pith. sign in

arxiv: 2605.17377 · v1 · pith:GMXURTN5new · submitted 2026-05-17 · 🌌 astro-ph.HE

Precise and Rapid Parameter Inference of Kilonova with Conditional Variational Autoencoder

Surojit Saha , Albert K.H Kong This is my paper

Pith reviewed 2026-05-19 23:01 UTC · model grok-4.3

classification 🌌 astro-ph.HE
keywords kilonovaparameter inferenceconditional variational autoencoderlight curvesneutron star mergersvariational inferencegravitational wavesmulti-messenger astronomy
0
0 comments X

The pith

A conditional variational autoencoder enables rapid parameter inference for kilonovae from light curves.

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

The paper proposes training a conditional variational autoencoder on publicly available kilonova light curves paired with their physical parameters. This setup lets the model approximate the likelihood through variational inference, allowing quick estimates of parameters for new observations. If correct, the method would reduce the time for such analyses from days or longer down to a total of under three hours from start to finish. The result would support faster follow-up studies of neutron star merger events detected through gravitational waves.

Core claim

By conditioning a variational autoencoder on kilonova light curve data and the associated physical parameters during training, the authors obtain a model that performs rapid parameter inference on new light curves, with the full process of training plus inference completing in under approximately three hours.

What carries the argument

The conditional variational autoencoder that learns to map light curve observations to parameter distributions by approximating the posterior via variational inference.

Load-bearing premise

The training light curves must adequately represent the full range of physical parameters in the chosen kilonova model, and the variational approximation must be accurate enough for reliable inference.

What would settle it

Running the trained CVAE on a new kilonova light curve and comparing the inferred parameter values and uncertainties directly against results from conventional Bayesian sampling on the same data; consistent mismatches beyond statistical expectations would falsify the precision claim.

Figures

Figures reproduced from arXiv: 2605.17377 by Albert K.H Kong, Surojit Saha.

Figure 1
Figure 1. Figure 1: In this plot, corresponding to the data set DA, the histogram depicts the distribution of the physical parameters that has been used for the training, testing and validation. As is evident, in the sets of physical parameters, we have only specific values. The light curve shown in the g band, is a sample light curve and has a single physical parameter set with md=0.001M⊙, vd=0.05c, mw=0.001M⊙ and vw=0.05c i… view at source ↗
Figure 2
Figure 2. Figure 2: The plot outlines the kernel density estimation between the true and CVAE-generated physical parameters for TP-wind 1,TP-wind 2,TS-wind 1 and TS-wind 2 models. Primarily, g-band light curves from the test data have been used to generate these physical parameters. The true and CVAE-generated distribution across the models for each of the physical parameters overlaps with each other hence representing a robu… view at source ↗
Figure 3
Figure 3. Figure 3: This polar plot represents the kernel density approximation of the true and CVAE-generated physical parameters calculated in each filter band across the whole parameter space for md, vd, mw and vw respectively for TP-wind 1 model. The different filter bands are shown along the angular axis. The violin plots in the inset corresponding to each polar plot compare the distribution of the true and CVAE-generate… view at source ↗
Figure 4
Figure 4. Figure 4: Mean squared error plot for all physical parameters grouped according to the models for all the filter bands. In the consecutive frames shows the MSE for md, vd, mw and vw where in each frame, MSEs are grouped according to the KNe models across the filter bands. the training data. Utilizing a denser angular grid is expected to improve interpolation between adjacent viewing angles and should consequently re… view at source ↗
Figure 5
Figure 5. Figure 5: This plot represents the mean squared error calculated between the true and CVAE-generated physical parameters for chirp mass, mass ratio, fraction of the remnant disk and viewing angle while grouped according to the filter bands [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Kernel density estimation plot for all physical parameters grouped according to the filter bands. This comprehensive plot demonstrates the extent of intersection between the true and generated physical parameters. In this subplot, down the rows, for each filter bands and across the rows for physical parameters, the KDEs for chirp mass, mass ratio, fraction of the remnant disk and viewing angle are depicted… view at source ↗
Figure 7
Figure 7. Figure 7: Polar plot combined with KDE plot for all physical parameters across all filter bands for TP-wind 2 model [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Polar plot combined with KDE plot for all physical parameters across all filter bands for TS-wind 1 model [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Polar plot combined with KDE plot for all physical parameters across all filter bands for TS-wind 2 model [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
read the original abstract

The coalescence of binary neutron stars in the GW170817 event led to the generation of gravitational waves, accompanied by the electromagnetic counterpart known as a kilonova (KN). Since then, it has been a prime topic of interest, as it has provided much insight into multi-messenger astronomy. Apart from existing methods for parameter estimation, we propose an alternative technique for it, utilizing the strength and flexibility of a conditional variational autoencoder. Publicly available light curves are used as training data, conditioning on the corresponding physical parameters for a chosen model; after training, we carry out rapid parameter inferences. As this approach approximates the likelihood through variational inference, it yields results more efficiently. Through this innovative approach, we demonstrated that the total time, from training to parameter inference, is under $\approx3$h. We showed that for a given KN light curve, we can rapidly perform parameter inference based on the required model.

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 proposes using a conditional variational autoencoder (CVAE) trained on publicly available kilonova light curves, conditioned on the corresponding physical parameters of a chosen model, to enable rapid parameter inference. The central claim is that this approach approximates the likelihood via variational inference and completes the full process from training to inference in under approximately 3 hours, offering an efficient alternative to existing methods for kilonova parameter estimation.

Significance. If the reported precision and speed are validated with quantitative benchmarks, the method could meaningfully accelerate multi-messenger analyses of future binary neutron star mergers by providing near-real-time posterior estimates. The approach builds on standard supervised training of a CVAE and does not appear to introduce new physical modeling, so its primary value would lie in computational efficiency rather than novel scientific insight.

major comments (3)
  1. [Abstract / Results] Abstract and results summary: The claim of 'precise' inference is not accompanied by any quantitative metrics (e.g., parameter recovery errors, posterior coverage fractions, or direct comparisons to MCMC or nested sampling), validation plots, or error bars. Without these in the results section, it is impossible to assess whether the central claim of accurate rapid inference holds.
  2. [Training data / §2] Training data section (likely §2): The manuscript relies on publicly available light curves as training data but does not demonstrate that these curves densely sample the full physical parameter space (ejecta mass, velocity, composition, viewing angle) of the chosen kilonova model. Sparse or biased coverage would cause the variational posterior to be unreliable outside the sampled region, directly undermining the asserted precision.
  3. [Timing / §4] Timing claim (likely §4): The statement that the total time from training to inference is under ≈3 h requires explicit specification of the hardware, batch sizes, and breakdown between training and inference phases. This information is load-bearing for the 'rapid' aspect of the central claim.
minor comments (2)
  1. [Methodology] The notation used for the CVAE evidence lower bound (ELBO) and conditioning variables should be introduced with an explicit equation to improve clarity.
  2. [Figures] Figure captions for any light-curve or posterior plots should include the specific kilonova model parameters and the number of samples used.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have helped us improve the clarity and rigor of our manuscript. We address each major comment point by point below and have revised the manuscript to incorporate the requested information and validations.

read point-by-point responses

  1. Referee: [Abstract / Results] Abstract and results summary: The claim of 'precise' inference is not accompanied by any quantitative metrics (e.g., parameter recovery errors, posterior coverage fractions, or direct comparisons to MCMC or nested sampling), validation plots, or error bars. Without these in the results section, it is impossible to assess whether the central claim of accurate rapid inference holds.

    Authors: We agree that quantitative metrics are necessary to substantiate the claim of precise inference. The original manuscript focused primarily on demonstrating the speed of the CVAE approach but did not include explicit error metrics or comparisons in the results. In the revised version, we have added a dedicated subsection to the results with quantitative benchmarks, including mean absolute percentage errors for key parameters (ejecta mass, velocity, and viewing angle), 68% and 95% posterior coverage fractions on a held-out test set, and side-by-side comparisons of CVAE posteriors versus MCMC sampling for representative light curves. Corresponding validation plots have also been included. revision: yes

  2. Referee: [Training data / §2] Training data section (likely §2): The manuscript relies on publicly available light curves as training data but does not demonstrate that these curves densely sample the full physical parameter space (ejecta mass, velocity, composition, viewing angle) of the chosen kilonova model. Sparse or biased coverage would cause the variational posterior to be unreliable outside the sampled region, directly undermining the asserted precision.

    Authors: The referee raises a valid concern regarding the representativeness of the training data. While the manuscript states that publicly available light curves conditioned on the model parameters were used, it did not explicitly verify the density of coverage across the full parameter space. We have revised §2 to include a new figure showing the marginal and joint distributions of the training parameters (ejecta mass, velocity, composition, and viewing angle) and added text confirming that the sampled points provide dense coverage within the physically motivated ranges of the kilonova model, with no large gaps that would affect interpolation. revision: yes

  3. Referee: [Timing / §4] Timing claim (likely §4): The statement that the total time from training to inference is under ≈3 h requires explicit specification of the hardware, batch sizes, and breakdown between training and inference phases. This information is load-bearing for the 'rapid' aspect of the central claim.

    Authors: We thank the referee for highlighting the need for implementation details to support the timing claim. The original manuscript reported the total time of under ≈3 h but omitted the supporting specifications. In the revised §4, we now explicitly state that all computations were performed on an NVIDIA A100 GPU with 40 GB memory, using a batch size of 256 for training. We provide a breakdown: approximately 2 hours and 40 minutes for model training (including data loading and 100 epochs) and under 20 minutes for inference on a new light curve, including posterior sampling. revision: yes

Circularity Check

0 steps flagged

No circularity: standard CVAE training on external catalogs yields independent inference

full rationale

The paper trains a conditional variational autoencoder on publicly available kilonova light-curve catalogs paired with physical parameters, then uses the trained model for rapid posterior inference on new inputs via variational approximation. This pipeline does not reduce any claimed prediction or result to a quantity defined by the same inputs or by self-citation; the inference step is a forward pass through a model whose parameters were optimized on separate training data. No load-bearing uniqueness theorems, ansatzes smuggled via prior work, or fitted quantities renamed as predictions appear. The approach is self-contained against external benchmarks and follows ordinary supervised ML practice.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach rests on standard variational-inference assumptions and the representativeness of public training light curves; no new physical entities or ad-hoc constants are introduced in the abstract.

axioms (1)
  • domain assumption Variational inference provides a tractable approximation to the posterior over physical parameters given observed light curves
    Core justification for using CVAE instead of exact likelihood sampling; invoked in the description of the method.

pith-pipeline@v0.9.0 · 5686 in / 1294 out tokens · 41178 ms · 2026-05-19T23:01:23.303789+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

48 extracted references · 48 canonical work pages · 13 internal anchors

  1. [1]

    Wollaeger , R. T. and Fryer , C. L. and Chase , E. A. and Fontes , C. J. and Ristic , M. and Hungerford , A. L. and Korobkin , O. and O'Shaughnessy , R. and Herring , A. M. A Broad Grid of 2D Kilonova Emission Models. , keywords =. 2021. doi:10.3847/1538-4357/ac0d03. arXiv:2105.11543

    work page doi:10.3847/1538-4357/ac0d03 2021

  2. [2]

    Mass ejection in neutron star mergers

    Rosswog , S. and Liebend \"o rfer , M. and Thielemann , F. -K. and Davies , M. B. and Benz , W. and Piran , T. Mass ejection in neutron star mergers. , keywords =. 1999. arXiv:astro-ph/9811367

    work page internal anchor Pith review Pith/arXiv arXiv 1999

  3. [3]

    The Electromagnetic Counterpart of the Binary Neutron Star Merger LIGO/Virgo GW170817. I. Dark Energy Camera Discovery of the Optical Counterpart

    Soares-Santos , M. and Holz , D. E. and Annis , J. and Chornock , R. and Herner , K. and Berger , E. and Brout , D. and Chen , H. -Y. and Kessler , R. and Sako , M. and Allam , S. and Tucker , D. L. and Butler , R. E. and Palmese , A. and Doctor , Z. and Diehl , H. T. and Frieman , J. and Yanny , B. and Lin , H. and Scolnic , D. and Cowperthwaite , P. and...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/2041-8213/aa9059 2017

  4. [4]

    1998, ApJL, 507, L59, doi: 10.1086/311680

    Li-Xin Li and Bohdan Paczy \' n ski. Transient Events from Neutron Star Mergers. 1998. doi:10.1086/311680

    work page doi:10.1086/311680 1998

  5. [5]

    N., Kokubo E., Inutsuka S.-I., Matsumoto T., 2010, @doi [ ] 10.1111/j.1365-2966.2010.16527.x , http://adsabs.harvard.edu/abs/2010MNRAS.405.1227M 405, 1227

    Electromagnetic counterparts of compact object mergers powered by the radioactive decay of r-process nuclei. , keywords =. doi:10.1111/j.1365-2966.2010.16864.x , archivePrefix =. 1001.5029 , primaryClass =

    work page doi:10.1111/j.1365-2966.2010.16864.x 2010

  6. [6]

    Kingma and Max Welling

    Kingma, Diederik P. and Welling, Max , year=. An Introduction to Variational Autoencoders , volume=. Foundations and Trends® in Machine Learning , publisher=. doi:10.1561/2200000056 , number=

    work page doi:10.1561/2200000056

  7. [7]

    arXiv e-prints , keywords =

    Auto-Encoding Variational Bayes. arXiv e-prints , keywords =

    work page

  8. [8]

    V. M. Lipunov and E. Gorbovskoy and V. G. Kornilov and N . Tyurina and P. Balanutsa and A. Kuznetsov and D. Vlasenko and D. Kuvshinov and I. Gorbunov and D. A. H. Buckley and A. V. Krylov and R. Podesta and C. Lopez and F. Podesta and H. Levato and C. Saffe and C. Mallamachi and S. Potter and N. M. Budnev and O. Gress and Yu. Ishmuhametova and V. Vladimir...

    work page doi:10.3847/2041-8213/aa92c0 2041

  9. [9]

    N. R. Tanvir and A. J. Levan and C. Gonz. The Emergence of a Lanthanide-rich Kilonova Following the Merger of Two Neutron Stars , journal =. doi:10.3847/2041-8213/aa90b6 , url =

    work page doi:10.3847/2041-8213/aa90b6 2041

  10. [10]

    Iair Arcavi and Curtis McCully and Griffin Hosseinzadeh and D. Andrew Howell and Sergiy Vasylyev and Dovi Poznanski and Michael Zaltzman and Dan Maoz and Leo Singer and Stefano Valenti and Daniel Kasen and Jennifer Barnes and Tsvi Piran and Wen-fai Fong , title =. ApJ , abstract =. doi:10.3847/2041-8213/aa910f , url =

    work page doi:10.3847/2041-8213/aa910f 2041

  11. [11]

    The discovery of the electromagnetic counterpart of GW170817: kilonova AT 2017gfo/DLT17ck

    The Discovery of the Electromagnetic Counterpart of GW170817: Kilonova AT 2017gfo/DLT17ck. , keywords =. doi:10.3847/2041-8213/aa8edf , archivePrefix =. 1710.05854 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/2041-8213/aa8edf 2041

  12. [12]

    D. A. Coulter and R. J. Foley and C. D. Kilpatrick and M. R. Drout and A. L. Piro and B. J. Shappee and M. R. Siebert and J. D. Simon and N. Ulloa and D. Kasen and B. F. Madore and A. Murguia-Berthier and Y.-C. Pan and J. X. Prochaska and E. Ramirez-Ruiz and A. Rest and C. Rojas-Bravo , title =. Science , volume =. 2017 , doi =. https://www.science.org/do...

    work page doi:10.1126/science.aap9811 2017

  13. [13]

    A Short History of Markov Chain Monte Carlo: Subjective Recollections from Incomplete Data , volume =

    Christian Robert and George Casella , journal =. A Short History of Markov Chain Monte Carlo: Subjective Recollections from Incomplete Data , volume =

    work page

  14. [14]

    Markov Chain Monte Carlo Methods for Bayesian Data Analysis in Astronomy

    Sharma, Sanjib. Markov Chain Monte Carlo Methods for Bayesian Data Analysis in Astronomy. 2017. doi:https://doi.org/10.1146/annurev-astro-082214-122339

    work page doi:10.1146/annurev-astro-082214-122339 2017

  15. [15]

    Learning Structured Output Representation using Deep Conditional Generative Models , url =

    Sohn, Kihyuk and Lee, Honglak and Yan, Xinchen , booktitle =. Learning Structured Output Representation using Deep Conditional Generative Models , url =

    work page

  16. [16]

    2021 , eprint=

    Tutorial on Variational Autoencoders , author=. 2021 , eprint=

    work page 2021

  17. [17]

    Bond-Taylor, Sam and Leach, Adam and Long, Yang and Willcocks, Chris G. , year=. Deep Generative Modelling: A Comparative Review of VAEs, GANs, Normalizing Flows, Energy-Based and Autoregressive Models , volume=. IEEE Transactions on Pattern Analysis and Machine Intelligence , publisher=. doi:10.1109/tpami.2021.3116668 , number=

    work page doi:10.1109/tpami.2021.3116668 2021

  18. [18]

    Nature Phys.18(1), 112–117 (2022) https://doi.org/10.1038/s41567-021-01425-7 arXiv:1909.06296 [astro-ph.IM]

    Gabbard, Hunter and Messenger, Chris and Heng, Ik Siong and Tonolini, Francesco and Murray-Smith, Roderick , year=. Bayesian parameter estimation using conditional variational autoencoders for gravitational-wave astronomy , volume=. Nature Physics , publisher=. doi:10.1038/s41567-021-01425-7 , number=

    work page doi:10.1038/s41567-021-01425-7

  19. [19]

    arXiv e-prints , keywords =

    Practical Guidance for Bayesian Inference in Astronomy. arXiv e-prints , keywords =. doi:10.48550/arXiv.2302.04703 , archivePrefix =. 2302.04703 , primaryClass =

    work page doi:10.48550/arxiv.2302.04703

  20. [20]

    Thrane and C

    Thrane, Eric and Talbot, Colm , year=. An introduction to Bayesian inference in gravitational-wave astronomy: Parameter estimation, model selection, and hierarchical models , volume=. doi:10.1017/pasa.2019.2 , journal=

    work page doi:10.1017/pasa.2019.2 2019

  21. [21]

    Bayesian Astrostatistics: A Backward Look to the Future

    Loredo, Thomas J. Bayesian Astrostatistics: A Backward Look to the Future. Astrostatistical Challenges for the New Astronomy. 2013. doi:10.1007/978-1-4614-3508-2_2

    work page doi:10.1007/978-1-4614-3508-2_2 2013

  22. [22]

    Learning from Examples in Astronomy and Physics

    Bayesian Methods for the Physical Sciences. Learning from Examples in Astronomy and Physics

    work page

  23. [23]

    2014 , publisher=

    Statistical methods for astronomical data analysis , author=. 2014 , publisher=

    work page 2014

  24. [24]

    Markov Chain Monte Carlo Methods for Bayesian Data Analysis in Astronomy , volume =

    Sharma, Sanjib , year =. Markov Chain Monte Carlo Methods for Bayesian Data Analysis in Astronomy , volume =. , doi =

    work page

  25. [25]

    Loredo, T. J. From Laplace to Supernova SN 1987A: Bayesian Inference in Astrophysics. Maximum Entropy and Bayesian Methods. 1990. doi:10.1007/978-94-009-0683-9_6

    work page doi:10.1007/978-94-009-0683-9_6 1990

  26. [26]

    , keywords =

    Parameter estimation in astronomy through application of the likelihood ratio. , keywords =

    work page

  27. [27]

    10.1051/0004-6361:20034401

    On Maximum Likelihood Estimation of averaged power spectra , DOI= "10.1051/0004-6361:20034401", url= "https://doi.org/10.1051/0004-6361:20034401", journal =

    work page doi:10.1051/0004-6361:20034401

  28. [28]

    , keywords =

    Parameter estimation in X-ray astronomy using maximum likelihood. , keywords =

    work page

  29. [29]

    and Fryer, Christopher L

    Korobkin, Oleg and Wollaeger, Ryan T. and Fryer, Christopher L. and Hungerford, Aimee L. and Rosswog, Stephan and Fontes, Christopher J. and Mumpower, Matthew R. and Chase, Eve A. and Even, Wesley P. and Miller, Jonah and Misch, G. Wendell and Lippuner, Jonas , title =. 2021 , month =. doi:10.3847/1538-4357/abe1b5 , url =

    work page doi:10.3847/1538-4357/abe1b5 2021

  30. [30]

    Full Transport Model of GW170817-Like Disk Produces a Blue Kilonova

    Full transport model of GW170817-like disk produces a blue kilonova. , keywords =. doi:10.1103/PhysRevD.100.023008 , archivePrefix =. 1905.07477 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physrevd.100.023008 1905

  31. [31]

    The Astropy Project: Sustaining and Growing a Community-oriented Open-source Project and the Latest Major Release (v5.0) of the Core Package

    The Astropy Project: Sustaining and Growing a Community-oriented Open-source Project and the Latest Major Release (v5.0) of the Core Package. , keywords =. doi:10.3847/1538-4357/ac7c74 , archivePrefix =. 2206.14220 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4357/ac7c74

  32. [32]

    The Astropy Project: Building an inclusive, open-science project and status of the v2.0 core package

    The Astropy Project: Building an Open-science Project and Status of the v2.0 Core Package. , keywords =. doi:10.3847/1538-3881/aabc4f , archivePrefix =. 1801.02634 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-3881/aabc4f

  33. [33]

    Astropy: A Community Python Package for Astronomy

    Astropy: A community Python package for astronomy. , keywords =. 2013. doi:10.1051/0004-6361/201322068 , archivePrefix =. 1307.6212 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201322068 2013

  34. [34]

    , keywords =

    SExtractor: Software for source extraction. , keywords =. 1996

    work page 1996

  35. [35]

    Quantifying the Observational Effort Required for the Radial Velocity Characterization of TESS Planets

    Quantifying the Observational Effort Required for the Radial Velocity Characterization of TESS Planets. , keywords =. 2018. doi:10.3847/1538-3881/aacea9 , archivePrefix =. 1807.01263 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-3881/aacea9 2018

  36. [36]

    X-ray Scattering Echoes and Ghost Halos from the Intergalactic Medium: Relation to the nature of AGN variability

    X-Ray Scattering Echoes and Ghost Halos from the Intergalactic Medium: Relation to the Nature of AGN Variability. , keywords =. 2015. doi:10.1088/0004-637X/805/1/23 , archivePrefix =. 1503.01475 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/805/1/23 2015

  37. [37]

    , keywords =

    The 2013 Release of Cloudy. , keywords =. 2013

    work page 2013

  38. [38]

    1989", month =

    T _ E X and LAT _ E X Macro Definition Files for Astronomical Publications. , year = "1989", month = "Mar", pages =

    work page 1989

  39. [39]

    LaTeX: A Document Preparation System. 1994

    work page 1994

  40. [40]

    Quasi-periodic Fast Propagating Magnetoacoustic Waves during the Magnetic Reconnection Between Solar Coronal Loops

    Quasi-periodic Fast Propagating Magnetoacoustic Waves during the Magnetic Reconnection Between Solar Coronal Loops. , keywords =. 2018. doi:10.3847/2041-8213/aaf167 , archivePrefix =. 1811.08553 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/2041-8213/aaf167 2018

  41. [41]

    Nominal values for selected solar and planetary quantities: IAU 2015 Resolution B3

    Nominal Values for Selected Solar and Planetary Quantities: IAU 2015 Resolution B3. , keywords =. 2016. doi:10.3847/0004-6256/152/2/41 , archivePrefix =. 1605.09788 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/0004-6256/152/2/41 2015

  42. [42]

    Swift X-Ray Observations of Classical Novae. II. The Super Soft Source Sample. , keywords =. 2011. doi:10.1088/0067-0049/197/2/31 , archivePrefix =. 1110.6224 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0067-0049/197/2/31 2011

  43. [43]

    Galaxy emission line classification using 3D line ratio diagrams

    Galaxy Emission Line Classification Using Three-dimensional Line Ratio Diagrams. , keywords =. 2014. doi:10.1088/0004-637X/793/2/127 , archivePrefix =. 1406.5186 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/793/2/127 2014

  44. [44]

    , keywords =

    Tight multimessenger constraints on the neutron star equation of state from GW170817 and a forward model for kilonova light-curve synthesis. , keywords =. doi:10.1093/mnras/stab1523 , archivePrefix =. 2102.02229 , primaryClass =

    work page doi:10.1093/mnras/stab1523

  45. [45]

    2015 , howpublished=

    Keras , author=. 2015 , howpublished=

    work page 2015

  46. [46]

    Hunter, J. D. , Title =. Computing in Science & Engineering , Volume =

    work page

  47. [47]

    and Varoquaux, G

    Pedregosa, F. and Varoquaux, G. and Gramfort, A. and Michel, V. and Thirion, B. and Grisel, O. and Blondel, M. and Prettenhofer, P. and Weiss, R. and Dubourg, V. and Vanderplas, J. and Passos, A. and Cournapeau, D. and Brucher, M. and Perrot, M. and Duchesnay, E. , journal=. Scikit-learn: Machine Learning in

    work page

  48. [48]

    doi:10.21105/joss.03021

    Michael L. Waskom , title =. 2021 , publisher =. doi:10.21105/joss.03021 , url =

    work page doi:10.21105/joss.03021 2021