Pith. sign in

REVIEW 3 major objections 5 minor 57 references

Real-time Light Curve Classification Framework for the Wide Field Survey Telescope Using Modified Semi-supervised Variational Auto-Encoder

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper claims that adding an unsupervised reconstruction path to a supervised classifier improves early transient classification: on simulated WFST data it reaches 83.1% accuracy at three days after trigger, versus 78.7% for an RNN…

desk verdict Useful early-classification framework for WFST simulations, but the claimed 5.59% semi-supervised gain is not actually demonstrated and needs an ablation before I'd trust it. read the letter →

arxiv 2506.01216 v1 pith:7AVXELBM submitted 2025-06-01 astro-ph.IM

Yongling Tang , Lulu Fan , Zhen Wan , Yating Liu , Yan Lu This is my paper
classification astro-ph.IM
keywords LightcurveclassificationSemi-supervisedlearningVariationalauto-encoderEarlytransientWideFieldSurveyTelescopeGaussianprocessregressionPLAsTiCCRecurrentneuralnetwork
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 argues that a classifier given only the first three days of a transient's light curve can be made more accurate by simultaneously training it, through a shared variational encoder, to reconstruct the full 90-day light curve. On simulated Wide Field Survey Telescope data spanning nine transient classes, the framework reaches 83.1 percent accuracy at three days after trigger, compared with 78.7 percent for a GRU-based recurrent neural network benchmark, a 5.59 percent relative gain, with higher precision and recall in nearly every class. The authors' explanation is that the reconstruction task forces the early-phase representation to encode information about the later evolution of the event, which helps distinguish classes whose early light curves look similar. If the result carries over to real data, survey telescopes could flag interesting transients for follow-up within days of detection, when rapid spectroscopic and multi-wavelength observations are still possible. The framework also outputs a reconstructed full light curve and a compact latent representation, which the paper proposes as tools for downstream tasks such as anomaly detection.

What carries the argument

The load-bearing object is a semi-supervised variational auto-encoder with one shared encoder and two branches. The supervised branch maps the latent code to class probabilities through a softmax layer; the unsupervised branch decodes the latent code into the full 90-day light curve. The input is the three-day partial light curve in the u, g, and r bands, with later time steps masked to zero, and the reconstruction target is the full light curve. These branches are trained together with a total loss $\mathcal{L}_{\mathrm{tot}} = W_r \mathcal{L}_{\mathrm{rec}} + W_k \mathcal{L}_{\mathrm{KL}} + W_c \mathcal{L}_{\mathrm{cla}}$, where $\mathcal{L}_{\mathrm{rec}}$ is reconstruction mean squared error, $\mathcal{L}_{\mathrm{KL}}$ is the Kullback-Leibler divergence to a Gaussian prior, and $\mathcal{L}_{\mathrm{cla}}$ is categorical cross-entropy, with $W_c = 10$ so that the classification term dominates. The variational formulation keeps the latent space continuous and structured, so neighboring latent points decode to similar light curves.

What would settle it

Train both the framework and the RNN benchmark on the same simulated sample, then apply them to a set of spectroscopically confirmed transients from WFST's real alert stream, classifying each at three days after trigger with identical preprocessing. If the framework's accuracy does not exceed the benchmark's by a comparable margin, or falls below it, the central claim that the reconstruction path improves early classification in practice is falsified.

Watch

Extended reading notes

Core claim

On simulated WFST-like light curves drawn from PLAsTiCC models via SNANA, the paper finds that coupling a supervised classification branch with an unsupervised variational auto-encoder that reconstructs the full light curve from the partial early light curve improves early classification. Across 10-fold cross-validation, the framework reaches 83.1 percent accuracy at three days after trigger, compared with 78.7 percent for a GRU-based RNN benchmark, and raises macro precision and recall from 0.836 to 0.858. The authors attribute the gain to the reconstruction task: the shared encoder must encode enough information about the full evolution to reconstruct it, so the early-phase representation carries more class-discriminative content than a purely supervised RNN would learn. They also show that the latent space clusters by class and that reconstruction of the full curve is possible, with an average reconstruction loss of 19.56 percent.

Load-bearing premise

The load-bearing premise is that the simulated light curves—generated from PLAsTiCC models with the SNANA package, smoothed by Gaussian processes, and split by redshift—are representative enough of real WFST observations that the measured 5.59 percent accuracy gain will transfer to the alert stream.

Editorial extensions

If this is right

  • At three days after trigger, the framework separates nine transient classes at 83.1 percent accuracy on simulated data, making early follow-up of fast-evolving events such as kilonovae and tidal disruption events feasible.
  • Because the encoder is shared, every classified alert also carries a reconstructed full light curve and a latent code; the paper identifies anomaly detection and observation-phase determination as downstream uses.
  • The framework can be updated with newly collected unlabeled data without discarding previously learned structure, which the paper argues matters for a new survey whose labeled sample is small.
  • Precision and recall improve in almost all subclasses relative to the GRU benchmark, with the largest absolute gains in classes such as AGN and TDE.
  • The performance advantage over the benchmark appears from the early phase onward and remains consistent as more data accumulate, according to the paper's phase-resolved comparison.

Reading between the lines

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

  • A natural test the paper does not run is to vary the amount of unlabeled data and the label fraction; the hypothesis that reconstruction helps most when labels are scarce is directly measurable and would tell observers how much spectroscopic follow-up is needed.
  • Because the reconstruction target is the full light curve, the same architecture can be read as a forecasting device: at three days it predicts the upcoming 87 days, so its output could seed trigger decisions for time-critical follow-up even before the classifier's probability is used.
  • The paper's own latent-space plot of the unseen CART class suggests a practical anomaly detector: alerts that land far from known clusters, or that reconstruct poorly, could be prioritized for human review even if the classifier assigns them a known label.
  • The relative comparison between the two models on identical simulated input is probably a safer quantity to carry into the real survey than the absolute 83.1 percent accuracy, since simulation bias affects both models similarly; the paper makes this point only implicitly.
Share X Bluesky LinkedIn Reddit HN

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 presents a semi-supervised variational auto-encoder (VAE) framework for early classification of simulated Wide Field Survey Telescope (WFST) light curves. The framework combines a supervised GRU-based classification path with an unsupervised decoder path that reconstructs a full 90-day light curve from a partial input covering only the first three days after trigger. Evaluated on SNANA-generated, PLAsTiCC-based simulated light curves, the authors report an accuracy of 0.831 versus 0.787 for a GRU benchmark model, with improvements in precision and recall for most subclasses. The paper also presents latent-space visualizations and light-curve reconstructions, and states that the code and model weights are publicly available. The core claim is that the unsupervised reconstruction path and the use of unlabeled data drive the observed accuracy improvement.

Significance. If the reported improvement is real and can be causally attributed to the proposed semi-supervised reconstruction mechanism, the framework would be a practically useful contribution to early transient classification, where rapid follow-up decisions matter. The paper addresses a relevant problem, uses a realistic simulation pipeline (SNANA with PLAsTiCC models and WFST cadence), makes code and model weights available, and explicitly acknowledges the limitations of simulated data for real-world transfer. However, the central causal claim—that the unsupervised path is responsible for the accuracy gain—is not established by the experiments as presented, because the comparison against the benchmark is confounded by multiple simultaneous differences and no ablation studies are reported. The significance of the work therefore hinges on additional controlled experiments that isolate the contribution of each component.

major comments (3)
  1. [Section 4.4, Table 4, Eq. (6)] The headline improvement (accuracy 0.831 vs 0.787) is attributed to the unsupervised auto-encoder path, but the comparison does not isolate this mechanism. The full framework differs from the GRU benchmark in at least three ways: the decoder branch and reconstruction loss L_rec (Eq. 3), the KL loss L_KL (Eq. 4), and the inclusion of an additional unlabeled training dataset. No experiments vary the loss weights W_r, W_k, W_c in Eq. (6) or the amount/composition of the unlabeled data, and no parameter counts or training budgets are reported for the two models. As a result, the measured gain could be due to added model capacity, multi-task regularization from the reconstruction branch, or simply seeing more training examples, rather than the claimed semi-supervised representation learning. Please add ablations: (a) the supervised path trained alone (no decoder or KL terms), (b) the full framework trained with labeled data only, and (c) the full framework trained with varying amounts of unlabeled data, to demonstrate the specific contribution of each component.
  2. [Section 4, first paragraph] The evaluation protocol is described as "10-fold cross-validation (each fold contains approximately 2000 light curves, randomly selected from the test dataset without specific division)". This is not the standard use of cross-validation, which is typically performed on the training set for model selection, with the test set held out for final evaluation. As written, it is unclear whether the reported accuracy and uncertainties come from training on the test set, from repeated random subsampling of the test set, or from some other procedure. This ambiguity directly affects the validity of the reported accuracy uncertainty (0.012) and the fairness of the comparison with the benchmark model. Please clarify the exact protocol: how many models were trained, how the folds were created, whether the same folds were used for both models, and whether any training data were used in the cross-validation procedure.
  3. [Section 3.3.2, Eq. (5)] The framework is described as semi-supervised, and Section 4.4 implies that the use of unlabeled data contributes to the accuracy gain, but the loss function L_cla (Eq. 5) is defined only for labeled samples, and the text does not specify how unlabeled samples enter the training objective. If unlabeled samples contribute only through L_rec and L_KL, then the claimed benefit of unlabeled data is not independent of the reconstruction regularization, and the term "semi-supervised" is potentially misleading. Please state explicitly whether unlabeled samples enter L_cla (e.g., via pseudo-labels or a weighted cross-entropy term) and, if they do not, revise the claims about the role of unlabeled data accordingly.
minor comments (5)
  1. [Section 3.3.2] In the paragraph after Eq. (6), the text says the overall loss is "combined as outlined in Equation 5", but the total loss is defined in Eq. (6); the equation number should be corrected.
  2. [Section 4.2] The passage "we exclude the CART (Calcium-rich gap transients) class in the training dataset" is confusing because CART is not listed among the nine classes in Table 1 or in the dataset description in Section 2.2; please clarify whether CART was part of the simulated data and, if so, why it is excluded, or remove this passage if it is an error.
  3. [Tables 2 and 3] The reported uncertainties are given to three significant figures (e.g., 4.8E-04), which seems overly precise given that the standard deviation across folds is not described; consider reporting fewer significant digits or explaining explicitly how these uncertainties were computed from the evaluation procedure.
  4. [Section 3.3.1] The phrase "masking the values after a certain time step with zeros" is ambiguous: are zeros used as input features alongside a masking layer in the GRU, or is the zero-padding itself the only indicator? The architecture in Figure 5 includes a masking operation, but the text should specify that zeros are used with a mask to distinguish unobserved time steps.
  5. [Abstract / Code availability] The abstract and GitHub mention state that code and model weights are publicly available, but no repository URL or data availability statement is provided in the manuscript; adding the URL would help reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the headline accuracy gain is an empirical comparison on a held-out simulated test set, and the reconstruction path is a trained objective, not an independent prediction; the missing ablation is an attribution concern, not circularity.

full rationale

The paper's central claim is an empirical model comparison: the full semi-supervised framework reaches accuracy 0.831 versus 0.787 for a benchmark RNN on the same simulated test set, evaluated by 10-fold cross-validation. This is not a derivation from assumptions, and the test set is held out from training, so the comparison is not circular by construction. The reconstruction path is trained with the explicit mean-square-error objective in Eq. 3 to map a partial light curve to the full light curve; reporting a test reconstruction loss of 19.56% is a standard evaluation of that same objective, not a claim that the reconstruction is an independent, untrained prediction. The paper does not invoke a uniqueness theorem, does not define any central quantity in terms of another claimed result, and does not rely on a load-bearing self-citation chain; the only co-authored citation (Cai et al. 2025, which includes author L. Fan) concerns the WFST alert stream and is not used to justify the classification improvement. The absence of an ablation isolating the unsupervised path means the causal attribution in Section 4.4 is underdetermined, but that is a correctness/interpretability limitation, not circularity, because the headline gain is still a direct measurement on a separate test set rather than a quantity that reduces to a fitted input or to a self-citation. The paper's own limitation note that simulated-data performance may be overly optimistic concerns external validity, not circular derivation. Therefore no circular step can be exhibited, and the score is 0.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The framework's central claims rest on the fidelity of the SNANA/PLAsTiCC simulation to real WFST observations, the unbiasedness of the GP preprocessing, and the (untested) assumption that the unsupervised path leverages unlabeled data. The free parameters are largely ML hyperparameters; no physical constants are fitted.

free parameters (8)
  • Loss weights Wr, Wk, Wc = Wr=1, Wk=1, Wc=10
    Hand-chosen to balance reconstruction, KL, and classification loss scales (Section 3.3.2).
  • Network architecture hyperparameters (layers, units) = Not given numerically in text; topology shown in Figures 6-7
    Selected by simple grid search based on values from similar models (Section 3.3.2).
  • Latent dimensionality = 2 (implied by latent space plots)
    The 2D latent plots imply a 2-dimensional latent space, but the paper never states the latent dimension explicitly.
  • GP kernel hyperparameters = Matern-3/2 length scales, fitted per light curve
    GP regression implemented with george and scikit-learn; length scales fitted to each light curve during interpolation (Section 3.1).
  • Early classification window = 3 days after trigger (phase j=6)
    Design choice defining the 'early' input; results depend on this window.
  • Global normalization constant F = Maximum flux across all training data
    Used in Equations 1-2 to preserve absolute brightness while normalizing; data-derived.
  • Trigger threshold = 5-sigma in imaging difference
    Definition of trigger from WFST pipeline (Hu et al. 2022a); affects what data fall in the early window.
  • GP augmentation wavelength shifts = Described as low/high 'z' scenarios; exact shift values not given
    Augmentation by modifying the wavelength parameter of the GP model (Figure 3), a free choice.
assumptions (6)
  • domain assumption PLAsTiCC simulation models faithfully represent the nine transient classes and their light-curve properties.
    Section 2.2: all training and test data come from SNANA simulations using PLAsTiCC models; no real data are used.
  • domain assumption SNANA with custom SIMLIB and survey definition reproduces WFST cadence, filters, and depth.
    Section 2.2; the WFST-like simulation is created in-house and not validated against real WFST data.
  • domain assumption GP interpolation with a Matern-3/2 kernel provides unbiased 90-day light curves from sparse early data.
    Section 3.1 and Figure 4; authors note GP performs poorly at forecasting future points but assume interpolation is consistent.
  • ad hoc to paper The reconstruction target, the GP-interpolated full light curve, is the correct learning target for the unsupervised path.
    Section 3.3.1-3.3.2; the model reconstructs the GP-interpolated and normalized full curve, so GP artifacts are part of the target.
  • ad hoc to paper Zero-padding unobserved future time steps, without feeding the phase index to the network, is sufficient for the model to know how much of the curve is observed.
    Section 3.3.1; the network sees masked inputs with zeros after phase j, and the model must infer the observation phase implicitly.
  • domain assumption Train and test sets with identical class distributions, split only by redshift, are adequate for evaluating generalization.
    Section 2.2 and 4.4; the authors explicitly state this deviates from real-world settings.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Real-time Light Curve Classification Framework for the Wide Field Survey Telescope Using Modified Semi-supervised Variational Auto-Encoder." pith.science (2026). https://pith.science/paper/7AVXELBM

@misc{pith2026250601216,
  author       = {Pith},
  title        = {Pith review of: Real-time Light Curve Classification Framework for the Wide Field Survey Telescope Using Modified Semi-supervised Variational Auto-Encoder},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7AVXELBM}},
  note         = {Machine review of arXiv:2506.01216}
}
read the original abstract

Modern time-domain astronomy will benefit from the vast data collected by survey telescopes. The 2.5 m Wide Field Survey Telescope (WFST), with its powerful capabilities, is promising to make significant contributions in the era of large sky surveys. To harness the full potential of the enormous amount of unlabeled light curve data that the WFST will collect, we have developed a semisupervised light curve classification framework. This framework showcases several unique features. First, it is optimized for classifying events based on the early phase of the light curve (three days after trigger), which can help identify interesting events early and enable efficient follow-up observations. Second, the semisupervised nature of our framework allows it to leverage valuable information from large volumes of unlabeled data, potentially bridging the gap between simulations and real observations and achieving better generalization in practical scenarios. Compared to the commonly used Recurrent Neural Network models, our framework has shown a 5.59% improvement in accuracy for early classification tasks, as well as improvements in precision and recall in almost all subclasses. Moreover, our approach provides a reconstructed light curve, along with a compact latent representation, offering a different perspective that can be used for further downstream tasks beyond classification. The code and model weights used in this work are maintained and publicly available on our GitHub repository.

Figures

Figures reproduced from arXiv: 2506.01216 by the authors.

Figure 1
Figure 1. Redshift distributions of three datasets representing the following transient classes: KN, SLSN, SNIa, SNIa-91bg, SNIax, SNIbc, SNII and TDE. We generated three datasets in total: a labeled training dataset, a test dataset, and an additional unlabeled training dataset. The ratio of these three datasets is 1:1:2 after preprocessing, as described in Section 3.3.1. We deviate from the commonly used ratio of training to… view at source ↗
Figure 2
Figure 2. Illustrations of well sampled simulated light curves in expected WFST cadence. From left to right, the light curves represent the following classes: AGN, KN and SNIa (the first line); SNIa-91bg, SNIax and SNIbc (the second line); SNII, SLSN and TDE (the third line). days before the trigger), which we refer to as a “full light curve”. The definition of a “trigger” for an alert event in the light curve is any observat… view at source ↗
Figure 3
Figure 3. Illustration of GP Augmentation Applied to Simulated Light Curves. Each band displays the mean Gaussian Process (GP) flux prediction as a solid line, with a shaded contour representing one-standard￾deviation uncertainty in the flux prediction. The middle column presents the initial fit. The left column shows the augmented data for a ”low redshift” scenario (achieved by reducing the wavelength parameter in the GP mod… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Example of GP real-time interpolation applied to simulated light curves. A solid vertical line divides the light curve into segments for early classification (left) and full classification (right). Each band’s mean GP flux prediction is shown as a solid line, surrounde…
Figure 5
Figure 5. Figure 5: The semi-supervised classification architecture comprises two distinct pathways. The unsuper￾vised component (denoted as the orange pathway) is purposed to discern the low-dimensional distribution inherent in the data. It functions by reconstructing the original full l…
Figure 6
Figure 6. Figure 6: The detailed structure of the benchmark model (left panel) and the encoder module of our framework (right panel). Lcla = − 1 N X N i=1 X K k=1 y i k log(ˆy i k ) (5) Ltot = Wr × Lrec + Wk × LKL + Wc × Lcla (6) In these loss functions, Fˆ ij (t) represents the reconstru…
Figure 7
Figure 7. Figure 7: Decoder module of our framework were selected for their effectiveness in balancing adequate model capacity with overfitting prevention, meeting the demands of our current demonstration framework. Future tests with real-world data are expected to further refine these ch…
Figure 8
Figure 8. Figure 8: Comparison of confusion matrices (normalized over true labels) between the benchmark model (right panel) and our framework (left panel). We further examine the Receiver Operating Characteristics (ROC) curve of our framework. Com￾pared to the confusion matrix, the ROC c…
Figure 9
Figure 9. Figure 9: Comparison of ROC curve and corresponding AUC between the benchmark model (right panel) and our framework (left panel) [PITH_FULL_IMAGE:figures/full_fig_p022_9.png]
Figure 10
Figure 10. Figure 10: The macro precision, recall, and F1 scores of our framework are compared with those of the benchmark model across different phases of the light curve. space. Different classes display varying degrees of dispersion in the latent space, which may indicate the inherent v…
Figure 11
Figure 11. Figure 11: The latent space representation of the early light curve in the test dataset. The two axes are the latent features learned by the unsupervised path. Finally, we exclude the CART (Calcium-rich gap transients) class in the training dataset (as they are relatively rare, …
Figure 12
Figure 12. Figure 12: Latent space distribution. The left panel is the different distribution of the Cart and test data set; The right panel is the different distribution of the train dataset and the test dataset. that the model can effectively reconstruct the original light curve given th…
Figure 13
Figure 13. Figure 13: The latent space analysis reveals distinct patterns in model performance: the left panel highlights misclassified data points within the context of the full test set, while the right panel visualizes reconstructed light curves sampled from diverse regions of the laten…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

57 extracted references · 14 canonical work pages

  1. [1]

    2024, PASP, 136, 124504, doi: 10.1088/1538-3873/ad9a7d

    Abdelhadi, B., & Rubin, D. 2024, PASP, 136, 124504, doi: 10.1088/1538-3873/ad9a7d

    work page doi:10.1088/1538-3873/ad9a7d 2024

  2. [2]

    2019, MNRAS, 482, 5078, doi: 10.1093/mnras/sty2836

    Aguirre, C., Pichara, K., & Becker, I. 2019, MNRAS, 482, 5078, doi: 10.1093/mnras/sty2836

    work page doi:10.1093/mnras/sty2836 2019

  3. [3]

    Allam, T., & McEwen, J. D. 2024, RAS Techniques and Instruments, 3, 209, doi: 10.1093/rasti/rzad046

    doi:10.1093/rasti/rzad046 2024

  4. [4]

    Allam, Jr., T., Peloton, J., & McEwen, J. D. 2023, arXiv e-prints, arXiv:2303.08951, doi: 10.48550/arXiv.2303.08951

    work page Pith review arXiv doi:10.48550/arxiv.2303.08951 2023

  5. [5]

    W., & O’Neil, M

    Greengard, L., Hogg, D. W., & O’Neil, M. 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence, 38, 252, doi: 10.1109/TPAMI.2015.2448083

    arXiv 2015

  6. [6]

    2019, arXiv e-prints, arXiv:1904.07248, doi: 10.48550/arXiv.1904.07248

    Baron, D. 2019, arXiv e-prints, arXiv:1904.07248, doi: 10.48550/arXiv.1904.07248

    arXiv doi:10.48550/arxiv.1904.07248 2019

  7. [7]

    2020, MNRAS, 493, 2981, doi: 10.1093/mnras/staa350

    Becker, I., Pichara, K., Catelan, M., et al. 2020, MNRAS, 493, 2981, doi: 10.1093/mnras/staa350

    doi:10.1093/mnras/staa350 2020

  8. [8]

    C., Kulkarni, S

    Bellm, E. C., Kulkarni, S. R., Graham, M. J., et al. 2019, PASP, 131, 018002, doi: 10.1088/1538-3873/aaecbe

    doi:10.1088/1538-3873/aaecbe 2019

Show all 57 references
  1. [9]

    A., Abe, F., Dodd, R

    Bond, I. A., Abe, F., Dodd, R. J., et al. 2001, MNRAS, 327, 868, doi: 10.1046/j.1365-8711.2001.04776.x

    2001

  2. [10]

    2019, AJ, 158, 257, doi: 10.3847/1538-3881/ab5182 —

    Boone, K. 2019, AJ, 158, 257, doi: 10.3847/1538-3881/ab5182 —. 2021, AJ, 162, 275, doi: 10.3847/1538-3881/ac2a2d 29

    2019 doi

  3. [11]

    F., & Maund, J

    Burhanudin, U. F., & Maund, J. R. 2023, MNRAS, 521, 1601, doi: 10.1093/mnras/stac3672

    2023 doi

  4. [12]

    2025, arXiv e-prints, arXiv:2501.15018, doi: 10.48550/arXiv.2501.15018

    Cai, M., Xu, Z., Fan, L., et al. 2025, arXiv e-prints, arXiv:2501.15018, doi: 10.48550/arXiv.2501.15018

    2025 doi

  5. [13]

    Chaini, S., & Kumar, S. S. 2020, arXiv e-prints, arXiv:2006.12333, doi: 10.48550/arXiv.2006.12333

    2020 doi arXiv

  6. [14]

    2014, arXiv e-prints, arXiv:1409.1259, doi: 10.48550/arXiv.1409.1259

    Bengio, Y. 2014, arXiv e-prints, arXiv:1409.1259, doi: 10.48550/arXiv.1409.1259

    2014 doi arXiv

  7. [15]

    2015, Keras, https://keras.io Dark Energy Survey Collaboration, Abbott, T.,

    Chollet, F., et al. 2015, Keras, https://keras.io Dark Energy Survey Collaboration, Abbott, T.,

    2015

  8. [16]

    B., et al

    Abdalla, F. B., et al. 2016, MNRAS, 460, 1270, doi: 10.1093/mnras/stw641

    2016 doi

  9. [17]

    2023, A&A, 670, A54, doi: 10.1051/0004-6361/202243928

    Donoso-Oliva, C., Becker, I., Protopapas, P., et al. 2023, A&A, 670, A54, doi: 10.1051/0004-6361/202243928

    2023 doi

  10. [18]

    J., Djorgovski, S

    Drake, A. J., Djorgovski, S. G., Mahabal, A., et al. 2012, in IAU Symposium, Vol. 285, New Horizons in Time Domain Astronomy, ed. E. Griffin, R. Hanisch, & R. Seaman, 306–308, doi: 10.1017/S1743921312000889

    2012 doi

  11. [19]

    2017, A&A, 603, A60, doi: 10.1051/0004-6361/201630240

    Frontera-Pons, J., Sureau, F., Bobin, J., & Le Floc’h, E. 2017, A&A, 603, A60, doi: 10.1051/0004-6361/201630240

    2017 doi

  12. [20]

    I., & Aleo, P

    Malz, A. I., & Aleo, P. D. 2023, ApJ, 954, 6, doi: 10.3847/1538-4357/ace326 Hloˇ zek, R., Malz, A. I., Ponder, K. A., et al. 2023, ApJS, 267, 25, doi: 10.3847/1538-4365/accd6a

    2023 doi

  13. [21]

    1997, Neural computation, 9, 1735

    Hochreiter, S., & Schmidhuber, J. 1997, Neural computation, 9, 1735

    1997

  14. [22]

    2020, MNRAS, 493, 6050, doi: 10.1093/mnras/staa642

    Hosenie, Z., Lyon, R., Stappers, B., Mootoovaloo, A., & McBride, V. 2020, MNRAS, 493, 6050, doi: 10.1093/mnras/staa642

    2020 doi

  15. [23]

    2022a, ApJ, 936, 157, doi: 10.3847/1538-4357/ac7394

    Hu, L., Wang, L., Chen, X., & Yang, J. 2022a, ApJ, 936, 157, doi: 10.3847/1538-4357/ac7394

    doi

  16. [24]

    2022b, Universe, 9, 7, doi: 10.3390/universe9010007 Ivezi´ c,ˇZ., Kahn, S

    Hu, M., Hu, L., Jiang, J.-a., et al. 2022b, Universe, 9, 7, doi: 10.3390/universe9010007 Ivezi´ c,ˇZ., Kahn, S. M., Tyson, J. A., et al. 2019, ApJ, 873, 111, doi: 10.3847/1538-4357/ab042c

    2019 doi

  17. [25]

    Jamal, S., & Bloom, J. S. 2020, ApJS, 250, 30, doi: 10.3847/1538-4365/aba8ff

    2020 doi

  18. [26]

    2012, in AAS/Division for Planetary Sciences Meeting

    Jedicke, R., Tonry, J., Veres, P., et al. 2012, in AAS/Division for Planetary Sciences Meeting

    2012

  19. [27]

    2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference

    Kaiser, N., Burgett, W., Chambers, K., et al. 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference

    2010

  20. [28]

    7733, Ground-based and Airborne Telescopes III, ed

    Series, Vol. 7733, Ground-based and Airborne Telescopes III, ed. L. M. Stepp, R. Gilmozzi, & H. J. Hall, 77330E, doi: 10.1117/12.859188

    doi

  21. [29]

    P., Cinabro, D., et al

    Kessler, R., Bernstein, J. P., Cinabro, D., et al. 2009, PASP, 121, 1028, doi: 10.1086/605984

    2009 doi

  22. [30]

    2019, PASP, 131, 094501, doi: 10.1088/1538-3873/ab26f1

    Kessler, R., Narayan, G., Avelino, A., et al. 2019, PASP, 131, 094501, doi: 10.1088/1538-3873/ab26f1

    2019 doi

  23. [31]

    2023, Research in Astronomy and Astrophysics, 23, 035013, doi: 10.1088/1674-4527/acb877 30

    Lei, L., Zhu, Q.-F., Kong, X., et al. 2023, Research in Astronomy and Astrophysics, 23, 035013, doi: 10.1088/1674-4527/acb877 30

    2023 doi

  24. [32]

    2022, MNRAS, 513, 2422, doi: 10.1093/mnras/stac946

    Lin, Z., Jiang, N., & Kong, X. 2022, MNRAS, 513, 2422, doi: 10.1093/mnras/stac946

    2022 doi

  25. [33]

    2023, ApJ, 947, 59, doi: 10.3847/1538-4357/acc73b

    Liu, Z.-Y., Lin, Z.-Y., Yu, J.-M., et al. 2023, ApJ, 947, 59, doi: 10.3847/1538-4357/acc73b

    2023 doi

  26. [34]

    2018, arXiv e-prints, arXiv:1806.00398, doi: 10.48550/arXiv.1806.00398

    Ma, Z., Zhu, J., Li, W., & Xu, H. 2018, arXiv e-prints, arXiv:1806.00398, doi: 10.48550/arXiv.1806.00398

    2018 doi arXiv

  27. [35]

    2022, Libraries & Recommended Citations for using PLAsTiCC Models, v4, Zenodo, doi: 10.5281/zenodo.6672739 M¨ oller, A., & de Boissi` ere, T

    Modelers, P. 2022, Libraries & Recommended Citations for using PLAsTiCC Models, v4, Zenodo, doi: 10.5281/zenodo.6672739 M¨ oller, A., & de Boissi` ere, T. 2020, MNRAS, 491, 4277, doi: 10.1093/mnras/stz3312

    2022 doi

  28. [36]

    2022, MNRAS, 517, 393, doi: 10.1093/mnras/stac2582

    Webb, S., & Narayan, G. 2022, MNRAS, 517, 393, doi: 10.1093/mnras/stac2582

    2022 doi

  29. [37]

    2019, PASP, 131, 118002, doi: 10.1088/1538-3873/ab1609

    Biswas, R., & Hloˇ zek, R. 2019, PASP, 131, 118002, doi: 10.1088/1538-3873/ab1609

    2019 doi

  30. [38]

    2018, in American Astronomical Society Meeting Abstracts, Vol

    Narayan, G., Zaidi, T., Soraisam, M., & ANTARES Collaboration. 2018, in American Astronomical Society Meeting Abstracts, Vol. 231, American Astronomical Society Meeting Abstracts #231, 332.08

    2018

  31. [39]

    S., P´ erez, F., & van der Walt, S

    Naul, B., Bloom, J. S., P´ erez, F., & van der Walt, S. 2018, Nature Astronomy, 2, 151, doi: 10.1038/s41550-017-0321-z

    2018 doi

  32. [40]

    2019, A&A, 627, A21, doi: 10.1051/0004-6361/201834473

    Fouchez, D. 2019, A&A, 627, A21, doi: 10.1051/0004-6361/201834473

    2019 doi

  33. [41]

    2011, Journal of Machine Learning Research, 12, 2825, doi: 10.48550/arXiv.1201.0490 Pimentel, ´O., Est´ evez, P

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825, doi: 10.48550/arXiv.1201.0490 Pimentel, ´O., Est´ evez, P. A., & F¨ orster, F. 2023, AJ, 165, 18, doi: 10.3847/1538-3881/ac9ab4

    2011 doi arXiv

  34. [42]

    Connolly, A. J. 2020, AJ, 160, 45, doi: 10.3847/1538-3881/ab9644

    2020 doi

  35. [43]

    2018, MNRAS, 481, 894, doi: 10.1093/mnras/sty2309

    Pursiainen, M., Childress, M., Smith, M., et al. 2018, MNRAS, 481, 894, doi: 10.1093/mnras/sty2309

    2018 doi

  36. [44]

    2022, AJ, 163, 57, doi: 10.3847/1538-3881/ac39a1

    Qu, H., & Sako, M. 2022, AJ, 163, 57, doi: 10.3847/1538-3881/ac39a1

    2022 doi

  37. [45]

    2019, AJ, 157, 16, doi: 10.3847/1538-3881/aaf101

    Reis, I., Baron, D., & Shahaf, S. 2019, AJ, 157, 16, doi: 10.3847/1538-3881/aaf101

    2019 doi

  38. [46]

    A., Trotta, R., & van Dyk, D

    Revsbech, E. A., Trotta, R., & van Dyk, D. A. 2018, MNRAS, 473, 3969, doi: 10.1093/mnras/stx2570

    2018 doi

  39. [47]

    W., Homrighausen, D., Freeman, P

    Richards, J. W., Homrighausen, D., Freeman, P. E., Schafer, C. M., & Poznanski, D. 2012, MNRAS, 419, 1121, doi: 10.1111/j.1365-2966.2011.19768.x

    2012

  40. [48]

    G., Filippenko, A

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

    1998 doi

  41. [49]

    L., Aleo, P

    Russeil, E., Malanchev, K. L., Aleo, P. D., et al. 2024, A&A, 683, A251, doi: 10.1051/0004-6361/202348158

    2024 doi

  42. [50]

    2008, AJ, 135, 348, doi: 10.1088/0004-6256/135/1/348 S´ anchez-S´ aez, P., Reyes, I., Valenzuela, C., et al

    Sako, M., Bassett, B., Becker, A., et al. 2008, AJ, 135, 348, doi: 10.1088/0004-6256/135/1/348 S´ anchez-S´ aez, P., Reyes, I., Valenzuela, C., et al. 2021, AJ, 161, 141, doi: 10.3847/1538-3881/abd5c1

    2008 doi

  43. [51]

    2024, ApJ, 976, 155, doi: 10.3847/1538-4357/ad86bc 31

    Su, Z.-B., Cai, Z.-Y., Wang, J.-X., et al. 2024, ApJ, 976, 155, doi: 10.3847/1538-4357/ad86bc 31

    2024 doi

  44. [52]

    B., & Crotts, A

    Tomaney, A. B., & Crotts, A. P. S. 1996, AJ, 112, 2872, doi: 10.1086/118228

    1996 doi

  45. [53]

    A., Cranmer, M., Contardo, G., Ho, S., & Yao-Yu Lin, J

    Villar, V. A., Cranmer, M., Contardo, G., Ho, S., & Yao-Yu Lin, J. 2020a, arXiv e-prints, arXiv:2010.11194, doi: 10.48550/arXiv.2010.11194

    2010 doi arXiv

  46. [54]

    A., Nicholl, M., & Berger, E

    Villar, V. A., Nicholl, M., & Berger, E. 2019, in American Astronomical Society Meeting

    2019

  47. [55]

    A., Hosseinzadeh, G., Berger, E., et al

    Villar, V. A., Hosseinzadeh, G., Berger, E., et al. 2020b, ApJ, 905, 94, doi: 10.3847/1538-4357/abc6fd

    doi

  48. [56]

    2023, Science China Physics, Mechanics, and Astronomy, 66, 109512, doi: 10.1007/s11433-023-2197-5

    Wang, T., Liu, G., Cai, Z., et al. 2023, Science China Physics, Mechanics, and Astronomy, 66, 109512, doi: 10.1007/s11433-023-2197-5

    2023 doi

  49. [57]

    2021, Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 11, e1425

    Yu, C., Li, K., Zhang, Y., et al. 2021, Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 11, e1425

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.