Pith. sign in

REVIEW 3 major objections 5 minor 3 cited by

ABC-SN: Attention Based Classifier for Supernova Spectra

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read ABC-SN claims an attention-based classifier for supernova spectra that beats a retrained DASH baseline, with macro F1 82.45% versus 58.86% on ten subtypes.

desk verdict Solid transformer-based SN classifier, but the DASH comparison is unfair and needs a retrain before the headline numbers are believable. read the letter →

arxiv 2507.22106 v3 pith:2CKUUIMO submitted 2025-07-29 astro-ph.IM astro-ph.SR

Willow Fox Fortino , Federica B. Bianco , Pavlos Protopapas , Daniel Muthukrishna , Austin Brockmeier This is my paper · ORCID
classification astro-ph.IMastro-ph.SR
keywords supernovaclassificationspectralattentionmechanismtransformerencoderDASHbenchmarkSNIDtemplateslow-resolutionspectroscopymaskedpretraining
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that an attention-based neural network, ABC-SN, can classify supernova spectra into ten subtypes more accurately than the leading deep-learning baseline DASH when both are trained on the same SNID-derived dataset. On a test set of 1,880 spectra, ABC-SN attains a macro F1-score of 82.45% against DASH's 58.86%, with SN Ia completeness rising from roughly 63% to roughly 88% and purity from roughly 83% to roughly 95%. The authors argue this matters because the coming wide-field survey will produce too many transients for human template matching, and low-resolution ($R=100$) spectra suffice for the attention model, easing telescope follow-up demands. The paper positions ABC-SN as an up-to-date general-purpose spectroscopic classifier and releases pretrained weights and code.

What carries the argument

The central object is ABC-SN's encoder-only transformer. Input spectra are embedded with a multilayer perceptron into (64, 128) tokens, a Fourier positional encoding injects wavelength information, six encoder blocks with multi-head self-attention and residual connections process the token sequence, and a three-layer feed-forward head with softmax outputs class probabilities. The load-bearing training device is generative pretraining: the encoder weights are first trained to reconstruct a randomly masked 15% of each spectrum (plus 2.5% perturbed points), then fine-tuned for classification; this masked-spectrum pretraining is what the paper credits for controlling overfitting. The other key mechanism is resolution reduction: spectra are convolved with a Gaussian and rebinned to $R=100$, making the $O(N_{\rm wvl}^2)$ attention mechanism memory-feasible without, the paper argues, losing classification power.

What would settle it

Retrain DASH on exactly the same ten subtypes with the same decontaminated split and early stopping protocol, then evaluate both models on a third set of spectra never used for early stopping or architecture selection; if DASH's macro F1 rises from 58.86% toward ABC-SN's 82.45%, the claimed advantage collapses. The paper's own Figure 1 already shows DASH's test F1 is sensitive to training schema, so this is a directly checkable comparison.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a transformer-style encoder with multi-head self-attention, a learned MLP embedding, Fourier positional encoding, and a masked-spectrum pretraining stage can learn the spectral feature combinations that distinguish ten SN subtypes from the SNID library, and that lowering the spectra from $R=738$ to $R=100$ does not hurt classification. Benchmarking against DASH retrained with early stopping and a decontaminated train/test split, ABC-SN reports macro F1 82.45% versus 58.86%, recall above 75% for every class, and stronger completeness and purity for almost every subtype, with Ibn purity as the only exception and that class containing only three SNe. The authors state that the model transfers to unseen spectra from a newer SNID sample and to fourteen normal SN Ia spectra.

Load-bearing premise

The load-bearing premise is that a DASH model trained on all 16 subtypes but evaluated on only 10 is a fair baseline; because DASH reserves probability for six excluded classes and the test set was used for early stopping and architecture selection, the reported margin is likely optimistic.

Editorial extensions

If this is right

  • If the reported accuracy holds, ABC-SN gives the community a general-purpose, open-source spectral classifier for ten subtypes that runs at roughly 128 microseconds per spectrum on a single T4 GPU.
  • SN Ia purity near 95% at no probability threshold means that even without human vetting, a large fraction of the recovered Ia sample is clean; users can trade completeness for purity with per-class thresholds using the provided ROC curves.
  • Demonstrating that $R=100$ spectra suffice suggests follow-up programs can use lower-resolution spectrographs and still obtain subtype-level classifications.
  • The decontaminated DASH retraining provides an updated baseline that future automated SN classifiers can compare against on the same SNID-derived dataset.
  • Transfer results on 194 unseen spectra (weighted precision 0.80, recall 0.82) support the claim that the model generalizes beyond its training objects.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the headline comparison may overstate ABC-SN's edge, because DASH was retrained on all 16 subtypes while Table 2 reports only its 10-class confusion matrices; the six excluded classes absorb probability mass, and a 10-class retrained DASH could close part of the 23.6-point macro F1 gap.
  • Editorial inference: since the same test set guided early stopping and architecture selection, the quoted 82.45% macro F1 is likely optimistic; an independent test set held out from all model selection decisions would give the operational accuracy.
  • Editorial inference: a natural extension is to use the same masked-spectrum pretraining on a larger, more balanced library, or to predict continuous physical parameters (phase, redshift, ejecta velocity) rather than discrete subtypes; the paper itself notes the discrete taxonomy is questionable.
  • Editorial inference: the attention mechanism's success at $R=100$ suggests a testable extension to real-time classification of alert-stream spectra, provided the 4,500--7,000 Angstrom padded window does not exclude high-redshift features.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper introduces ABC-SN, an attention-based encoder classifier for ten supernova subtypes, trained on a curated dataset of 3,764 spectra from 498 SNe derived from the SNID library and related catalogs. The authors retrain the DASH convolutional classifier under three protocols (original, early-stopped, and decontaminated) and report that ABC-SN attains a macro F1-score of 82.45% on the test set versus 58.86% for the decontaminated DASH benchmark, with per-class completeness and purity improvements in nearly all classes. The paper also includes an architecture search, a masked-spectrum pretraining stage, a study of performance versus spectral resolution, and a transferability test on unseen spectra from Magill et al. (2025) plus manually selected normal SN Ia spectra.

Significance. If the headline comparison were sound, the paper would be a useful contribution to spectroscopic supernova classification: ABC-SN is released as open source with pretrained weights, the authors carefully prevent train/test contamination at the level of individual supernovae, they demonstrate that low-resolution (R=100) spectra suffice, and they provide a detailed and reproducible account of architectural choices and hyperparameters. The paper also makes a sensible methodological point that benchmark comparisons should be performed on shared, identically preprocessed datasets. However, the central quantitative claim that ABC-SN outperforms DASH is currently not established because the DASH benchmark is not trained on the same label set as ABC-SN, and because the test set is used for early stopping, architecture selection, and final reporting without an independent holdout.

major comments (3)
  1. [§2, §5, Table 2] The central comparison is confounded by a label-set mismatch. Section 2 states explicitly that 'All three instances of DASH are trained with all original 16 SN subtypes,' while Figure 1 and Table 2 evaluate DASH only on the ten ABC-SN subtypes. A DASH model with a 16-class output layer must spread probability mass over six classes that are never correct in the reported evaluation, and any test spectrum predicted into one of those six excluded classes is counted as an error. ABC-SN, trained only on the ten classes, never incurs this penalty. The reported 82.45% versus 58.86% macro F1 gap, the per-class completeness and purity numbers in Table 2, and the abstract's 26% SN Ia completeness improvement therefore do not measure the same task. The authors should retrain DASH on exactly the ten ABC-SN subtypes, or alternatively renormalize DASH's probabilities over the ten classes and re-evaluate, before claiming that ABC-SN outperforms DASH.
  2. [§2, §4.1, Figure 7, Table 4] The reported test-set scores are not unbiased because the same 'test' set is used for model development and final evaluation. Section 2 explicitly defines the two-way split as 'the set used for hyperparameter tuning' and calls it 'test' following Bishop and Nasrabadi (2006); Section 4.1 states that 'It is the F1-score of the test set that we monitor during training'; the architecture selection in Figure 7 is performed on test-set F1; and Table 4's early stopping and Reduce-Learning-Rate-Plateau callbacks use test-set F1 or MSE. Consequently, the headline figures 82.45% and 58.86% are selected, at least in part, on the same data used to report them. The authors should either hold out a truly unseen set for final reporting after all selection decisions are frozen, or use nested cross-validation, and should state the variance over their five random-seed models for both ABC-SN and DASH.
  3. [§5, Table 3] The unseen-data transferability test does not rescue the DASH comparison and is itself incomplete for the most important class. Table 3 reports precision and recall for seven subtypes on the Magill et al. (2025) sample, but no score for Ia-norm or Ib-norm; the Ia-norm check rests on 14 manually selected spectra from five well-studied SNe, two of which are classified as Ia-91T. This is a reasonable sanity check, but it is not a benchmark against DASH on unseen data, and the sample is too small to support the strong statement in the Conclusion that 'The model transferability is confirmed on unseen data' for the dominant Ia-norm class. The authors should either report DASH on the same unseen sample or soften the transferability claim accordingly.
minor comments (5)
  1. [§1 and §2] The number of DASH subtypes is inconsistent: Section 2 says 'all original 16 SN subtypes,' while the text describing DASH's output in Section 1 refers to '306 type-age classes (18 age classes for 17 subtypes).' Please clarify which number is correct.
  2. [§5, paragraph on thresholds] The sentence 'at classification threshold p = 0.92 we obtain >3% purity on SN Ia with a completeness of >76%' appears to contain a typo: the SN Ia purity should presumably read '>93%,' since the next clause says 'for all other classes, purity is better than 3%.' As written, the sentence is self-contradictory.
  3. [Table 2] The row labeled 'Ia' overlaps with the rows 'Ia-norm,' 'Ia-91T,' 'Ia-91bg,' and 'Iax.' Please define explicitly whether 'Ia' is the aggregate of those four subtypes or a distinct class, and apply the same clarification to 'Ibc.'
  4. [Figure 1] The caption should state how predictions into the six excluded DASH subtypes are treated in the displayed confusion matrices: if such predictions are omitted from the matrices, the row and column percentages do not sum to 100%, and the matrices do not fully represent DASH's behavior; if they are counted as errors, that should be stated explicitly. This is closely tied to Major Comment 1 but should be fixed in the figure regardless.
  5. [§3.3 and §5] The paper reports that inverting the train/test split gives 'virtually identical performance' but provides no numerical result. Since the small classes consist of only three to six supernovae, a quantitative statement of the two-fold results, including per-class scores for the smallest classes, would help the reader assess robustness.

Circularity Check

0 steps flagged · score 1.0 of 10

No material circularity: ABC-SN is an empirical benchmark against external SNID labels; minor self-citations and a fairness mismatch in the DASH comparison do not reduce the result to its inputs.

full rationale

ABC-SN's central claims are macro F1, completeness, and purity measured on held-out SNID-labeled spectra and on unseen Magill et al. (2025) spectra. These are external ground-truth evaluations, not quantities defined in terms of the model's own fitted parameters. The DASH comparison is an empirical baseline: the authors retrain DASH themselves using code from Muthukrishna et al. (2019), so the comparison is not a self-citation carrying the conclusion. The Fourier positional encoding is selected from a six-way architecture sweep (Figure 7), with the Moreno-Cartagena et al. (2023) citation (which includes a coauthor) serving only as a menu of options, not as an unverified premise. The paper itself flags its principal validity limitations: Section 2 uses the 'test' set for hyperparameter tuning and early stopping, and Section 3.3 states that for classes populated by only three (Ibn) or six (Iax, IIP) SNe 'there is no way to ensure that the classification power will remain as measured' if the sample is not representative. These are correctness risks, not circularity. The most serious concern is benchmark fairness: 'All three instances of DASH are trained with all original 16 SN subtypes. However, in Figure 1 we present confusion matrices for these models including only the ten subtypes seen in Figure 2' (Section 2), so DASH's softmax reserves probability for six excluded classes and its 10-class scores may be artificially depressed. This is an apples-to-oranges experimental comparison that could inflate the reported gap, but it is not a reduction of the claimed result to a fitted parameter or to a self-citation chain. No equation in the paper equates the headline F1 to an input by construction. Accordingly, no circular step is exhibited and the circularity score is low.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The central claim rests on a supervised learning setup whose main assumptions are label reliability, the chosen class taxonomy, the resolution reduction, and the validity of using the test set for model selection. The hyperparameters and architecture choices are fitted to the test set, which is a significant free-parameter burden that affects the reported scores. No new physical entities are introduced.

free parameters (8)
  • Learning rates = 1e-4 (pretrain), 1e-5 (fine-tune)
    Chosen by hand with reduce-on-plateau; affect convergence and final performance.
  • Dropout rate = 0.5
    Applied in encoder and head; chosen to combat overfitting as discussed in Section 4.2.3.
  • L2 regularization = 0.01
    Applied to each Dense layer in the classification head (Section 4.2.4).
  • Number of encoder blocks = 6
    Architecture choice; the encoder block can be repeated arbitrarily and 6 was adopted (Figure 6).
  • Architecture selection (embedding, PE, head) = MLP embedding, Fourier PE, FF head
    Selected by comparing test-set F1 across 5 embeddings, 6 positional encodings, and 3 heads (Figure 7), i.e., tuned on the test set.
  • Data augmentation parameters = noise sigma 0.10 (Section 3.4) or 0.01 (Table 4); shift +/-5 bins; 0-4 spikes; spike amplitude 2x spectrum sigma; 80%…
    Chosen to balance classes and increase robustness; the paper's two stated noise sigma values are inconsistent.
  • Masking fraction = 15% plus 2.5% perturbation
    Pretraining hyperparameter for masked-spectrum reconstruction (Section 4.3).
  • Early stopping and RLRP thresholds = Patience 25; delta F1 0.005; lr factor 0.5
    Stopping criteria determine the final epoch (75 of 100) and thus final weights.
assumptions (5)
  • domain assumption SNID labels are treated as ground truth for training and evaluation.
    Section 3 states 'we start with the SNID labels as our ground truth.' The Magill test reveals some labels are wrong, so this is load-bearing.
  • domain assumption All spectra from the same SN must be in the same train/test split to prevent contamination.
    Section 2: spectra from the same SN are correlated; enforced in the split.
  • domain assumption Lowering resolution to R=100 preserves classification information for ABC-SN.
    Section 3.2 demonstrates this for DASH only (Figure 4), then assumes it applies to ABC-SN.
  • domain assumption The ten selected subtypes form a coherent classification target.
    Section 3 removes seven subtypes for low counts or because 'peculiar' labels were catch-alls; this choice defines the problem.
  • ad hoc to paper The two-way split called 'test' can be used for hyperparameter selection and still yield unbiased final scores.
    Section 2 explicitly names the hyperparameter-tuning set 'test'; the reported headline numbers come from this set.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ABC-SN: Attention Based Classifier for Supernova Spectra." pith.science (2026). https://pith.science/paper/2CKUUIMO

@misc{pith2026250722106,
  author       = {Pith},
  title        = {Pith review of: ABC-SN: Attention Based Classifier for Supernova Spectra},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2CKUUIMO}},
  note         = {Machine review of arXiv:2507.22106}
}
abstract

While significant advances have been made in photometric classification ahead of the millions of transient events and hundreds of supernovae (SNe) each night that the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) will discover, classifying SNe spectroscopically remains the best way to determine most subtypes of SNe. Traditional spectrum classification tools use template matching techniques (eg., SNID, SuperFit) and require significant human supervision. Two deep learning spectral classifiers, DASH and SNIascore, define the state of the art, but SNIascore is a binary classifier devoted to maximizing the purity of the SN Ia-norm sample, while DASH is no longer maintained and the original work suffers from contamination of multi-epoch spectra in the training and test sets. We have explored several neural network architectures in order to create a new automated method for classifying SN subtypes, settling on an attention-based model we call ABC-SN. We benchmark our results against an updated version of DASH, thus providing the community with an up-to-date general-purpose SN classifier. Our dataset is comprised of ten different SN subtypes, including subtypes of SN Ia, core collapse, and interacting SNe. We find that ABC-SN outperforms DASH for nearly all classes, including an improvement of $26\%$ in SN Ia completeness ($\sim88\%$) and $2.4\%$ in SN Ia purity ($\sim95\%$) when unthresholded (improvements for each class can further be obtained by tuned thresholds), and we discuss the limitations of current SN datasets for benchmarking performance.

Figures

Figures reproduced from arXiv: 2507.22106 by the authors.

Figure 1
Figure 1. DASH (Muthukrishna et al. 2019) SN classifier performance under different training schemas for the classes considered in this work (section 3). The top row shows model completeness (i.e., recall) by normalizing across true labels. The bottom row shows purity (i.e., precision) by normalizing across the predictions. In each confusion matrix, correct predictions (diagonal cells) are plotted in shades of blue and incorr… view at source ↗
Figure 2
Figure 2. A treemap plot showing the composition of our dataset of SN spectra after preprocessing. The size of each rectangle corresponds to the proportion of the ten SN sub￾types in our dataset. The color of each rectangle corresponds to which main SN type — Ia, Ib, Ic or II — the subtype be￾longs to. There are 3, 764 total spectra from 498 SNe. This plot shows the extreme class imbalance in this dataset, with ∼ 73% of the s… view at source ↗
Figure 4
Figure 4. This figure shows how the performance of DASH changes as a function of the spectral resolution of the dataset it is trained on. We lower the spectral resolution of our dataset to 40 different values between the native resolution, R = 738 and R = 5 in order to examine how the perfor￾mance of DASH would change. In blue, the macro F1-score on the test set is plotted at 41 different spectral resolutions. In orange, the … view at source ↗
Figures from the paper (11 more)
Figure 3
Figure 3. Figure 3: Three spectra from SN1998dt, a type Ib SN, ob￾served 1.8 days after peak brightness. The blackbody contin￾uum has been removed. Type Ib SNe do not show hydrogen spectral lines nor the Si IIλ6355 that characterizes SNe Ia. Top: The spectrum is plotted at the original hi…
Figure 5
Figure 5. Figure 5: A flowchart that summarizes the operations we perform on our spectral data. All preprocessing steps are described in subsection 3.1. more friendly to the Transformer architecture. The first was to bin each flux value in the spectra according to what quantile that flux …
Figure 6
Figure 6. Figure 6: The conceptual diagram of the ABC-SN architec￾ture described in subsection 4.2. that both the GAF and MAF generally performed worse than having no embedding. The fourth embedding method is a learned embedding which uses a Conv1D layer with a kernel size of 1 and Nembed…
Figure 7
Figure 7. Figure 7: Results of tests performed on different architecture element combinations. Top: each box corresponds to a com￾bination embedding (y axis) — positional encoding (x axis) as labeled and described in sections 4.2.1 and 4.2.2, respectively and average over the three tested…
Figure 8
Figure 8. Figure 8: The loss curve during ABC-SN pretraining. The mean squared error (MSE) is shown for the training and test sets during pretraining. The model was pretrained for 141 epochs on reconstruction of masked spectra (see subsec￾tion 4.3) before reverting back to epoch 116 due t…
Figure 9
Figure 9. Figure 9: The loss and metric curves for ABC-SN during training. Top: The categorical cross-entropy loss. Bottom: The categorical accuracy (solid lines) and macro F1-score (dotted lines). ABC-SN was set to train for 100 epochs before reverting back to epoch 75 due to an early st…
Figure 10
Figure 10. Figure 10: Confusion matrix showing the performance of ABC-SN in the classification of SNe in the SNID dataset. On the horizontal axis, the predicted class. On the vertical axis, the label. As in figure [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: ROC curve for ABC-SN One-vs-rest classifica￾tion for each considered class. The insert shows a zoom in of the top left region of the ROC space. We note that the smallest classes (IIP and Ib-n, 6 objects, and Iax, 3 objects, dashed or dotted lines) stand out in the ROC…
Figure 12
Figure 12. Figure 12: shows a holistic view of ABC-SN performance with three test set confusion matrices. The first shows the absolute performance of ABC-SN, providing raw values for the number of classified spectra. The second confusion matrix shows the same data but normalized by the num…
Figure 13
Figure 13. Figure 13: Same as [PITH_FULL_IMAGE:figures/full_fig_p020_13.png]
Figure 14
Figure 14. Figure 14: Same as [PITH_FULL_IMAGE:figures/full_fig_p020_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. AGNFormer I: Reconstruction of AGN spectra using a probabilistic transformer model

    astro-ph.GA 2026-07 conditional novelty 6.0 of 10

    An uncertainty-aware transformer reconstructs masked AGN broad lines and spectral halves with 4-16% flux errors and beats eleven purpose-built Lyα-reconstruction algorithms on a blind benchmark.

  2. Catching Disguised Transients with ASTRANet: Anomaly-Aware Spectroscopic Classification and Conformal Calibration

    astro-ph.IM 2026-07 conditional novelty 6.0 of 10

    ASTRANet combines a redshift-free spectral classifier, a 16-score anomaly detector, and conformal prediction to identify and calibrate uncertainty for out-of-taxonomy astronomical transients.

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

    astro-ph.CO 2025-11 conditional novelty 4.0 of 10

    In mock CSST-UDF data, an LSTM classifier plus JLA-like cuts yields a >99.5% pure Type Ia sample and recovers Ω_M and w to 14% and 18% in a flat wCDM model.

Reference graph

Works this paper leans on

65 extracted references · 21 canonical work pages · cited by 3 Pith papers

  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]

    E., Tripathi, A., et al

    Ahmed, S., Nielsen, I. E., Tripathi, A., et al. 2022, Transformers in Time -series Analysis : A Tutorial , arXiv, 10.48550/arXiv.2205.01138

    arXiv doi:10.48550/arxiv.2205.01138 2022

  5. [5]

    E., et al

    Andrew Howell, D., Sullivan, M., Nugent, P. E., et al. 2006, Nature, 443, 308

    work page 2006

  6. [6]

    D., Walters, R., et al

    Blagorodnova, N., Neill, J. D., Walters, R., et al. 2018, Publications of the Astronomical Society of the Pacific, 130, 035003, 10.1088/1538-3873/aaa53f

    doi:10.1088/1538-3873/aaa53f 2018

  7. [7]

    Blondin, S., & Tonry, J. L. 2007, The Astrophysical Journal, 666, 1024, 10.1086/520494

    doi:10.1086/520494 2007

  8. [8]

    P., et al

    Blondin, S., Matheson, T., Kirshner, R. P., et al. 2012, The Astronomical Journal, 143, 126, 10.1088/0004-6256/143/5/126

    doi:10.1088/0004-6256/143/5/126 2012

Show all 65 references
  1. [9]

    2019, Astronomical Journal, 158, 257, 10.3847/1538-3881/ab5182

    Boone , K. 2019, Astronomical Journal, 158, 257, 10.3847/1538-3881/ab5182

    2019 doi

  2. [10]

    2022, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 745--755

    Cai, Y., Lin, J., Lin, Z., et al. 2022, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 745--755

    2022

  3. [11]

    2015, Keras, https://keras.io

    Chollet, F., et al. 2015, Keras, https://keras.io

    2015

  4. [12]

    Couch, S. M. 2017, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 375, 20160271, 10.1098/rsta.2016.0271

    2017

  5. [13]

    2019, BERT : Pre -training of Deep Bidirectional Transformers for Language Understanding , arXiv, 10.48550/arXiv.1810.04805

    Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. 2019, BERT : Pre -training of Deep Bidirectional Transformers for Language Understanding , arXiv, 10.48550/arXiv.1810.04805

    2019 doi arXiv

  6. [14]

    2020, arXiv preprint arXiv:2010.11929

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., et al. 2020, arXiv preprint arXiv:2010.11929

    2020 arXiv

  7. [15]

    2020, Journal of Machine Learning Research, 21, 1

    Faouzi, J., & Janati, H. 2020, Journal of Machine Learning Research, 21, 1. http://jmlr.org/papers/v21/19-763.html

    2020

  8. [16]

    Filippenko, A. V. 1997, Annual Review of Astronomy and Astrophysics, 35, 309, 10.1146/annurev.astro.35.1.309

    1997 doi

  9. [17]

    J., Coughlin, M

    Fremling, C., Hall, X. J., Coughlin, M. W., et al. 2021, The Astrophysical Journal, 917, L2, 10.3847/2041-8213/ac116f

    2021 doi

  10. [18]

    2017, in Handbook of Supernovae, ed

    Gal-Yam , A. 2017, in Handbook of Supernovae, ed. A. W. Alsabti & P. Murdin , 195, 10.1007/978-3-319-21846-5_35

    2017 doi

  11. [19]

    Gehring, J., Auli, M., Grangier, D., Yarats, D., & Dauphin, Y. N. 2017, Convolutional Sequence to Sequence Learning , arXiv, 10.48550/arXiv.1705.03122

    2017 doi arXiv

  12. [20]

    2022, Transient Name Server AstroNote, 191, 1

    Goldwasser , S., Yaron , O., Sass , A., et al. 2022, Transient Name Server AstroNote, 191, 1

    2022

  13. [21]

    H., Pfahler, P., Pastorello, A., et al

    Harutyunyan, A. H., Pfahler, P., Pastorello, A., et al. 2008, Astronomy & Astrophysics, 488, 383, 10.1051/0004-6361:20078859

    2008 doi

  14. [22]

    2021, Remote Sensing, 13, 498

    He, X., Chen, Y., & Lin, Z. 2021, Remote Sensing, 13, 498

    2021

  15. [23]

    2007, The Astrophysical Journal, 669, L17

    Hicken, M., Garnavich, P., Prieto, J., et al. 2007, The Astrophysical Journal, 669, L17

    2007

  16. [24]

    A., Sullivan, M., Perrett, K., et al

    Howell, D. A., Sullivan, M., Perrett, K., et al. 2005, The Astrophysical Journal, 634, 1190, 10.1086/497119

    2005 doi

  17. [25]

    2012, Progress of Theoretical and Experimental Physics, 2012, 01A309

    Janka, H.-T., Hanke, F., H \"u depohl, L., et al. 2012, Progress of Theoretical and Experimental Physics, 2012, 01A309

    2012

  18. [26]

    2011, The Astrophysical Journal, 734, 102

    Kasen, D., Woosley, S., & Heger, A. 2011, The Astrophysical Journal, 734, 102

    2011

  19. [27]

    B., Modjaz, M., et al

    Khakpash, S., Bianco, F. B., Modjaz, M., et al. 2024, The Astrophysical Journal Supplement Series, 275, 37

    2024

  20. [28]

    2024, SpectraFM : Tuning into Stellar Foundation Models , arXiv, 10.48550/arXiv.2411.04750

    Koblischke, N., & Bovy, J. 2024, SpectraFM : Tuning into Stellar Foundation Models , arXiv, 10.48550/arXiv.2411.04750

    2024 doi arXiv

  21. [29]

    2020, Astrophysical Journal, 890, 177, 10.3847/1538-4357/ab6601

    Kou , S., Chen , X., & Liu , X. 2020, Astrophysical Journal, 890, 177, 10.3847/1538-4357/ab6601

    2020 doi

  22. [30]

    2020, The Astrophysical Journal, 890, 177

    Kou, S., Chen, X., & Liu, X. 2020, The Astrophysical Journal, 890, 177

    2020

  23. [31]

    1987, International Astronomical Union Circular, 4316, 1

    Kunkel, W., Madore, B., Shelton, I., et al. 1987, International Astronomical Union Circular, 4316, 1. https://ui.adsabs.harvard.edu/abs/1987IAUC.4316....1K

    1987

  24. [32]

    2015, SuperNova IDentification spectral templates of 70 stripped-envelope core-collapse supernovae, arXiv, 10.48550/arXiv.1405.1437

    Liu, Y., & Modjaz, M. 2015, SuperNova IDentification spectral templates of 70 stripped-envelope core-collapse supernovae, arXiv, 10.48550/arXiv.1405.1437

    2015 doi arXiv

  25. [33]

    B., & Graur, O

    Liu, Y.-Q., Modjaz, M., Bianco, F. B., & Graur, O. 2016, The Astrophysical Journal, 827, 90, 10.3847/0004-637X/827/2/90

    2016 doi

  26. [34]

    P., Challis, P., et al

    Matheson, T., Kirshner, R. P., Challis, P., et al. 2008, The Astronomical Journal, 135, 1598, 10.1088/0004-6256/135/4/1598

    2008 doi

  27. [35]

    A., Röpke, F

    Mazzali, P. A., Röpke, F. K., Benetti, S., & Hillebrandt, W. 2007, Science, 315, 825, 10.1126/science.1136259

    2007 doi

  28. [36]

    1941, Publications of the Astronomical Society of the Pacific, 53, 224, 10.1086/125315

    Minkowski, R. 1941, Publications of the Astronomical Society of the Pacific, 53, 224, 10.1086/125315

    1941 doi

  29. [37]

    P., & Arcavi , I

    Modjaz , M., Guti \'e rrez , C. P., & Arcavi , I. 2019, Nature Astronomy, 3, 717, 10.1038/s41550-019-0856-2

    2019 doi

  30. [38]

    Q., Bianco, F

    Modjaz, M., Liu, Y. Q., Bianco, F. B., & Graur, O. 2016, The Astrophysical Journal, 832, 108, 10.3847/0004-637X/832/2/108

    2016 doi

  31. [39]

    P., et al

    Modjaz, M., Blondin, S., Kirshner, R. P., et al. 2014, The Astronomical Journal, 147, 99, 10.1088/0004-6256/147/5/99

    2014 doi

  32. [40]

    2020, Monthly Notices of the RAS, 491, 4277, 10.1093/mnras/stz3312

    M \"o ller , A., & de Boissiére , T. 2020, Monthly Notices of the RAS, 491, 4277, 10.1093/mnras/stz3312

    2020 doi

  33. [41]

    2023, Positional Encodings for Light Curve Transformers : Playing with Positions and Attention , arXiv, 10.48550/arXiv.2308.06404

    Moreno-Cartagena, D., Cabrera-Vives, G., Protopapas, P., et al. 2023, Positional Encodings for Light Curve Transformers : Playing with Positions and Attention , arXiv, 10.48550/arXiv.2308.06404

    2023 doi arXiv

  34. [42]

    J., Sorokina, E

    Moriya, T. J., Sorokina, E. I., & Chevalier, R. A. 2018, Space Science Reviews, 214, 59

    2018

  35. [43]

    Muthukrishna, D., Parkinson, D., & Tucker, B. E. 2019, The Astrophysical Journal, 885, 85, 10.3847/1538-4357/ab48f4

    2019 doi

  36. [44]

    2024, Monthly Notices of the Royal Astronomical Society, 528, 5890, 10.1093/mnras/stae068

    Pan, J.-S., Ting, Y.-S., & Yu, J. 2024, Monthly Notices of the Royal Astronomical Society, 528, 5890, 10.1093/mnras/stae068

    2024 doi

  37. [45]

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

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

    1999 doi

  38. [46]

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

    Qu , H., Sako , M., M \"o ller , A., & Doux , C. 2021, Astronomical Journal, 162, 67, 10.3847/1538-3881/ac0824

    2021 doi

  39. [47]

    G., Filippenko, A

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

    1998 doi

  40. [48]

    2007, Thermonuclear Supernovae, Tech

    R \"o pke, F. 2007, Thermonuclear Supernovae, Tech. rep

    2007

  41. [49]

    de Souza, R., Thorp, S., Galbany, L., et al

    S. de Souza, R., Thorp, S., Galbany, L., et al. 2023, Astronomy and Computing, 44, 100715, 10.1016/j.ascom.2023.100715

    2023

  42. [50]

    Sasdelli , M., Ishida , E. E. O., Vilalta , R., et al. 2016, Monthly Notices of the RAS, 461, 2044, 10.1093/mnras/stw1228

    2016 doi

  43. [51]

    G., Gagliano , A., Malanchev , K., Narayan , G., & The LSST Dark Energy Science Collaboration

    Shah , V. G., Gagliano , A., Malanchev , K., Narayan , G., & The LSST Dark Energy Science Collaboration . 2025, arXiv e-prints, arXiv:2501.01496, 10.48550/arXiv.2501.01496

    2025 doi

  44. [52]

    A., Sollerman, J., et al

    Sharma, Y., Mahabal, A. A., Sollerman, J., et al. 2025, CCSNscore : A multi-input deep learning tool for classification of core-collapse supernovae using SED - Machine spectra, arXiv, 10.48550/arXiv.2412.08601

    2025 doi arXiv

  45. [53]

    M., Foley, R

    Silverman, J. M., Foley, R. J., Filippenko, A. V., et al. 2012, Monthly Notices of the Royal Astronomical Society, 425, 1789, 10.1111/j.1365-2966.2012.21270.x

    2012

  46. [54]

    2015, Very Deep Convolutional Networks for Large - Scale Image Recognition , arXiv, 10.48550/arXiv.1409.1556

    Simonyan, K., & Zisserman, A. 2015, Very Deep Convolutional Networks for Large - Scale Image Recognition , arXiv, 10.48550/arXiv.1409.1556

    2015 doi arXiv

  47. [55]

    2017, Attention Is All You Need , arXiv, 10.48550/arXiv.1706.03762

    Vaswani, A., Shazeer, N., Parmar, N., et al. 2017, Attention Is All You Need , arXiv, 10.48550/arXiv.1706.03762

    2017 doi arXiv

  48. [56]

    A., de Soto, K., & Gagliano, A

    Villar, V. A., de Soto, K., & Gagliano, A. 2023, arXiv preprint arXiv:2312.02266

    2023 arXiv

  49. [57]

    Williamson, M., Modjaz, M., & Bianco, F. B. 2019, The Astrophysical Journal Letters, 880, L22

    2019

  50. [58]

    2017, The Astrophysical Journal, 836, 244

    Woosley, S. 2017, The Astrophysical Journal, 836, 244

    2017

  51. [59]

    2005, Nature Physics, 1, 147

    Woosley, S., & Janka, T. 2005, Nature Physics, 1, 147

    2005

  52. [60]

    2011, The Astrophysical Journal, 734, 38

    Woosley, S., & Kasen, D. 2011, The Astrophysical Journal, 734, 38

    2011

  53. [61]

    E., & Weaver, T

    Woosley, S. E., & Weaver, T. A. 1986, IN: Annual review of astronomy and astrophysics. Volume 24 (A87-26730 10-90). Palo Alto, CA, Annual Reviews, Inc., 1986, p. 205-253., 24, 205

    1986

  54. [62]

    2019, in Information Processing in Medical Imaging , ed

    Yan, Y., Kawahara, J., & Hamarneh, G. 2019, in Information Processing in Medical Imaging , ed. A. C. S. Chung, J. C. Gee, P. A. Yushkevich, & S. Bao (Cham: Springer International Publishing), 793--804, 10.1007/978-3-030-20351-1_62

    2019 doi

  55. [63]

    2012, Publications of the Astronomical Society of the Pacific, 124, 668, 10.1086/666656

    Yaron, O., & Gal-Yam, A. 2012, Publications of the Astronomical Society of the Pacific, 124, 668, 10.1086/666656

    2012 doi

  56. [64]

    T., Mishra-Sharma , S., & Ashley Villar , V

    Zhang , G., Helfer , T., Gagliano , A. T., Mishra-Sharma , S., & Ashley Villar , V. 2024, Machine Learning: Science and Technology, 5, 045069, 10.1088/2632-2153/ad990d

    2024 doi

  57. [65]

    2021, Informer: Beyond efficient transformer for long sequence time-series forecasting

    Zhou, H., Zhang, S., Peng, J., et al. 2021, Informer: Beyond efficient transformer for long sequence time-series forecasting

    2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.