Pith. sign in

REVIEW 4 major objections 5 minor 42 references

The classification of real and bogus transients using active learning and semi-supervised learning

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A real/bogus transient classifier built from only 1000 labels can match fully supervised models trained on 2500.

desk verdict Solid, honest AL+SSL real/bogus paper whose accuracy claims are limited by knowingly non-representative labels; deserves review but not the deployment framing. read the letter →

arxiv 2412.02409 v2 pith:7F7CCVSD submitted 2024-12-03 astro-ph.IM

Yating Liu , Lulu Fan , Lei Hu , Junqiang Lu , Yan Lu , Zelin Xu , Jiazheng Zhu , Haochen Wang
show 1 more author
Xu Kong
This is my paper · ORCID
classification astro-ph.IM
keywords real-bogusclassificationactivelearningsemi-supervisedpseudo-labelingZwickyTransientFacilitytime-domainsurveysconvolutionalneuralnetworksWFST
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 aims to show that a transient real/bogus classifier competitive with fully supervised models can be trained using only 1000 human-annotated examples. The proposed method, RB-C1000, first trains a network on 900 labels, asks experts to label the 100 most ambiguous remaining samples, then repeatedly pseudo-labels the most confident unlabeled samples and retrains. On three newly compiled ZTF datasets, RB-C1000 reaches average accuracies of 98.8%, 98.8%, and 98.6%, and it matches or outperforms a fully supervised ResNet-18 trained on 2500 random labels on the hardest dataset. The result matters because large upcoming surveys such as WFST and LSST need real-time filtering of transient candidates at a time when labeled data are scarce. The authors caution that their datasets contain only supernovae as real sources and random detections as bogus, so the reported numbers may not transfer directly to a real alert stream.

What carries the argument

The central mechanism is the three-stage RB-C1000 pipeline built on a ResNet-18 backbone taking three input channels (science, reference, and difference image stamps of 63x63 pixels). The initial training stage (ITS) trains on M=900 labeled samples and ranks unlabeled data by prediction confidence to identify the K=100 hardest samples. The active learning stage (ALS) retrains on the labeled set of 1000 samples and selects V pseudo-labeled samples whose confidence exceeds a threshold of τ=0.95. The semi-supervised learning stage (SSLS) retrains from scratch on labeled plus pseudo-labeled data and repeats R=3 times. This combination lets the model exploit both expert-selected hard examples and abundant unlabeled data without extra human annotation.

What would settle it

Take a fresh sample of alert streams from a running survey, label every candidate by expert visual inspection (including non-supernova real sources such as AGN variability and moving objects), and run the RB-C1000 model trained on the paper's recipe; if accuracy on this expert-labeled test set falls well below 98% or the false-negative rate on non-supernova real sources is much higher than the reported ~1.8%, the claimed generalization does not hold.

Watch

Extended reading notes

Core claim

The paper's central claim is that a real/bogus classifier built with only 1000 labeled samples can perform as well as a fully supervised model trained on thousands of random labels. In the RB-C1000 pipeline the initial supervised model selects the K=100 hardest unlabeled samples for expert annotation; the active model then picks V high-confidence samples and assigns them pseudo-labels; and the semi-supervised model retrains on the expanded set for R=3 iterations. On the ZTF-NEWg, ZTF-NEWr, and ZTF-NEWm datasets this yields average accuracies of 98.8%, 98.8%, and 98.6%, with the mixed-band ZTF-NEWm result (98.6%) exceeding the fully supervised ResNet-18 baseline trained on 2500 random labels (98.4%). The authors frame this as a practical route to deploying deep-learning real/bogus classification at the start of a survey, when annotation resources are limited.

Load-bearing premise

The results depend on the labels being trustworthy: real sources are only supernovae reported to TNS, and the bogus set is a random sample of detection candidates that the authors say may contain a few real detections, so label contamination would inflate the measured accuracy.

Editorial extensions

If this is right

  • Surveys like WFST and LSST can deploy a deep-learning real/bogus filter early in operations with only about 1000 human labels, reducing annotation cost and enabling rapid follow-up.
  • The RB-C1000 recipe improves over random labeling by 1.4-7.2 percentage points in summed performance and over labeled-only training by 4.1-8.7 points, depending on the band.
  • A fully supervised model needs roughly 2500 random labels to match RB-C1000's 1000-label performance on the mixed-band dataset, so the method is more label-efficient.
  • The pseudo-labeling iterations saturate around R=3, and using 5000-10000 unlabeled samples balances performance and training time.
  • The classifier is intended to be integrated into the WFST data processing pipeline for early-survey transient classification.

Reading between the lines

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

  • The same two-stage annotation strategy (label hardest, then pseudo-label confident rest) could be transferred to other label-scarce astronomical classification problems such as anomaly detection or supernova typing, where the gains from active selection on harder data appear largest.
  • Because the mixed-band dataset showed the largest active-learning gain, the method is likely most valuable when the data distribution is heterogeneous; a direct test would be to run RB-C1000 on multi-band or multi-telescope inputs with different noise properties.
  • The reported band asymmetry (r-to-g transfer fails while g-to-r works) implies that a single cross-band model is not safe; an inference is that WFST could train per-band RB-C1000 models using 1000 labels each, which the pipeline's cost makes feasible.
  • If the pseudo-label threshold or iteration count is tuned on one telescope's stamps, the optimal values may shift for another telescope; a testable extension is to re-scan τ and R on LSST-like simulated stamps.
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

4 major / 5 minor

Summary. The paper proposes RB-C1000, a deep-learning pipeline for real/bogus transient classification that combines active learning with pseudo-labeling-based semi-supervised learning. The method starts from M=900 labeled samples, has experts label K=100 hard samples, then iteratively pseudo-labels V high-confidence unlabeled samples over R rounds. Using three newly compiled ZTF datasets (ZTF-NEWg, ZTF-NEWr, ZTF-NEWm), each with 13,000 real sources (TNS supernovae) and 30,000 unvetted bogus candidates, the authors report mean accuracies of 98.8%, 98.8%, and 98.6% with only 1,000 labels, and argue that this surpasses a fully supervised ResNet-18 trained on 2,500 random labels. The paper includes repeated experiments (five trials), confidence intervals on metrics, and publicly available code and data. The authors repeatedly and explicitly caution that the datasets are not representative of the general alert stream.

Significance. If the reported efficiency gain is real, the method is a useful contribution to time-domain astronomy: it would allow early-survey operations to build a competitive classifier with far fewer manual labels. The algorithmic pipeline is coherent, the experiments are repeated five times with reported standard deviations, and the code and datasets are public, which are strengths. The central limitation is that the constructed labels do not match the stated deployment target: real is defined as TNS-reported supernovae, while bogus is a random, unvetted sample of detection candidates possibly containing real astrophysical sources. The paper acknowledges this in Section 2 and again in Sections 4 and 5, but the abstract and conclusion nevertheless frame the results as enabling general real-source discovery in WFST. The method's internal comparisons (active learning vs. random selection, SSL vs. no SSL) are meaningful on the constructed datasets, but the absolute performance and the headline efficiency claim require either representative labels or a considerably more cautious interpretation.

major comments (4)
  1. [Section 2, Section 5] The label definitions undermine the central deployment claim. Real sources are only TNS supernovae from 2019–2022, and bogus sources are randomly drawn detection candidates that, as the text acknowledges, 'might contain a few real detections.' The reported 98.8%/98.8%/98.6% accuracies therefore measure separation between TNS supernovae and unvetted artifacts, not performance on the general alert stream (stellar flares, AGN variability, faint SNe, moving objects, etc.). The paper already states this caveat, but the abstract and conclusion still claim the method can 'discover real sources that vary in brightness over time in the early stage of WFST.' Please either provide a validation set with representative, vetted real and bogus labels (e.g., a human-vetted alert sample or a set including known non-SN real sources), or explicitly restrict the performance claims to 'SNe-like real sources' and remove the implication of general WFST readiness.
  2. [Section 4.3, Table A.3, Table A.4, Table A.5] Several hyperparameters were selected using the same test sets on which final performance is reported. The text states that R=3 was chosen because it 'often shows the best performance on the overall indicator on these three datasets'; Table A.3 selects 10,000 unlabeled samples partly from test-set performance; Table A.4 selects M/K ratio 0.9 based on the highest sum on ZTF-NEWm. This makes the reported numbers optimistic and the efficiency comparison in Table A.6 circular in the sense that the configuration was tuned on the same test set. Please use a separate validation split for hyperparameter selection and report test-set results only for the selected configuration, or present the full set of configurations with an appropriate correction for multiple comparisons.
  3. [Table A.6, Section 4.3] The headline efficiency claim is not statistically established. RB-C1000 (ZTF-NEWm, sum = 587.5 ± 0.8) is compared with a fully supervised model using 2,500 random labels (sum = 586.3 ± 2.0). The difference in means is 1.2 points; using the reported standard deviations and five repeats, the difference is within approximately 1.3 standard errors, i.e., not significant at the usual 0.05 level. Please report a paired comparison across the five repeats (e.g., paired t-test or Wilcoxon signed-rank test) or otherwise quantify the uncertainty of the improvement, and adjust the claim 'significantly higher efficiency' accordingly.
  4. [Section 4.3, Table A.2] The active-learning component's standalone benefit is not consistently demonstrated. On ZTF-NEWr, adding the K hard samples degrades the ALS result relative to ITS (sum 581.4 vs. 581.7), and the final gain appears only after semi-supervised learning. The current separate attribution of '1.4%, 3.4%, and 7.2%' to active learning and '4.1%, 8.7%, and 7.3%' to semi-supervised learning is a post-hoc decomposition of a pipeline with interacting components. Please present an ablation that isolates each component's marginal contribution under the same hyperparameter selection procedure, or interpret the gains as joint effects.
minor comments (5)
  1. [Section 5] In the concluding paragraph, 'newly complied datasets' should be 'newly compiled datasets.'
  2. [Figure 6] The figure labels contain 'T rainingCNN' and 'InferenceCNN' with irregular spacing; please clean up the figure text.
  3. [Section 4.1] The validation set is taken from the (M+K) labeled samples rather than from a separate pool; this is acceptable for model selection but should be stated more explicitly, since it means the 'validation' performance is not an unbiased estimate of generalization to the unlabeled pool.
  4. [Section 2] The bogus sample construction uses only fields 774–780 on a single date (January 2, 2021). The paper should note whether any real sources from those fields on that date were excluded from the bogus pool, and how the random selection avoids overlap with the real-source set from the same fields.
  5. [Table A.1] The 'overall performance' metric (sum of six indicators) is not a standard evaluation measure and mixes quantities with different scales; consider reporting it as a secondary diagnostic or justifying its use.

Circularity Check

1 steps flagged · score 6.0 of 10

The reported 98.8/98.8/98.6 accuracies are maxima over hyperparameters (R, unlabeled-pool size, M/K ratio) selected on the same test set used for evaluation, so the headline numbers are partly fitted rather than independent held-out predictions.

  1. fitted input called prediction [Section 4.3, 'Effect of the count of unlabeled data usage', 'Effect of stability on different selections of M and K', and Table A.1/A.3/A.4]
    "We notice that SSLS ( R = 3) often shows the best performance on the overall indicator on these three datasets we collected, so we set R = 3 in this paper and the remaining experiments ... are all based on the condition R = 3. ... It can be seen from experiments that 10000 unlabeled sample usage size often give the best results. ... All other experimental results ... were obtained using the case of 10000 unlabeled data usage. ... The highest overall performance is achieved at ratios of 0.9, resulting in performances of 587.5%."

    The headline accuracies (98.8%, 98.8%, 98.6%) are computed on the same 10% test split that was used to choose R=3, the 10000-sample unlabeled pool, and the M:K=900:100 ratio. The paper explicitly selects each configuration as the one giving the best or near-best overall performance on the three evaluation datasets, then reports the resulting test-set accuracy as the method's result. Thus the reported metric is, by construction, a maximum over the test set rather than an unbiased prediction for an independent test set. The efficiency comparison against the fully supervised model in Table A.6 inherits these test-selected settings, so the claimed gain of using only 1000 labels is partly a fitted selection effect rather than a standalone prediction of the method.

full rationale

The paper is otherwise empirically self-contained: it builds three new ZTF datasets, trains the RB-C1000 pipeline with 1000 labels, and measures accuracy on held-out test samples; there is no load-bearing self-citation chain and no uniqueness theorem smuggled in. The dataset-construction limitation in Section 2, namely that 'real' sources are only TNS-reported supernovae and that the randomly drawn bogus candidates 'might contain a few real detections,' is a real external-validity caveat, but it is not circularity because the reported metrics are measured against those same constructed labels and the authors explicitly warn that performance may not transfer to the general alert stream. The principal circularity concern is the test-set-based selection of R, unlabeled-pool size, and the M/K ratio; picking the configuration with the best test performance and then presenting that test performance as the method's result makes the headline accuracies partially fitted choices rather than independent predictions. This warrants a score of 6 under the fitted-input-called-prediction pattern, while noting that the active-learning and semi-supervised-learning components still have independent algorithmic content.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The empirical results rest on hyperparameter choices that were tuned on the test sets (M, K, tau, R, unlabeled pool size) and on unvalidated assumptions about label reliability and transferability to real survey streams.

free parameters (5)
  • M (initial labeled sample count) = 900
    Selected from a sweep of M/K splits on the ZTF-NEWm test set (Table A.4).
  • K (hard samples labeled by experts) = 100
    Same sweep, keeping M+K=1000.
  • tau (confidence threshold) = 0.95
    Chosen by hand; no sensitivity analysis reported.
  • R (SSLS retraining rounds) = 3
    Chosen because it 'often shows the best performance' on the three test sets (Section 4.3).
  • Unlabeled data pool size = 10000
    Chosen from a sweep of 5000/10000/15000/20000; 10000 'often give[s] the best results' (Table A.3).
assumptions (4)
  • domain assumption Pseudo-labels above tau=0.95 are accurate enough to train on.
    The semi-supervised stage assumes high-confidence self-labels are reliable; no independent measure of pseudo-label accuracy is given.
  • domain assumption The ZTF datasets are a sufficient proxy for WFST/LSST alert streams.
    The paper's motivation is WFST/LSST, but it acknowledges the datasets contain only TNS supernovae and random candidates; cross-band test results suggest limited transfer.
  • domain assumption 63x63 triplet stamps with 1.012 arcsec/pixel encode the needed information.
    Input size and pixel scale are fixed with no ablation, so classification performance depends on this representation.
  • standard math ResNet-18 is a suitable backbone.
    Backbone follows common practice; it is not the paper's contribution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The classification of real and bogus transients using active learning and semi-supervised learning." pith.science (2026). https://pith.science/paper/7F7CCVSD

@misc{pith2026241202409,
  author       = {Pith},
  title        = {Pith review of: The classification of real and bogus transients using active learning and semi-supervised learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7F7CCVSD}},
  note         = {Machine review of arXiv:2412.02409}
}
read the original abstract

Deep-learning-based methods have been favored in astrophysics owing to their adaptability and remarkable performance and have been applied to the task of the classification of real and bogus transients. Different from most existing approaches which necessitate massive yet expensive annotated data, We aim to leverage training samples with only 1000 labels available to discover real sources that vary in brightness over time in the early stage of the WFST 6-year survey. Methods. We present a novel deep-learning method that combines active learning and semi-supervised learning to construct a competitive real/bogus classifier. Our method incorporates an active learning stage, where we actively select the most informative or uncertain samples for annotation. This stage aims to achieve higher model performance by leveraging fewer labeled samples, thus reducing annotation costs and improving the overall learning process efficiency. Furthermore, our approach involves a semi-supervised learning stage that exploits the unlabeled data to enhance the model's performance and achieve superior results compared to using only the limited labeled data.

Figures

Figures reproduced from arXiv: 2412.02409 by the authors.

Figure 1
Figure 1. Process of collecting real sources. The outputs of this process are real sources and each real detection includes a 63 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Pipeline of gathering bogus sources. The outputs of this pipeline are bogus detections and each bogus detection includes a [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Construction pipeline of the ZTF-NEWm dataset. Both the ZTF-NEWg dataset and the ZTF-NEWr dataset contain 13000 [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Examples of 63 pixels × 63 pixels triplet image stamps of our compiled datasets. The top-left part of the figure depicts two real sources in g-band and each row represents a real detection, showing the science image stamp, the reference image stamp, and the difference …
Figure 5
Figure 5. Figure 5: Distribution of magnitudes for real and bogus sources of the three datasets. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: Architecture of our method. In the ITS, each labeled sample undergoes convolutional neural network processing to train [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Process of labeling or pseudo-labeling for unlabeled samples. Following the training phase in the ITS, we select [PITH_FULL_IMAGE:figures/full_fig_p005_7.png]
Figure 8
Figure 8. Figure 8: Definitions of T N, FP, FN, and T P. pixels difference image stamp. Here, each pixel corresponds to 1.012 arcseconds. In our active and semi-supervised setting, we allocated 10% of each dataset for testing (Ntest = 4300). The amount of initial labeled data and the coun…
Figure 9
Figure 9. Figure 9: (a) Relationship between the quantity of labeled samples ( [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: Results of confusion matrices using RB-C1000 on our three datasets. The top half of this figure represents the unnormalized [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 26 canonical work pages

  1. [1]

    A., Allison, J., Anderson, S

    Abell, P. A., Allison, J., Anderson, S. F., et al. 2009, arXiv preprint arXiv:0912.0201 https://arxiv.org/abs/0912.0201

    arXiv 2009

  2. [2]

    What's the Difference? The potential for Convolutional Neural Networks for transient detection without template subtraction

    Acero-Cuellar, T., Bianco, F., Dobler, G., et al. 2022, AJ, 166, 115 https://arxiv.org/abs/2203.07390

    work page Pith review arXiv 2022

  3. [3]

    Identifying Transients in the Dark Energy Survey using Convolutional Neural Networks

    Ayyar, V., Knop Jr, R., Awbrey, A., et al. 2022, astro-ph.IM arXiv:2203.09908 https://arxiv.org/abs/2203.09908

    work page Pith review arXiv 2022

  4. [4]

    How to Find More Supernovae with Less Work: Object Classification Techniques for Difference Imaging

    Bailey, S., Aragon, C., Romano, R., et al. 2007, AJ, 665, 1246 https://arxiv.org/abs/0705.0493

    work page Pith review arXiv 2007

  5. [5]

    C., Kulkarni, S

    Bellm, E. C., Kulkarni, S. R., Graham, M. J., et al. 2018, Publ. Astron. Soc. Pac., 131, 018002 https://arxiv.org/abs/1902.01932

    arXiv 2018

  6. [6]

    1998, NeurIPS, 11 https://proceedings.neurips.cc/paper/1998/hash/b710915795b9e9c02cf10d6d2bdb688c-Abstract.html

    Bennett, K., & Demiriz, A. 1998, NeurIPS, 11 https://proceedings.neurips.cc/paper/1998/hash/b710915795b9e9c02cf10d6d2bdb688c-Abstract.html

    work page 1998

  7. [7]

    1996, ASTROPHYS J SUPPL S, 117, 393 https://aas.aanda.org/articles/aas/abs/1996/08/ds1060/ds1060.html

    Bertin, E., & Arnouts, S. 1996, ASTROPHYS J SUPPL S, 117, 393 https://aas.aanda.org/articles/aas/abs/1996/08/ds1060/ds1060.html

    work page 1996

  8. [8]

    2010, Astrophysics Source Code Library, ascl-1010

    Bertin, E. 2010, Astrophysics Source Code Library, ascl-1010. https://www.astromatic.net/software/swarp/

    work page 2010

Show all 42 references
  1. [9]

    W., Poznanski, D., et al

    Brink, H., Richards, J. W., Poznanski, D., et al. 2013, MNRAS, 435, 1047 https://arxiv.org/abs/1209.3775

    2013 arXiv

  2. [10]

    2017, AJ, 836, 97 https://doi.org/10.48550/arXiv.1701.00458

    Cabrera-Vives, G., Reyes, I., Förster, F., et al. 2017, AJ, 836, 97 https://doi.org/10.48550/arXiv.1701.00458

    2017 doi arXiv

  3. [12]

    C., Magnier, E

    Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2016, arXiv preprint arXiv:1612.05560 https://arxiv.org/abs/1612.05560

    2016 arXiv

  4. [13]

    2009, IEEE trans

    Chapelle, O., Scholkopf, B., & Zien, A. 2009, IEEE trans. neural netw., 20, 542 https://doi.org/10.1109/TNN.2009.2015974

    2009

  5. [14]

    Dark Energy Survey Collaboration, 2005, Int J Mod Phys A, 20, 3121 https://arxiv.org/abs/astro-ph/0510346

    2005 arXiv

  6. [15]

    A., Mahabal, A., Masci, F

    Duev, D. A., Mahabal, A., Masci, F. J., et al. 2019, MNRAS, 489, 3582 https://doi.org/10.1093/mnras/stz2357

    2019 doi

  7. [16]

    C., San Martín, J., et al

    Förster, F., Maureira, J. C., San Martín, J., et al. 2016, AJ, 832, 155 https://arxiv.org/abs/1609.03567

    2016 arXiv

  8. [17]

    A., D’Andrea, C

    Goldstein, D. A., D’Andrea, C. B., Fischer, J. A., et al. 2015, AJ, 150, 82 https://arxiv.org/abs/1504.02936

    2015 arXiv

  9. [18]

    2017, MNRAS, 499, 3130 https://arxiv.org/abs/2004.13877

    Gómez, C., Neira, M., Hernández Hoyos, M., et al. 2017, MNRAS, 499, 3130 https://arxiv.org/abs/2004.13877

    2017 arXiv

  10. [19]

    2022, MNRAS, 513, 1742 https://doi.org/10.1093/mnras/stac983

    Goode, S., Cooke, J., Zhang, J., et al. 2022, MNRAS, 513, 1742 https://doi.org/10.1093/mnras/stac983

    2022 doi

  11. [20]

    2017, ICCV, 2961 https://arxiv.org/abs/1703.06870

    He, K., Gkioxari, G., Dollár, P., et al. 2017, ICCV, 2961 https://arxiv.org/abs/1703.06870

    2017 arXiv

  12. [21]

    2016, CVPR, 770 https://arxiv.org/abs/1512.03385

    He, K., Zhang, X., Ren, S. 2016, CVPR, 770 https://arxiv.org/abs/1512.03385

    2016 arXiv

  13. [22]

    2021, EXP ASTRON, 51, 319 https://arxiv.org/abs/2104.13950

    Hosenie, Z., Bloemen, S., Groot, P., et al. 2021, EXP ASTRON, 51, 319 https://arxiv.org/abs/2104.13950

    2021 arXiv

  14. [23]

    2022, Universe, 9, 7 https://arxiv.org/abs/2212.11535

    Hu, M., Hu, L., Jiang, J., et al. 2022, Universe, 9, 7 https://arxiv.org/abs/2212.11535

    2022 arXiv

  15. [24]

    1999, ICML, 99, 200 https://proceedings.neurips.cc/paper/1998/hash/b710915795b9e9c02cf10d6d2bdb688c-Abstract.html

    Joachims, T. 1999, ICML, 99, 200 https://proceedings.neurips.cc/paper/1998/hash/b710915795b9e9c02cf10d6d2bdb688c-Abstract.html

    1999

  16. [25]

    L., Lyman, J., Steeghs, D., et al

    Killestein, T. L., Lyman, J., Steeghs, D., et al. 2021, MNRAS, 503, 4838 https://arxiv.org/abs/2102.09892

    2021 arXiv

  17. [26]

    LeCun, Y., Boser, B., Denker, J. S. 1989, Neural Comput, 1, 541 https://doi.org/10.1162/neco.1989.1.4.541

    1989 doi

  18. [27]

    2022, A&A 664, A81 https://arxiv.org/abs/2112.10280

    Makhlouf, K., Turpin, D., Corre, D., et al. 2022, A&A 664, A81 https://arxiv.org/abs/2112.10280

    2022 arXiv

  19. [28]

    2016, PASJ, psw096 https://doi.org/10.1093/pasj/psw096

    Morii, M., Ikeda, S., Tominaga, N., et al. 2016, PASJ, psw096 https://doi.org/10.1093/pasj/psw096

    2016 doi

  20. [29]

    2017, Automatic differentiation in pytorch https://openreview.net/forum?id=BJJsrmfCZ

    Paszke, A., Gross, S., Chintala, S., et al. 2017, Automatic differentiation in pytorch https://openreview.net/forum?id=BJJsrmfCZ

    2017

  21. [30]

    Prince, M. 2004, J. Eng. Educ., 93, 223 https://onlinelibrary.wiley.com/doi/abs/10.1002/j.2168-9830.2004.tb00809.x

    2004

  22. [31]

    2015, NeurIPS, 28 https://arxiv.org/abs/1506.01497

    Ren, S., He, K., Girshick, R., et al. 2015, NeurIPS, 28 https://arxiv.org/abs/1506.01497

    2015 arXiv

  23. [32]

    A., Reyes, I., et al

    Reyes, E., Estévez, P. A., Reyes, I., et al. 2018, IJCNN, 1 https://arxiv.org/abs/1808.03626

    2018 arXiv

  24. [33]

    Reyes-Jainaga, I., Förster, F., Arancibia, A. M. M., et al. 2023, APJL, 952, L43 https://arxiv.org/abs/2304.13080

    2023 arXiv

  25. [34]

    2017, arXiv preprint arXiv:1710.01422 https://arxiv.org/abs/1710.01422

    Sedaghat, N., & Mahabal, A. 2017, arXiv preprint arXiv:1710.01422 https://arxiv.org/abs/1710.01422

    2017 arXiv

  26. [35]

    2009, Active learning literature survey https://minds.wisconsin.edu/handle/1793/60660

    Settles, B. 2009, Active learning literature survey https://minds.wisconsin.edu/handle/1793/60660

    2009

  27. [36]

    2020, MNRAS, 497, 2641 https://ui.adsabs.harvard.edu/abs/2020MNRAS.497.2641T/abstract

    Turpin, D., Ganet, M., Antier, S., et al. 2020, MNRAS, 497, 2641 https://ui.adsabs.harvard.edu/abs/2020MNRAS.497.2641T/abstract

    2020

  28. [37]

    2023, SCI CHINA PHYS MECH, 66, 109512 https://arxiv.org/abs/2306.07590

    Wang, T., Liu, G., and Cai, Z., et al. 2023, SCI CHINA PHYS MECH, 66, 109512 https://arxiv.org/abs/2306.07590

    2023 arXiv

  29. [38]

    E., Lintott, C

    Wright, D. E., Lintott, C. J., Smartt, S. J., et al. 2017, MNRAS, 472, 1315 https://arxiv.org/abs/1707.05223

    2017 arXiv

  30. [39]

    E., Smartt, S

    Wright, D. E., Smartt, S. J., Smith, K. W., et al. 2015, MNRAS, 449, 451 https://arxiv.org/abs/1501.05470

    2015 arXiv

  31. [40]

    2016, PMLR, 40 https://proceedings.mlr.press/v48/yanga16.html

    Yang, Z., Cohen, W., & Salakhudinov, R. 2016, PMLR, 40 https://proceedings.mlr.press/v48/yanga16.html

    2016

  32. [41]

    Zhu, X. J. 2005, Semi-supervised learning literature survey https://minds.wisconsin.edu/handle/1793/60444

    2005

  33. [42]

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

    ENTRY address archiveprefix author booktitle chapter edition editor howpublished institution eprint journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sent...

  34. [43]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.