existing LVK pipelines could provide rapid initial posteriors that are later refined by slower samplers.

Load-bearing premise

Converting BBH spectrograms into RGB images and training exclusively on IMRPhenomXPHM simulations is sufficient to recover accurate posteriors on real detector data without explicit noise conditioning.

What would settle it

A side-by-side comparison of the model's posterior distributions against LVK results for mass ratio, effective spin, or luminosity distance on a fresh set of real events would reveal whether agreement holds or breaks for specific parameters.

Figures

Figures reproduced from arXiv: 2505.08089 by Daniel Lanchares, Joaqu\'in Gonz\'alez-Nuevo, Jos\'e A. Font, Luigi Toffolatti, Lysiane Mornas, Osvaldo G. Freitas, Pietro Vischia.

Figure 1
Figure 1. Figure 1: FIG. 1: Example of time-frequency representation of a [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2: Model diagram depicting the flow of information [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The average training loss decreases monotonically [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4: Jensen-Shannon divergences for all three-detector events of GWTC-2.1 [5] with posterior distributions match [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5: Jensen-Shannon divergences for all three-detector events of GWTC-3 [4] with posterior distributions matching [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6: Comparison (90% credible intervals) between the estimations of all compatible events and parameters of [31]. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
read the original abstract

The speed-up of parameter estimation is an active field of research in gravitational-wave data analysis. In this paper we present GP15, a deep-learning method that merges residual networks and normalizing flows into a general-purpose, image-based estimator of binary black hole (BBH) parameters. Building on our early work, we map BBH spectrograms from the Advanced LIGO and Advanced Virgo detectors to color channels in an RGB image amenable to be processed with residual networks. GP15 is trained on simulated data for BBH mergers obtained with the \texttt{IMRPhenomXPHM} waveform approximant and tested for all three-detector events from the GWTC-3 and GWTC-2.1 catalogs reported by the LIGO-Virgo-KAGRA (LVK) collaboration. Overall, our model yields good agreement with the LVK results over most parameters. Our simple model can produce large amounts of posterior samples in the order of a second, complementing existing approaches with normalizing flows based on time or frequency representation of gravitational-wave data. We also discuss current shortcomings of our model and possible improvements for future extensions (e.g. including noise conditioning from the detectors' PSD or splitting the parameter space into intrinsic and extrinsic subspaces).

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 GP15, a deep-learning method that converts BBH spectrograms from Advanced LIGO and Advanced Virgo into RGB images and employs residual networks combined with normalizing flows to estimate binary black hole parameters. The model is trained exclusively on noise-free simulations generated with the IMRPhenomXPHM waveform approximant and evaluated on three-detector events from the GWTC-3 and GWTC-2.1 catalogs, with the central claim being overall good agreement with LVK collaboration results alongside rapid generation of large numbers of posterior samples in approximately one second.

Significance. If validated with quantitative metrics, this image-based normalizing flow approach offers a potentially useful complement to existing time- or frequency-domain machine-learning methods for gravitational-wave parameter estimation, particularly for its speed in producing posterior samples. The work acknowledges current limitations and suggests future directions such as noise conditioning, which helps frame its scope as an initial assessment rather than a complete replacement for standard Bayesian pipelines.

major comments (2)
  1. [Abstract] Abstract: The claim that the model 'yields good agreement with the LVK results over most parameters' is presented without any quantitative metrics (e.g., median errors, credible-interval coverage fractions, or posterior overlap measures), error budgets, or ablation studies. This absence makes it impossible to assess whether the reported agreement is statistically meaningful or affected by post-hoc selections.
  2. [Methods] Methods (training procedure): The model is trained solely on clean IMRPhenomXPHM simulations without any conditioning on detector PSDs or marginalization over noise realizations. Real GWTC events contain colored Gaussian noise whose statistics vary by event and detector; the resulting distribution shift is load-bearing for the claim of applicability to real data, especially for extrinsic parameters and events near detection thresholds. The abstract itself identifies noise conditioning as a required future extension.
minor comments (1)
  1. [Abstract] Abstract: The description of the spectrogram-to-RGB conversion lacks sufficient detail on channel assignment, frequency scaling, and time-frequency resolution to allow independent reproduction of the input representation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have helped clarify the scope and presentation of our work. We address each major comment below and outline the revisions we will make to the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that the model 'yields good agreement with the LVK results over most parameters' is presented without any quantitative metrics (e.g., median errors, credible-interval coverage fractions, or posterior overlap measures), error budgets, or ablation studies. This absence makes it impossible to assess whether the reported agreement is statistically meaningful or affected by post-hoc selections.

    Authors: We agree that the abstract statement would be more precise if supported by explicit quantitative references. In the revised manuscript we will update the abstract to note that agreement is assessed via direct comparison of median posterior values and 90% credible-interval widths for the full set of three-detector events in GWTC-2.1 and GWTC-3, with the complete event list and all parameter comparisons shown in the results section (Figures 3–5 and associated tables). We will also state that no post-hoc event selection was performed. These changes will allow readers to evaluate the strength of the agreement directly from the reported figures and tables. revision: yes

  2. Referee: [Methods] Methods (training procedure): The model is trained solely on clean IMRPhenomXPHM simulations without any conditioning on detector PSDs or marginalization over noise realizations. Real GWTC events contain colored Gaussian noise whose statistics vary by event and detector; the resulting distribution shift is load-bearing for the claim of applicability to real data, especially for extrinsic parameters and events near detection thresholds. The abstract itself identifies noise conditioning as a required future extension.

    Authors: We concur that the absence of noise conditioning and PSD marginalization constitutes a genuine limitation when applying the model to real data. The manuscript already frames the work as an initial assessment and explicitly lists noise conditioning as a required future extension. In the revision we will expand the methods section to describe the clean-simulation training set in greater detail and add a dedicated paragraph in the discussion that addresses the expected impact of the distribution shift, particularly on extrinsic parameters. We will also reiterate that the current results on GWTC events should be viewed as a proof-of-concept demonstration rather than a fully noise-robust pipeline. revision: yes

Circularity Check

0 steps flagged

No significant circularity; validation relies on external LVK benchmarks

full rationale

The paper presents an empirical ML pipeline: spectrograms from IMRPhenomXPHM simulations are converted to RGB images, a residual-network + normalizing-flow model is trained, and posteriors are generated for real GWTC events then compared directly to independent LVK catalog results. No equations, fitted parameters, or self-citations reduce the reported agreement to a quantity defined by the model's own inputs. The comparison to LVK provides external falsifiability outside any fitted values in the present work. The mention of 'building on our early work' for the RGB mapping is a minor self-citation that does not bear the central performance claim.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The approach depends on standard machine-learning assumptions plus domain-specific modeling choices whose validity is not independently verified in the provided abstract.

free parameters (1)
  • neural-network weights and hyperparameters
    The residual network and normalizing-flow parameters are learned from simulated data; their specific values are not reported.
axioms (1)
  • domain assumption Simulated BBH signals generated with IMRPhenomXPHM are statistically representative of real Advanced LIGO/Virgo observations for the purpose of training and validation.
    The entire training and testing pipeline rests on this assumption.

pith-pipeline@v0.9.0 · 5785 in / 1385 out tokens · 35875 ms · 2026-05-22T15:13:39.365963+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

75 extracted references · 75 canonical work pages · 14 internal anchors

  1. [1]

    B. P. Abbott, R. Abbott, T. D. Abbott, M. R. Aber- nathy, F. Acernese, K. Ackley, C. Adams, T. Adams, P. Addesso, R. X. Adhikari, et al. (LIGO Scientific Col- laboration and Virgo Collaboration), Phys. Rev. Lett. 116, 061102 (2016), URLhttps://link.aps.org/doi/ 10.1103/PhysRevLett.116.061102

    work page doi:10.1103/physrevlett.116.061102 2016

  2. [2]

    B. P. Abbott, R. Abbott, T. D. Abbott, S. Abra- ham, F. Acernese, K. Ackley, C. Adams, R. X. Ad- hikari, V. B. Adya, C. Affeldt, et al. (LIGO Scientific Collaboration and Virgo Collaboration), Phys. Rev. X 9, 031040 (2019), URLhttps://link.aps.org/doi/10. 1103/PhysRevX.9.031040

    work page 2019

  3. [3]

    GWTC-2: Compact Binary Coalescences Observed by LIGO and Virgo During the First Half of the Third Observing Run

    R. Abbott, T. D. Abbott, S. Abraham, F. Acernese, K. Ackley, A. Adams, C. Adams, R. X. Adhikari, V. B. Adya, C. Affeldt, et al., Physical Review X11, 021053 (2021), 2010.14527

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  4. [4]

    Abbott, T

    R. Abbott, T. D. Abbott, F. Acernese, K. Ackley, C. Adams, N. Adhikari, R. X. Adhikari, V. B. Adya, C. Affeldt, D. Agarwal, et al. (LIGO Scientific Collab- oration, Virgo Collaboration, and KAGRA Collabora- tion), Phys. Rev. X13, 041039 (2023), URLhttps: //link.aps.org/doi/10.1103/PhysRevX.13.041039

    work page doi:10.1103/physrevx.13.041039 2023

  5. [5]

    GWTC-2.1: Deep Extended Catalog of Compact Binary Coalescences Observed by LIGO and Virgo During the First Half of the Third Observing Run

    R. Abbott, T. D. Abbott, F. Acernese, K. Ackley, C. Adams, N. Adhikari, R. X. Adhikari, V. B. Adya, C. Affeldt, D. Agarwal, et al., Phys. Rev. D109, 022001 (2024), 2108.01045

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  6. [6]

    The LIGO Scientific Collaboration, the Virgo Col- laboration, the KAGRA Collaboration, A. G. Abac, I. Abouelfettouh, F. Acernese, K. Ackley, C. Adam- cewicz, S. Adhicary, D. Adhikari, et al., arXiv e-prints arXiv:2508.18082 (2025), 2508.18082

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  7. [7]

    Abbott, T

    R. Abbott, T. D. Abbott, S. Abraham, F. Acernese, K. Ackley, A. Adams, C. Adams, R. X. Adhikari, V. B. Adya, C. Affeldt, et al., The Astrophysical Journal Let- ters915, L5 (2021)

    work page 2021

  8. [8]

    B. P. Abbott, R. Abbott, T. D. Abbott, F. Acernese, K. Ackley, C. Adams, T. Adams, P. Addesso, R. X. Adhikari, V. B. Adya, et al. (LIGO Scientific Col- laboration and Virgo Collaboration), Phys. Rev. Lett. 119, 161101 (2017), URLhttps://link.aps.org/doi/ 10.1103/PhysRevLett.119.161101

    work page doi:10.1103/physrevlett.119.161101 2017

  9. [9]

    B. P. Abbott, R. Abbott, T. D. Abbott, S. Abraham, F. Acernese, K. Ackley, C. Adams, R. X. Adhikari, V. B. Adya, C. Affeldt, et al., The Astrophysical Journal Let- ters892, L3 (2020)

    work page 2020

  10. [10]

    B. P. Abbott et al. (LIGO Scientific, Virgo), Class. Quant. Grav.37, 055002 (2020), 1908.11170

    work page arXiv 2020

  11. [11]

    Christensen and R

    N. Christensen and R. Meyer, Rev. Mod. Phys.94, 025001 (2022)

    work page 2022

  12. [12]

    J. Skilling, AIP Conference Proceedings735, 395 (2004), ISSN 0094-243X, https://pubs.aip.org/aip/acp/article- pdf/735/1/395/11702789/395 1 online.pdf, URLhttps: //doi.org/10.1063/1.1835238

    work page doi:10.1063/1.1835238 2004

  13. [13]

    J. S. Speagle, Monthly Notices of the RAS493, 3132 (2020), 1904.02180

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  14. [14]

    Metropolis, A

    N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller, The Journal of Chemical Physics21, 1087 (1953), ISSN 0021-9606, URLhttps: //doi.org/10.1063/1.1699114

    work page doi:10.1063/1.1699114 1953

  15. [15]

    W. K. Hastings, Biometrika57, 97 (1970), ISSN 0006-3444, https://academic.oup.com/biomet/article- pdf/57/1/97/23940249/57-1-97.pdf, URLhttps: //doi.org/10.1093/biomet/57.1.97

    work page doi:10.1093/biomet/57.1.97 1970

  16. [16]

    I. M. Romero-Shaw, C. Talbot, S. Biscoveanu, V. D’Emilio, G. Ashton, C. P. L. Berry, S. Cough- lin, S. Galaudage, C. Hoy, M. H¨ ubner, et al., Monthly 10 Notices of the Royal Astronomical Society499, 3295 (2020), ISSN 0035-8711, URLhttps://doi.org/10. 1093/mnras/staa2850

    work page 2020

  17. [17]

    Cosmic Explorer: The U.S. Contribution to Gravitational-Wave Astronomy beyond LIGO

    D. Reitze, R. X. Adhikari, S. Ballmer, B. Barish, L. Bar- sotti, G. Billingsley, D. A. Brown, Y. Chen, D. Coyne, R. Eisenstein, et al., inBulletin of the American Astro- nomical Society(2019), vol. 51, p. 35, 1907.04833

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  18. [18]

    Punturo, M

    M. Punturo, M. Abernathy, F. Acernese, B. Allen, N. An- dersson, K. Arun, F. Barone, B. Barr, M. Barsuglia, M. Beker, et al., Classical and Quantum Gravity27, 194002 (2010)

    work page 2010

  19. [19]

    A Horizon Study for Cosmic Explorer: Science, Observatories, and Community

    M. Evans, R. X. Adhikari, C. Afle, S. W. Ballmer, S. Bis- coveanu, S. Borhanian, D. A. Brown, Y. Chen, R. Eisen- stein, A. Gruson, et al., arXiv e-prints arXiv:2109.09882 (2021), 2109.09882

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  20. [20]

    Walker, R

    K. Walker, R. Smith, E. Thrane, and D. J. Reardon, Physical Review D110, 043013 (2024), 2401.02604

    work page arXiv 2024

  21. [21]

    Cuoco et al., Mach

    E. Cuoco et al., Mach. Learn. Sci. Tech.2, 011002 (2021), 2005.03745

    work page arXiv 2021

  22. [22]

    Benedetto, F

    V. Benedetto, F. Gissi, G. Ciaparrone, and L. Troiano, Applied Sciences13, 9886 (2023)

    work page 2023

  23. [23]

    Stergioulas, arXiv preprint (2024), arXiv:2401.07406

    N. Stergioulas, arXiv preprint (2024), arXiv:2401.07406

    work page arXiv 2024

  24. [24]

    L. Cai, L. Ren, Y. Wang, W. Xie, G. Zhu, and H. Gao, Royal Society Open Science8(2021)

    work page 2021

  25. [25]

    Nolan, A

    S. Nolan, A. Smerzi, and L. Pezz` e, npj Quantum In- formation7, 169 (2021), ISSN 2056-6387, URLhttps: //doi.org/10.1038/s41534-021-00497-w

    work page doi:10.1038/s41534-021-00497-w 2021

  26. [26]

    M. Khan, S. Raza Naqvi, Z. Ullah, S. Ali Ammar Taqvi, M. Nouman Aslam Khan, W. Farooq, M. Taqi Mehran, D. Juchelkov´ a, and L.ˇStˇ epanec, Fuel332, 126055 (2023), ISSN 0016-2361, URLhttps://www.sciencedirect. com/science/article/pii/S0016236122028794

    work page 2023

  27. [27]

    George and E

    D. George and E. Huerta, Physics Letters B 778, 64 (2018), ISSN 0370-2693, URLhttps: //www.sciencedirect.com/science/article/pii/ S0370269317310390

    work page 2018

  28. [28]

    Matching matched filtering with deep networks in gravitational-wave astronomy

    H. Gabbard, M. Williams, F. Hayes, and C. Messenger, Phys. Rev. Lett.120, 141103 (2018), 1712.06041

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  29. [29]

    A. J. K. Chua and M. Vallisneri, Phys. Rev. Lett.124, 041102 (2020), 1909.05966

    work page arXiv 2020

  30. [30]

    S. R. Green, C. Simpson, and J. Gair, Phys. Rev. D102, 104057 (2020), 2002.07656

    work page arXiv 2020

  31. [31]

    J. D. ´Alvares, J. A. Font, F. F. Freitas, O. G. Freitas, A. P. Morais, S. Nunes, A. Onofre, and A. Torres-Forn´ e, Classical and Quantum Gravity38, 155010 (2021), URL https://dx.doi.org/10.1088/1361-6382/ac0455

    work page doi:10.1088/1361-6382/ac0455 2021

  32. [32]

    P. G. Krastev, K. Gill, V. A. Villar, and E. Berger, Physics Letters B815, 136161 (2021), ISSN 0370- 2693, URLhttps://www.sciencedirect.com/science/ article/pii/S0370269321001015

    work page 2021

  33. [33]

    H. Shen, E. A. Huerta, E. O’Shea, P. Kumar, and Z. Zhao, Machine Learning: Science and Technology3, 015007 (2022), 1903.01998

    work page arXiv 2022

  34. [34]

    M. Dax, S. R. Green, J. Gair, M. P¨ urrer, J. Wildberger, J. H. Macke, A. Buonanno, and B. Sch¨ olkopf, Phys. Rev. Lett.130, 171403 (2023), 2210.05686

    work page arXiv 2023

  35. [35]

    Kolmus, J

    A. Kolmus, J. Janquart, T. Baka, T. van Laarhoven, C. V. D. Broeck, and T. Heskes,Tuning neural poste- rior estimation for gravitational wave inference(2024), 2403.02443

    work page arXiv 2024

  36. [36]

    Andr´ es-Carcasona, M

    M. Andr´ es-Carcasona, M. Mart´ ınez, and L. M. Mir, Monthly Notices of the Royal Astronomical Society527, 2887 (2023), ISSN 0035-8711, URLhttps://doi.org/ 10.1093/mnras/stad3448

    work page doi:10.1093/mnras/stad3448 2023

  37. [37]

    De Santi, M

    F. De Santi, M. Razzano, F. Fidecaro, L. Muccillo, L. Pa- palini, and B. Patricelli, Phys. Rev. D109, 102004 (2024), 2404.12028

    work page arXiv 2024

  38. [38]

    Gramaxo Freitas, J

    O. Gramaxo Freitas, J. Calder´ on Bustillo, J. A. Font, S. Nunes, A. Onofre, and A. Torres-Forn´ e, Machine Learning: Science and Technology5, 015036 (2024), 2307.16668

    work page arXiv 2024

  39. [39]

    Berbel, M

    M. Berbel, M. Miravet-Ten´ es, S. S. Chaudhary, S. Al- banesi, M. Cavagli` a, L. M. Zertuche, D. Tseneklidou, Y. Zheng, M. W. Coughlin, and A. Toivonen, Clas- sical and Quantum Gravity41, 085012 (2024), URL https://dx.doi.org/10.1088/1361-6382/ad3279

    work page doi:10.1088/1361-6382/ad3279 2024

  40. [40]

    M. Dax, S. R. Green, J. Gair, N. Gupte, M. P¨ urrer, V. Raymond, J. Wildberger, J. H. Macke, A. Buonanno, and B. Sch¨ olkopf, Nature (London)639, 49 (2025), 2407.09602

    work page arXiv 2025

  41. [41]

    Papamakarios, E

    G. Papamakarios, E. Nalisnick, D. J. Rezende, S. Mo- hamed, and B. Lakshminarayanan, Journal of Machine Learning Research22, 1 (2021), URLhttp://jmlr.org/ papers/v22/19-1028.html

    work page 2021

  42. [42]

    Papamakarios, T

    G. Papamakarios, T. Pavlakou, and I. Murray, in Advances in Neural Information Processing Systems, edited by I. Guyon, U. V. Luxburg, S. Bengio, H. Wal- lach, R. Fergus, S. Vishwanathan, and R. Garnett (Curran Associates, Inc., 2017), vol. 30, URLhttps: //proceedings.neurips.cc/paper_files/paper/2017/ file/6c1da886822c67822bcf3679d04369fa-Paper.pdf

    work page 2017

  43. [43]

    Durkan, A

    C. Durkan, A. Bekasov, I. Murray, and G. Papamakarios, arXiv e-prints arXiv:1906.04032 (2019), 1906.04032

    work page arXiv 1906

  44. [44]

    M. Dax, S. R. Green, J. Gair, J. H. Macke, A. Buo- nanno, and B. Sch¨ olkopf, Phys. Rev. Lett.127, 241103 (2021), URLhttps://link.aps.org/doi/10. 1103/PhysRevLett.127.241103

    work page 2021

  45. [45]

    LIGO Scientific Collaboration, J. Aasi, B. P. Abbott, R. Abbott, T. Abbott, M. R. Abernathy, K. Ackley, C. Adams, T. Adams, P. Addesso, et al., Classical and Quantum Gravity32, 074001 (2015)

    work page 2015

  46. [46]

    Acernese, M

    F. Acernese, M. Agathos, K. Agatsuma, D. Aisa, N. Alle- mandou, A. Allocca, J. Amarni, P. Astone, G. Balestri, G. Ballardin, et al., Classical and Quantum Gravity 32, 024001 (2015), URLhttps://dx.doi.org/10.1088/ 0264-9381/32/2/024001

    work page 2015

  47. [47]

    Computationally efficient models for the dominant and sub-dominant harmonic modes of precessing binary black holes

    G. Pratten, C. Garc´ ıa-Quir´ os, M. Colleoni, A. Ramos- Buades, H. Estell´ es, M. Mateu-Lucena, R. Jaume, M. Haney, D. Keitel, J. E. Thompson, et al., Phys. Rev. D103, 104056 (2021), 2004.06503

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  48. [48]

    Bilby: A user-friendly Bayesian inference library for gravitational-wave astronomy

    G. Ashton, M. H¨ ubner, P. D. Lasky, C. Talbot, K. Ack- ley, S. Biscoveanu, Q. Chu, A. Divakarla, P. J. Easter, B. Goncharov, et al., Astrophysical Journal, Supplement 241, 27 (2019), 1811.02042

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  49. [49]

    Klimenko, G

    S. Klimenko, G. Vedovato, M. Drago, F. Salemi, V. Ti- wari, G. A. Prodi, C. Lazzaro, K. Ackley, S. Ti- wari, C. F. Da Silva, et al., Phys. Rev. D93, 042004 (2016), URLhttps://link.aps.org/doi/10. 1103/PhysRevD.93.042004

    work page 2016

  50. [50]

    Drago, S

    M. Drago, S. Klimenko, C. Lazzaro, E. Milotti, G. Mitsel- makher, V. Necula, B. O’Brian, G. A. Prodi, F. Salemi, M. Szczepanczyk, et al., SoftwareX14, 100678 (2021), ISSN 2352-7110, URLhttps://www.sciencedirect. com/science/article/pii/S2352711021000236

    work page 2021

  51. [51]

    Meireles, P

    M. Meireles, P. Almeida, and M. Simoes, IEEE Transac- tions on Industrial Electronics50, 585 (2003)

    work page 2003

  52. [52]

    A. A. M. Al-Saffar, H. Tao, and M. A. Talab, 11 in2017 International Conference on Radar, An- tenna, Microwave, Electronics, and Telecommunications (ICRAMET)(2017), pp. 26–31

    work page 2017

  53. [53]

    X. Yuan, C. Ma, M. Hu, R. L. J. Qiu, E. Salari, R. Martini, and X. Yang, Journal of Applied Clinical Medical Physics26, e14559 (2025), URL https://aapm.onlinelibrary.wiley.com/doi/abs/10. 1002/acm2.14559

    work page 2025

  54. [54]

    Rezende, G

    E. Rezende, G. Ruppert, T. Carvalho, F. Ramos, and P. de Geus, in2017 16th IEEE International Conference on Machine Learning and Applications (ICMLA)(2017), pp. 1011–1014

    work page 2017

  55. [55]

    L. Gao, X. Zhang, T. Yang, B. Wang, and J. Li, Elec- tronics12(2023), ISSN 2079-9292, URLhttps://www. mdpi.com/2079-9292/12/17/3677

    work page 2023

  56. [56]

    K. He, X. Zhang, S. Ren, and J. Sun,Deep residual learn- ing for image recognition(2015), 1512.03385

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  57. [57]

    D. M. Macleod, J. S. Areeda, S. B. Cough- lin, T. J. Massinger, and A. L. Urban, Soft- wareX13, 100657 (2021), ISSN 2352-7110, URL https://www.sciencedirect.com/science/article/ pii/S2352711021000029

    work page 2021

  58. [58]

    S. Walt, J. Sch¨ onberger, J. Nunez-Iglesias, F. Boulogne, J. Warner, N. Yager, E. Gouillart, and T. Yu,scikit- image: Image processing in python(2014)

    work page 2014

  59. [59]

    J. C. Brown, The Journal of the Acousti- cal Society of America89, 425 (1991), ISSN 0001-4966, https://pubs.aip.org/asa/jasa/article- pdf/89/1/425/12135575/425 1 online.pdf, URL https://doi.org/10.1121/1.400476

    work page doi:10.1121/1.400476 1991

  60. [60]

    S. R. Green and J. Gair, Machine Learning: Science and Technology2, 03LT01 (2021), URLhttps://dx.doi. org/10.1088/2632-2153/abfaed

    work page doi:10.1088/2632-2153/abfaed 2021

  61. [61]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, et al., inProceedings of the 33rd International Confer- ence on Neural Information Processing Systems(Curran Associates Inc., Red Hook, NY, USA, 2019)

    work page 2019

  62. [63]

    D. P. Kingma and J. Ba,Adam: A method for stochas- tic optimization(2017), revised preprint of 2014 paper, 1412.6980, URLhttps://arxiv.org/abs/1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  63. [64]

    SGDR: Stochastic Gradient Descent with Warm Restarts

    I. Loshchilov and F. Hutter,Sgdr: Stochastic gradi- ent descent with warm restarts(2017), 1608.03983, URL https://arxiv.org/abs/1608.03983

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  64. [65]

    Lin, IEEE Transactions on Information Theory37, 145 (1991)

    J. Lin, IEEE Transactions on Information Theory37, 145 (1991)

    work page 1991

  65. [66]

    A. H. Nitz, S. Kumar, Y.-F. Wang, S. Kastha, S. Wu, M. Sch¨ afer, R. Dhurkunde, and C. D. Capano, The As- trophysical Journal946, 59 (2023), URLhttps://doi. org/10.3847/1538-4357/aca591

    work page doi:10.3847/1538-4357/aca591 2023

  66. [67]

    M. d. L. Planas, A. Ramos-Buades, C. Garc´ ıa-Quir´ os, H. Estell´ es, S. Husa, and M. Haney, Phys. Rev. D112, 123004 (2025), 2504.15833

    work page arXiv 2025

  67. [68]

    R. J. E. Smith, G. Ashton, A. Vajpeyi, and C. Talbot, Monthly Notices of the Royal Astronomical Society498, 4492 (2020), ISSN 0035-8711, URLhttps://doi.org/ 10.1093/mnras/staa2483

    work page doi:10.1093/mnras/staa2483 2020

  68. [69]

    Meyer, M

    R. Meyer, M. C. Edwards, P. Maturana-Russel, and N. Christensen, WIREs Computational Statistics14, e1532 (2022), URLhttps://wires.onlinelibrary. wiley.com/doi/abs/10.1002/wics.1532

    work page doi:10.1002/wics.1532 2022

  69. [70]

    Open data from the third observing run of LIGO, Virgo, KAGRA and GEO

    R. Abbott et al. (KAGRA, VIRGO, LIGO Scientific), Astrophys. J. Suppl.267, 29 (2023), 2302.03676

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  70. [71]

    Planck Collaboration, P. A. R. Ade, N. Aghanim, M. Ar- naud, M. Ashdown, J. Aumont, C. Baccigalupi, A. J. Banday, R. B. Barreiro, J. G. Bartlett, et al., Astronomy & Astrophysics594, A13 (2016), 1502.01589

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  71. [72]

    M. Dax, S. R. Green, J. Gair, M. Deistler, B. Sch¨ olkopf, and J. H. Macke (2021), 2111.13139

    work page arXiv 2021

  72. [73]

    Sun, C.-Y

    T.-Y. Sun, C.-Y. Xiong, S.-J. Jin, Y.-X. Wang, J.-F. Zhang, and X. Zhang, Chinese Physics C48, 045108 (2024), URLhttps://dx.doi.org/10.1088/1674-1137/ ad2a5f

    work page doi:10.1088/1674-1137/ 2024

  73. [74]

    A. H. Nitz, I. Harry, D. Brown, C. M. Biwer, J. Willis, T. D. Canton, C. Capano, T. Dent, L. Pekowsky, S. De, et al.,gwastro/pycbc: v2.0.6 release of pycbc(2023), soft- ware package, URLhttps://doi.org/10.5281/zenodo. 7547919

    work page doi:10.5281/zenodo 2023

  74. [75]

    Hoy and V

    C. Hoy and V. Raymond, SoftwareX15, 100765 (2021), ISSN 2352-7110, URLhttps://www.sciencedirect. com/science/article/pii/S2352711021000856

    work page 2021

  75. [76]

    J. D. Hunter, Computing in Science & Engineering9, 90 (2007)

    work page 2007