Pith. sign in

REVIEW 3 major objections 6 minor 58 references

A recurrent neural network on photometric light curves can build a Type Ia supernova sample pure enough for dark-energy constraints.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-04 00:05 UTC pith:ES2N6JNF

load-bearing objection A competent and transparent mock forecast for CSST-UDF SN cosmology; the headline numbers are self-consistent rather than robust, but the paper deserves peer review. the 3 major comments →

arxiv 2511.02631 v2 pith:ES2N6JNF submitted 2025-11-04 astro-ph.CO

Supernova Classification using the Recurrent Neural Network in the CSST Ultra-Deep Field Survey

Minglin Wang , Yan Gong , Dejia Zhou , Xuelei Chen This is my paper
classification astro-ph.CO
keywords supernova classificationrecurrent neural networksLSTMCSST Ultra-Deep FieldType Ia supernovaedark energyphotometric surveyscosmological constraints
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper argues that a recurrent neural network can sort supernova light curves from the CSST Ultra-Deep Field survey well enough that photometric data alone—no spectroscopy—yields a cosmology-grade Type Ia sample. The classifier, combined with quality cuts on the fitted light-curve parameters, pushes sample purity above 99.5%: 2,193 Type Ia supernovae with only 4 core-collapse contaminants. After correcting selection-effect magnitude biases with a two-step binned method, the authors get 1σ constraints of 14% on matter density Ω_M and 18% on the dark-energy equation-of-state parameter w, assuming a flat wCDM model. These precisions are comparable to surveys that rely on spectroscopic confirmation. A sympathetic reader would care because this suggests next-generation photometric surveys can probe cosmic expansion without the cost of spectroscopy.

Core claim

The central discovery is a numerical result from simulated observations: a recurrent-neural-network classifier, trained on roughly 250,000 mock Type Ia and 250,000 mock core-collapse supernovae, classifies 9,445 simulated CSST-UDF light curves with 99.1% overall accuracy. After a template fit and quality cuts on chi-square, stretch, color, and time-of-peak error, the sample holds 2,193 Type Ia supernovae and 4 core-collapse contaminants — purity above 99.5%. With selection-effect bias corrected (up to ~0.036 mag at high redshift), an MCMC fit gives Ω_M = 0.304 (+0.030/−0.052) and w = −1.017 (+0.177/−0.189), i.e. 14% and 18% relative accuracy under flat wCDM. The paper presents this as eviden

What carries the argument

The load-bearing mechanism is a two-layer bidirectional LSTM recurrent neural network that reads each supernova's multi-band light curve as a time series and outputs a Type Ia probability. The second mechanism is a set of quality cuts on the template-fit parameters — reduced chi-square below 5, stretch and color inside model-valid ranges, and time-of-peak error below 2 days — that remove the remaining core-collapse contaminants. The third is a two-step binned bias-correction likelihood: first, distance-modulus offsets are fitted per redshift bin while nuisance parameters are marginalized; second, Ω_M and w are fitted from those offsets. The bias correction is needed because selection effects

Load-bearing premise

The result depends on the simulated light curves being a faithful stand-in for real CSST-UDF observations—same templates, same noise, same selection—so that the classifier's 99.5% purity and the resulting 14% and 18% parameter accuracies carry over to the real sky.

What would settle it

Take the same LSTM classifier and test it on a mock sample that includes host-galaxy extinction, realistic photometric-redshift errors, and a wider mix of core-collapse subtypes, then recompute the post-cut purity. If purity falls below roughly 98%—the paper's own stated expectation for real data—the 99.5% claim and the 14%/18% cosmological forecasts do not transfer to the actual survey. Alternatively, once CSST-UDF data exist, spectroscopically confirm a subsample of the classifier's Type Ia candidates and measure the contamination fraction directly.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • A purely photometric CSST-UDF survey can deliver a >99.5% pure Type Ia sample of roughly 2,200 supernovae, with no spectroscopic follow-up.
  • The flat-wCDM constraints — 14% on Ω_M and 18% on w at 1σ — are comparable to those from spectroscopically confirmed samples.
  • The quality cuts add roughly 10% more usable supernovae than template-only classification in earlier work, while cutting contamination far lower.
  • Selection-effect bias, which grows toward z~1.3, is corrected to the point that residual core-collapse contamination can be dropped from the likelihood.
  • The full chain — recurrent-net classification, quality cuts, bias correction, and MCMC — is built to process real CSST-UDF data when they arrive.

Where Pith is reading between the lines

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

  • Because the training and testing sets are generated from the same templates and noise model, the 99.5% purity is a closed-loop score; real dust, calibration errors, photometric-redshift scatter, and CCSN subtypes beyond the templates could lower it. The paper's own note that real accuracy may fall below 98% suggests the 14%/18% constraints are best read as an upper-bound forecast.
  • Host-galaxy extinction, deliberately neglected in the simulation, directly mimics redder colors; since color is both a classification input and a template-fit parameter used in the quality cuts, adding extinction to the mock pipeline would be a sharper test of the cuts.
  • The same two-step binned likelihood could be applied to other multi-band photometric surveys, but the classifier and cuts would need retraining for each survey's cadence, filters, and depth.
  • If a future real-data run finds purity only near 98%, the simplified likelihood that drops the core-collapse term would need to be replaced by one that models residual contaminants; the simulated 99.5% is what justifies the simpler form.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper presents a mock forecast for supernova cosmology in the CSST Ultra-Deep Field. Light curves of SNe Ia and core-collapse SNe are simulated with SNCosmo using SALT3 and CCSN templates, and a SuperNNova LSTM classifier is trained and tested on these mocks. After applying JLA-like quality cuts, the authors report a sample of 2,193 SNe Ia and 4 CCSNe, corresponding to a purity above 99.5%. They then apply a binned BEAMS-with-Bias-Corrections (BBC) framework and MCMC to obtain cosmological constraints, finding Omega_M = 0.304 (+0.030/-0.052) and w = -1.017 (+0.177/-0.189), quoted as 14% and 18% relative accuracies in a flat wCDM model. The central claim is that a purely photometric CSST-UDF survey can match the cosmological precision of spectroscopically confirmed samples.

Significance. If the claimed purity and cosmological constraints are robust, the paper would be a useful demonstration that RNN-based classification plus BBC can extract competitive dark-energy constraints from a future photometric-only survey. The workflow is reproducible in the sense that the simulation, classifier, and fitting steps are described in enough detail and use public tools (SNCosmo, SuperNNova, emcee). The paper is therefore potentially valuable for CSST-UDF survey design. However, the magnitude of this value depends on how well the mock pipeline approximates real observations, and the current manuscript leaves several load-bearing assumptions unchecked.

major comments (3)
  1. [§4.1–4.2, Eqs. (6)–(11)] The selection-effect biases estimated in §4.1 (e.g., 0.036 mag at z=1.275) are never inserted into the BBC likelihood in §4.2. Equations (6) through (11) contain no bias-correction term. The fitted binned offsets Delta_mu,zeta absorb the selection bias, and the second-step vector D_mu,zeta in Eq. (11) does not include a -bias_zeta term. Either show explicitly where the correction enters, or state that Delta_mu,zeta is meant to include it and then subtract the estimated bias in Eq. (11). As written, the claimed 'correction' is not implemented, so the 14%/18% accuracy is not supported by the equations.
  2. [§4.1, Eqs. (2)–(3)] The selection-bias estimates are computed from 50 simulations in which Delta_mu uses mu_fid(z, theta_fid) with theta_fid = (Omega_M=0.3, w=-1), the same cosmology used to generate all mock SNe. If the true cosmology differs from this fiducial point, the redshift-dependent selection bias will differ, and a correction calibrated only at the fiducial cosmology can pull the fit toward theta_fid. The two reference cosmologies tested in §4.2 vary only the first-step reference model, not the cosmology used in the bias simulations. A demonstration with an alternate input cosmology for both the mock data and the bias estimate, or a formal cosmology-independence argument, is needed before the 14%/18% claim can be considered a robust forecast.
  3. [§3.3 and Abstract] The >99.5% purity is a self-consistency result: the classifier is trained and tested on light curves generated from the same SNCosmo/SALT3/CCSN models and the same selection function described in §2. The paper's own §3.3 admits that real classification accuracy could fall below 98% once dust, calibration errors, and photometric-redshift systematics are included. The abstract and summary present 99.5% without this caveat. Please frame the purity as an in-simulation upper bound and add the caveat at the abstract level, or include a sensitivity test with perturbed CCSN templates, host-galaxy extinction, and photometric calibration errors.
minor comments (6)
  1. [§3.2] Typo: 'for conservation purpose' should be 'for consistency purpose' or similar.
  2. [Figure 5] Typo in caption: 'Distance Moudle' should be 'Distance Modulus'.
  3. [§4.2, Eq. (9)] The photometric-redshift uncertainty sigma_z enters sigma_mu,z, but no numerical value or prior is given for sigma_z. Please state the assumed photo-z error and whether it comes from SALT3 fitting or external host-galaxy redshifts.
  4. [§4.2, Eq. (8)] The parametrization M = M0 + 5 log10(c/H0) should be checked for dimensional consistency and the units of c/H0 clarified, since the numerical value affects the fixed absolute magnitude M0 = -19.25.
  5. [Abstract and §4.2] The quoted '14%' accuracy for Omega_M is not transparent for an asymmetric error bar (+0.030/-0.052). Specify whether this is the upward error, the downward error, a symmetricized value, or something else.
  6. [References] The Riess et al. (2019) entry is duplicated; consolidate.

Circularity Check

0 steps flagged

No significant circularity: the analysis is a simulation-based forecast using external tools and explicitly stated assumptions.

full rationale

The paper is a mock-data forecast, not an empirical measurement. The derivation chain is: (1) generate light curves with SNCosmo, SALT3, and CCSN templates under CSST-UDF assumptions; (2) train SuperNNova on a separate simulated training set and evaluate on a held-out simulated test set; (3) fit SALT3 parameters and apply JLA-like cuts; (4) use a two-step BBC binning analysis and MCMC to constrain Omega_M and w. Each step uses independent draws and standard external tools. The classification accuracy is a generalization measure within the assumed simulation distribution, not a fit to the test labels, so it is not circular by construction. The BBC selection-effect simulations in Section 4.1 use the same fiducial cosmology as the mock data, but this is the standard method for calibrating a forecast pipeline: it does not algebraically force the recovered parameters to equal the inputs, and the paper explicitly tests two different reference cosmologies in Section 4.2. The constraints are presented as expected statistical precisions under the assumed model, not as a real-data measurement. The paper also explicitly acknowledges in Section 3.3 that real observations may lower classification accuracy below 98%, so it does not overclaim transferability. The self-citation to Wang et al. (2024) is used for simulation parameters and SALT3 fitting procedures, not as a load-bearing uniqueness or circularity argument. No quoted equation or fitted parameter is renamed as a prediction in a way that reduces to its own input.

Axiom & Free-Parameter Ledger

5 free parameters · 7 axioms · 0 invented entities

The central forecast rests on simulator fidelity (SALT3/CCSN templates plus CSST survey model), in-simulation ML metrics, and bias corrections calibrated with the same fiducial cosmology used to make the data. No new physical entities are introduced; the free parameters are standard SN-likelihood nuisance parameters plus redshift-bin offsets.

free parameters (5)
  • α (stretch-luminosity coefficient) = not reported (prior 0.08–0.32)
    Fitted in first-step BBC MCMC (Eq. 7); nuisance parameter in distance modulus.
  • β (color-luminosity coefficient) = not reported (prior 1–5)
    Fitted in first-step BBC MCMC (Eq. 7); nuisance parameter in distance modulus.
  • Δμ_ζ (24 redshift-bin distance-modulus offsets) = not reported
    Free parameters in first-step BBC fit; these offsets are the key input to the second-step cosmological fit (Eqs. 6, 11).
  • M0 (absolute SN Ia magnitude) = -19.25
    Fixed to the fiducial value used in the simulation (Sec. 4.2); anchors μ_i through M = M0 + 5 log10(c/H0), effectively using knowledge of the true input in the fit.
  • Selection-bias corrections per redshift bin = 0.014, 0.021, 0.028, 0.036 mag at z = 1.125, 1.175, 1.225, 1.275
    Computed from 50 simulations, not fitted, but calibrated with the same fiducial cosmology as the data; these corrections steer the recovered distances toward the fiducial model.
axioms (7)
  • domain assumption SALT3 SN Ia model and CCSN templates capture real SN diversity
    Light curves are generated with SNCosmo using SALT3 (Kenworthy 2021) and CCSN templates (Vincenzi 2019); if real supernovae have subtypes or diversity not in these templates, classifier purity and bias corrections are optimistic (§2).
  • domain assumption CSST-UDF survey parameters (limiting magnitudes, cadence, 60×250 s exposures, 9 deg² area) are accurate
    All mock light curves use these instrument/survey assumptions; changes in delivered image quality, PSF, or cadence would change selection and classification performance (§1, §2).
  • domain assumption Training and testing data are drawn from the same simulator
    The ML accuracy is measured on data from the same SNCosmo/SALT3/CCSN pipeline used for training, so the 99.1% accuracy quantifies performance on the simulation distribution, not on real CSST data (§3.2–3.3).
  • domain assumption BBC selection-bias simulations use the same fiducial cosmology and SN parameters as the mock data
    The bias corrections in §4.1 are built using Ω_M=0.3, w=−1 and the same rates/parameters used to generate the test data; this makes the correction partly self-referential.
  • domain assumption Host-galaxy extinction is neglected
    The authors state this tends to increase CCSN contamination, making purity conservative, but it also removes a real source of correlated scatter in distances and classification (§2).
  • domain assumption Residual CCSN contamination after cuts is negligible; D_CC term is dropped
    Eq. 10 keeps only the SN Ia term in the BBC likelihood because only 4 contaminants remain; if real contamination is higher, the simplified likelihood biases the constraints (§4.2).
  • domain assumption Flat wCDM is the true cosmology and the fitting model
    The fiducial model and the fitted model are both flat wCDM with Ω_M=0.3, w=−1; if the true expansion history differs, the quoted precision and accuracy are not representative (§1, §4).

pith-pipeline@v1.3.0-alltime-deepseek · 12660 in / 12963 out tokens · 132617 ms · 2026-08-04T00:05:38.599934+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Supernova Classification using the Recurrent Neural Network in the CSST Ultra-Deep Field Survey." pith.science (2026). https://pith.science/paper/ES2N6JNF

@misc{pith2026251102631,
  author       = {Pith},
  title        = {Pith review of: Supernova Classification using the Recurrent Neural Network in the CSST Ultra-Deep Field Survey},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ES2N6JNF}},
  note         = {Machine review of arXiv:2511.02631}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We study supernova (SN) classification using the Recurrent Neural Networks (RNNs) within the Chinese Space-station Survey Telescope Ultra-Deep Field (CSST-UDF) photometric survey and explore the improvements in cosmological constraints. We simulate Type Ia supernovae (SNe Ia) and core-collapse supernovae (CCSNe) using SNCosmo with SALT3 SN Ia model and CCSN templates, and apply the SuperNNova (SNN) program for classification. Our study indicates that the SNN combined with the Joint Light-curve Analysis cuts can enhance the purity of the CSST-UDF SN Ia sample up to over 99.5% with 2,193 SNe Ia and 4 CCSNe, which can significantly increase the reliability of the cosmological constraints. The method based on the Bayesian Estimation Applied to Multiple Species with Bias Corrections framework is used to correct the SN Ia magnitude bias caused by the selection effect and CCSN contamination, and the Markov Chain Monte Carlo (MCMC) method is employed for cosmological constraints. We find that the accuracy of the constraints on the matter density $\Omega_{\rm M}$ and the equation of state of dark energy parameter $w$ can achieve 14% and 18%, respectively, assuming the flat $w$CDM model. This result is comparable to current surveys relying on spectroscopic confirmation. Our results indicate that our data analysis method is effective, and the CSST-UDF SN photometric survey is a powerful tool to explore the expansion history of the Universe.

Figures

Figures reproduced from arXiv: 2511.02631 by Dejia Zhou, Minglin Wang, Xuelei Chen, Yan Gong.

Figure 1
Figure 1. Figure 1: The mock light curve examples for SNe Ia in different CSST-UDF photometric bands at redshifts between z = 0.28 and 1.3. The solid lines correspond to the theoretical expectations derived from the fiducial model. of CCSNe. The machine learning classifier is trained to learn the variations in light curves of different SN types in order to distinguish them. Selection criteria are critical for supernova cosmol… view at source ↗
Figure 2
Figure 2. Figure 2: The mock light curve examples at z ≃ 0.5 for the six types of CCSNe considered in this study. The solid lines denote the theoretical light curves derived from the fiducial model [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: A simplified architecture of an LSTM network. The cell state (Ct) carries information through time. The input layer receives data at each time step, while hidden lay￾ers compute and update the hidden state (Ht). The output layer generates predictions. The diagram illustrates the flow of information across time steps t ∈ [1, 4] [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: An LSTM cell within the hidden layer, comprising the cell state (Ct), hidden state (Ht), and input (Xt). It includes key components, i.e. Input Gate, Forget Gate, and Output Gate, which enable the model to effectively capture long-term dependencies. In [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The Hubble diagram as a function of input red￾shifts for the 2197 SNe which are classified by the SNN. The blue and orange data points denote SNe Ia and CCSNe, re￾spectively. In the lower panel, we show the residuals of the distance modulus and errors relative to the fiducial cosmol￾ogy for the 24 redshift bins. 4. COSMOLOGICAL CONSTRAINT In the cosmological constraint only using the SN pho￾tometric data, … view at source ↗
Figure 6
Figure 6. Figure 6: The predicted 1σ and 2σ contour maps and 1-D PDFs of ΩM and w assuming the flat wCDM model in the CSST-UDF SN photometric survey. show the 1D marginalized probability distribution func￾tions (PDFs) and 2D contour maps (1σ and 2σ) of ΩM vs. w. These results indicate that, despite rely￾ing solely on photometrically classified SNe, our method achieves constraint precision comparable to those ob￾tained from sp… view at source ↗

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

58 extracted references · 6 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter doi edition editor eprint howpublished institution journal key month number organization pages publisher school series title misctitle type volume year version url label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts ...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION format.url url empty "" new.block "" url * "" * if FUNCTION format.eprint eprint empty "" archivePrefix empty "" archivePrefix "arXiv" = new.block " " eprint * " " * new.block " " eprint * " " * if if if FUNCTION format.doi doi empty "" " " doi * " " * if FUNCTION format.pid doi empty eprint empty ur...

  3. [3]

    - [1] #1 = = ^ ^ ^ .\!\!^ d .\!\!^ h .\!\!^ m .\!\!^ s .\!\!^ @mss

    thebibliography [1] 20pt to REFERENCES 6pt =0pt -12pt 10pt plus 3pt =0pt =0pt =1pt plus 1pt =0pt =0pt -12pt =13pt plus 1pt =20pt =13pt plus 1pt \@M =10000 =-1.0em =0pt =0pt 0pt =0pt =1.0em @enumiv\@empty 10000 10000 `\.\@m \@noitemerr \@latex@warning Empty `thebibliography' environment \@ifnextchar \@reference \@latexerr Missing key on reference command E...

    2021

  4. [4]

    Abbott , T. M. C., Allam , S., Andersen , P., et al. 2019, , 872, L30, 10.3847/2041-8213/ab04fa

    doi:10.3847/2041-8213/ab04fa 2019

  5. [5]

    M., Lim , P

    Astropy Collaboration , Price-Whelan , A. M., Lim , P. L., et al. 2022, , 935, 167, 10.3847/1538-4357/ac7c74

    doi:10.3847/1538-4357/ac7c74 2022

  6. [6]

    C., Vincenzi, M., Scolnic, D., et al

    Bailey, A. C., Vincenzi, M., Scolnic, D., et al. 2023, Monthly Notices of the Royal Astronomical Society, 524, 5432–5441, 10.1093/mnras/stad2179

    doi:10.1093/mnras/stad2179 2023

  7. [7]

    2025, SNCosmo, v2.12.1, Zenodo, 10.5281/zenodo.15019859

    Barbary, K., Bailey, S., Barentsen, G., et al. 2025, SNCosmo, v2.12.1, Zenodo, 10.5281/zenodo.15019859

    doi:10.5281/zenodo.15019859 2025

  8. [8]

    G., Casertano, S., et al

    Breuval, L., Riess, A. G., Casertano, S., et al. 2024, The Astrophysical Journal, 973, 30, 10.3847/1538-4357/ad630e

    doi:10.3847/1538-4357/ad630e 2024

  9. [9]

    2022, The Astrophysical Journal, 938, 110, 10.3847/1538-4357/ac8e04

    Brout, D., Scolnic, D., Popovic, B., et al. 2022, The Astrophysical Journal, 938, 110, 10.3847/1538-4357/ac8e04

    doi:10.3847/1538-4357/ac8e04 2022

  10. [10]

    M., Vincenzi, M., et al

    Camilleri, R., Davis, T. M., Vincenzi, M., et al. 2024, Monthly Notices of the Royal Astronomical Society, 533, 2615–2639, 10.1093/mnras/stae1988

    work page doi:10.1093/mnras/stae1988 2024

  11. [11]

    2022, , 511, 1830, 10.1093/mnras/stac151

    Cao , Y., Gong , Y., Liu , D., et al. 2022, , 511, 1830, 10.1093/mnras/stac151

    work page doi:10.1093/mnras/stac151 2022

  12. [12]

    2024, Evaluating Cosmological Biases using Photometric Redshifts for Type Ia Supernova Cosmology with the Dark Energy Survey Supernova Program

    Chen, R., Scolnic, D., Vincenzi, M., et al. 2024, Evaluating Cosmological Biases using Photometric Redshifts for Type Ia Supernova Cosmology with the Dark Energy Survey Supernova Program. 2407.16744

    Pith/arXiv arXiv 2024

  13. [13]

    Collaboration, D., Abbott, T. M. C., Acevedo, M., et al. 2025, Dark Energy Survey: implications for cosmological expansion models from the final DES Baryon Acoustic Oscillation and Supernova data. 2503.06712

    arXiv 2025

  14. [14]

    2025, arXiv e-prints, arXiv:2507.04618, 10.48550/arXiv.2507.04618

    CSST Collaboration , Gong , Y., Miao , H., et al. 2025, arXiv e-prints, arXiv:2507.04618, 10.48550/arXiv.2507.04618

    doi:10.48550/arxiv.2507.04618 2025

  15. [15]

    DES Collaboration , Abbott , T. M. C., Acevedo , M., et al. 2024, arXiv e-prints, arXiv:2401.02929, 10.48550/arXiv.2401.02929

    Pith/arXiv arXiv doi:10.48550/arxiv.2401.02929 2024

  16. [16]

    Elman, J. L. 1990, Cognitive science, 14, 179

    1990

  17. [17]

    Fitzpatrick , E. L. 1999, , 111, 63, 10.1086/316293

    doi:10.1086/316293 1999

  18. [18]

    W., Lang , D., & Goodman , J

    Foreman-Mackey , D., Hogg , D. W., Lang , D., & Goodman , J. 2013, , 125, 306, 10.1086/670067

    doi:10.1086/670067 2013

  19. [19]

    F., Bianco, F

    Fortino, W. F., Bianco, F. B., Protopapas, P., Muthukrishna, D., & Brockmeier, A. 2025, ABC-SN: Attention Based Classifier for Supernova Spectra. 2507.22106

    Pith/arXiv arXiv 2025

  20. [20]

    2025, Optimizing Supernova Classification with Interpretable Machine Learning Models

    Garg, A. 2025, Optimizing Supernova Classification with Interpretable Machine Learning Models. 2510.13765

    arXiv 2025

  21. [21]

    2010, , 709, 1420, 10.1088/0004-637X/709/2/1420

    Gong , Y., Cooray , A., & Chen , X. 2010, , 709, 1420, 10.1088/0004-637X/709/2/1420

    work page doi:10.1088/0004-637x/709/2/1420 2010

  22. [22]

    2019, , 883, 203, 10.3847/1538-4357/ab391e

    Gong , Y., Liu , X., Cao , Y., et al. 2019, , 883, 203, 10.3847/1538-4357/ab391e

    doi:10.3847/1538-4357/ab391e 2019

  23. [23]

    2025, Future Cosmology: New Physics and Opportunity from the China Space Station Telescope (CSST)

    Gong, Y., Miao, H., Zhou, X., et al. 2025, Future Cosmology: New Physics and Opportunity from the China Space Station Telescope (CSST). 2501.15023

    Pith/arXiv arXiv 2025

  24. [24]

    2012, The Astrophysical Journal, 752, 79, 10.1088/0004-637x/752/2/79

    Hlozek, R., Kunz, M., Bassett, B., et al. 2012, The Astrophysical Journal, 752, 79, 10.1088/0004-637x/752/2/79

    doi:10.1088/0004-637x/752/2/79 2012

  25. [25]

    Hopfield, J. J. 1982, Proceedings of the national academy of sciences, 79, 2554

    1982

  26. [26]

    M., Tyson , J

    Ivezi \'c , Z ., Kahn , S. M., Tyson , J. A., et al. 2019, , 873, 111, 10.3847/1538-4357/ab042c

    doi:10.3847/1538-4357/ab042c 2019

  27. [27]

    O., Scolnic , D

    Jones , D. O., Scolnic , D. M., Riess , A. G., et al. 2017, , 843, 6, 10.3847/1538-4357/aa767b

    doi:10.3847/1538-4357/aa767b 2017

  28. [28]

    D., Jones , D

    Kenworthy , W. D., Jones , D. O., Dai , M., et al. 2021, , 923, 265, 10.3847/1538-4357/ac30d8

    doi:10.3847/1538-4357/ac30d8 2021

  29. [29]

    2017, , 836, 56, 10.3847/1538-4357/836/1/56

    Kessler , R., & Scolnic , D. 2017, , 836, 56, 10.3847/1538-4357/836/1/56

    doi:10.3847/1538-4357/836/1/56 2017

  30. [30]

    2023, The Astrophysical Journal Letters, 952, L8, 10.3847/2041-8213/ace34d

    Kessler, R., Vincenzi, M., & Armstrong, P. 2023, The Astrophysical Journal Letters, 952, L8, 10.3847/2041-8213/ace34d

    doi:10.3847/2041-8213/ace34d 2023

  31. [31]

    A., & Sen, A

    Kumar, D., Mitra, A., Adil, S. A., & Sen, A. A. 2025, Physical Review D, 111, 10.1103/physrevd.111.043503

    doi:10.1103/physrevd.111.043503 2025

  32. [32]

    Leoni, M., Ishida, E. E. O., Peloton, J., & Möller, A. 2022, Astronomy & Astrophysics, 663, A13, 10.1051/0004-6361/202142715

    doi:10.1051/0004-6361/202142715 2022

  33. [33]

    2025, Machine Learning in Stellar Astronomy: Progress up to 2024

    Li, G., Lu, Z., Wang, J., & Wang, Z. 2025, Machine Learning in Stellar Astronomy: Progress up to 2024. 2502.15300

    Pith/arXiv arXiv 2025

  34. [34]

    2023, Science China Physics, Mechanics, and Astronomy, 66, 229511, 10.1007/s11433-022-2018-0

    Li , S.-Y., Li , Y.-L., Zhang , T., et al. 2023, Science China Physics, Mechanics, and Astronomy, 66, 229511, 10.1007/s11433-022-2018-0

    work page doi:10.1007/s11433-022-2018-0 2023

  35. [35]

    2024, Science China Physics, Mechanics & Astronomy, 67, 10.1007/s11433-024-2456-x

    Liu, C., Xu, Y., Meng, X., et al. 2024, Science China Physics, Mechanics & Astronomy, 67, 10.1007/s11433-024-2456-x

    work page doi:10.1007/s11433-024-2456-x 2024

  36. [36]

    P., Kessler , R., et al

    Marriner , J., Bernstein , J. P., Kessler , R., et al. 2011, , 740, 72, 10.1088/0004-637X/740/2/72

    doi:10.1088/0004-637x/740/2/72 2011

  37. [37]

    A., et al

    Mellier, Y., Abdurro’uf, Acevedo Barroso, J. A., et al. 2025, Astronomy & Astrophysics, 697, A1, 10.1051/0004-6361/202450810

    doi:10.1051/0004-6361/202450810 2025

  38. [38]

    2020, , 491, 4277, 10.1093/mnras/stz3312

    M \"o ller , A., & de Boissi \`e re , T. 2020, , 491, 4277, 10.1093/mnras/stz3312

    doi:10.1093/mnras/stz3312 2020

  39. [39]

    2022, , 514, 5159, 10.1093/mnras/stac1691

    M \"o ller , A., Smith , M., Sako , M., et al. 2022, , 514, 5159, 10.1093/mnras/stac1691

    doi:10.1093/mnras/stac1691 2022

  40. [40]

    2024, The Dark Energy Survey 5-year photometrically classified type Ia supernovae without host-galaxy redshifts

    Möller, A., Wiseman, P., Smith, M., et al. 2024, The Dark Energy Survey 5-year photometrically classified type Ia supernovae without host-galaxy redshifts. 2402.18690

    Pith/arXiv arXiv 2024

  41. [41]

    1999, The Astrophysical Journal, 517, 565–586, 10.1086/307221

    Perlmutter, S., Aldering, G., Goldhaber, G., et al. 1999, The Astrophysical Journal, 517, 565–586, 10.1086/307221

    doi:10.1086/307221 1999

  42. [42]

    2021, The Astronomical Journal, 162, 67, 10.3847/1538-3881/ac0824

    Qu, H., Sako, M., Möller, A., & Doux, C. 2021, The Astronomical Journal, 162, 67, 10.3847/1538-3881/ac0824

    doi:10.3847/1538-3881/ac0824 2021

  43. [44]

    G., Casertano, S., Yuan, W., Macri, L

    Riess, A. G., Casertano, S., Yuan, W., Macri, L. M., & Scolnic, D. 2019, The Astrophysical Journal, 876, 85, 10.3847/1538-4357/ab1422

    doi:10.3847/1538-4357/ab1422 2019

  44. [45]

    G., Filippenko , A

    Riess , A. G., Filippenko , A. V., Challis , P., et al. 1998, , 116, 1009, 10.1086/300499

    doi:10.1086/300499 1998

  45. [46]

    G., Yuan , W., Macri , L

    Riess , A. G., Yuan , W., Macri , L. M., et al. 2022, , 934, L7, 10.3847/2041-8213/ac5c5b

    doi:10.3847/2041-8213/ac5c5b 2022

  46. [47]

    A., Riess , A

    Rodney , S. A., Riess , A. G., Strolger , L.-G., et al. 2014, , 148, 13, 10.1088/0004-6256/148/1/13

    doi:10.1088/0004-6256/148/1/13 2014

  47. [48]

    M., Baltay, C., Hounsell, R., et al

    Rose, B. M., Baltay, C., Hounsell, R., et al. 2021, A Reference Survey for Supernova Cosmology with the Nancy Grace Roman Space Telescope. 2111.03081

    Pith/arXiv arXiv 2021

  48. [49]

    E., Hinton, G

    Rumelhart, D. E., Hinton, G. E., & Williams, R. J. 1986, Nature, 323, 533. https://api.semanticscholar.org/CorpusID:205001834

    1986

  49. [50]

    E., Barclay , T., Barnes , A., et al

    Schlieder , J. E., Barclay , T., Barnes , A., et al. 2024, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 13092, Space Telescopes and Instrumentation 2024: Optical, Infrared, and Millimeter Wave, ed. L. E. Coyle , S. Matsuura , & M. D. Perrin , 130920S, 10.1117/12.3020622

    doi:10.1117/12.3020622 2024

  50. [51]

    2022, , 938, 113, 10.3847/1538-4357/ac8b7a

    Scolnic , D., Brout , D., Carr , A., et al. 2022, , 938, 113, 10.3847/1538-4357/ac8b7a

    doi:10.3847/1538-4357/ac8b7a 2022

  51. [52]

    A., et al

    Strolger , L.-G., Dahlen , T., Rodney , S. A., et al. 2015, , 813, 93, 10.1088/0004-637X/813/2/93

    doi:10.1088/0004-637x/813/2/93 2015

  52. [53]

    E., et al

    Vincenzi , M., Sullivan , M., Firth , R. E., et al. 2019, , 489, 5802, 10.1093/mnras/stz2448

    doi:10.1093/mnras/stz2448 2019

  53. [54]

    2022, Monthly Notices of the Royal Astronomical Society, 518, 1106–1127, 10.1093/mnras/stac1404

    Vincenzi, M., Sullivan, M., Möller, A., et al. 2022, Monthly Notices of the Royal Astronomical Society, 518, 1106–1127, 10.1093/mnras/stac1404

    doi:10.1093/mnras/stac1404 2022

  54. [55]

    Vincenzi, M., Brout, D., Armstrong, P., et al. 2024. 2401.02945

    Pith/arXiv arXiv 2024

  55. [56]

    2024, Monthly Notices of the Royal Astronomical Society, 530, 4288, 10.1093/mnras/stae1119

    Wang, M., Gong, Y., Deng, F., et al. 2024, Monthly Notices of the Royal Astronomical Society, 530, 4288, 10.1093/mnras/stae1119

    work page doi:10.1093/mnras/stae1119 2024

  56. [57]

    2011, SCIENTIA SINICA Physica, Mechanica & Astronomica, 41, 1441

    Zhan, H. 2011, SCIENTIA SINICA Physica, Mechanica & Astronomica, 41, 1441. http://www.sciengine.com/publisher/Science China Press/journal/SCIENTIA SINICA Physica, Mechanica & Astronomica/41/12/10.1360/132011-961

    doi:10.1360/132011-961 2011

  57. [58]

    2021, Chinese Science Bulletin, 66, 1290, https://doi.org/10.1360/TB-2021-0016

    ---. 2021, Chinese Science Bulletin, 66, 1290, https://doi.org/10.1360/TB-2021-0016

    doi:10.1360/tb-2021-0016 2021

  58. [59]

    ?O' ]k!65fkMs z )5Vz'γ&4rnaAǧ˥溛y zM4\

    thebibliography [1] 20pt to REFERENCES 6pt =0pt -12pt 10pt plus 3pt =0pt =0pt =1pt plus 1pt =0pt =0pt -12pt =13pt plus 1pt =20pt =13pt plus 1pt \@M =10000 =-1.0em =0pt =0pt 0pt =0pt =1.0em @enumiv\@empty 10000 10000 `\.\@m \@noitemerr \@latex@warning Empty `thebibliography' environment \@ifnextchar \@reference \@latexerr Missing key on reference command E...

    arXiv 2021