Pith. sign in

REVIEW 4 major objections 5 minor 60 references

From Galaxy Zoo DECaLS to BASS/MzLS: detailed galaxy morphology classification with unsupervised domain adaption

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 ↗

Pith's one-line read Unsupervised domain adaptation transfers a detailed galaxy morphology classifier from DECaLS to BASS/MzLS without new labels.

desk verdict Useful UDA application and catalogue, but the headline parity claim rests on a small unmatched test set and needs revision. read the letter →

arxiv 2412.15533 v1 pith:W6KUD3RD submitted 2024-12-20 astro-ph.GA astro-ph.IMcs.CV

Renhao Ye , Shiyin Shen , Rafael S. de Souza , Quanfeng Xu , Mi Chen , Zhu Chen , Emille E. O. Ishida , Alberto Krone-Martins
show 1 more author
Rupesh Durgesh
This is my paper · ORCID
classification astro-ph.GAastro-ph.IMcs.CV
keywords galaxymorphologyunsuperviseddomainadaptationtransferlearningDESILegacySurveysDECaLSBASSMzdeep
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

The paper sets out to show that a deep-learning model for detailed galaxy morphology classification, trained on DECaLS images with Galaxy Zoo volunteer labels, can be adapted to BASS/MzLS images without collecting a single new morphology label. The authors achieve this by unsupervised domain adaptation (UDA): they fine-tune the model's feature extractor on unlabeled BMz galaxies, using pseudo-labels from spherical k-means clustering to align the target domain embedding distribution with the fixed source-domain classifier. On the 3,618 galaxies common to both surveys, the adapted model performs significantly better than directly applying the source model, reaching accuracy close to the source model's own performance on DECaLS. If this holds, it would mean that detailed morphology labels can be propagated across surveys with different telescopes and image quality for a modest computational cost.

What carries the argument

The load-bearing mechanism is a pseudo-label alignment loop built on spherical k-means and a fixed classifier. Each epoch, the target-domain feature embeddings are clustered in cosine space; galaxies whose embedding lies close enough to a cluster center (cosine similarity above a schedule threshold) receive the corresponding morphology pseudo-label. The feature extractor is then fine-tuned by a softmax cross-entropy loss (with temperature) that pushes these embeddings onto the fixed source classifier's weight vectors, which also serve as the initial cluster centers. This alignment transfers the source model's decision boundaries to the target domain without requiring any target labels.

What would settle it

Apply the fully trained UDA model to the 3,618 common DECaLS/BMz galaxies and compare accuracy per question to the source model without adaptation; if the UDA model does not substantially beat the source model on questions like 'Bar' and 'Spiral Arm Count', the claimed transfer has not occurred.

Watch

Extended reading notes

Core claim

The central claim is that UDA with pseudo-label alignment closes most of the 'data shift' gap between DECaLS and BMz imaging. The authors start from a source model (EfficientNet-v2-s with a Dirichlet-multinomial loss) that matches earlier GZD-5 results. They then freeze the classifier weights and fine-tune the feature extractor on 248,088 unlabeled BMz galaxies, assigning pseudo-labels only to embeddings whose cosine similarity to a spherical k-means cluster center (initialized from the classifier weight vectors) exceeds a threshold that decays during training. On the overlapping-footprint test set of 3618 BMz galaxies that were excluded from training, the UDA model improves accuracy over the direct source model on most questions (e.g., 'Bar' from 75.9% to 83.4%, close to the 82.1% source-domain value), confirming that the transfer works and that the released catalogue of 248,088 classifications is not merely source-model extrapolation.

Load-bearing premise

The method assumes that the DECaLS and BASS/MzLS galaxies are drawn from the same physical galaxy population, so aligning their feature embeddings with pseudo-labels is meaningful.

Editorial extensions

If this is right

  • If the UDA transfer works, the released catalogue provides detailed morphology probabilities for 248,088 BASS/MzLS galaxies, enabling studies of bar fraction, bulge size, and spiral structure in the northern DESI footprint.
  • The pipeline can be rerun for other surveys that share a physical galaxy population with DECaLS, such as future wide-field imaging, without waiting for volunteer labels.
  • The two-step training (source model then label-free fine-tune) means that new surveys need only unlabeled images and a matched source catalogue to obtain per-question morphology predictions.
  • Because the classifier weights are shared between domains, the target model is directly comparable with the source model, giving consistent morphology definitions across surveys.

Reading between the lines

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

  • The success of the pseudo-label alignment suggests that a broader class of cross-survey classifiers (e.g., for photometric redshifts or star-galaxy separation) could be transferred the same way, provided the physical domain overlap is controlled.
  • The probability-simplex analysis indicates that lower-resolution BMz images push predictions toward the center (ambiguous) region; a resolution-aware or multi-resolution augmentation might recover some of the tail of the distribution.
  • If the UDA model's pseudo-labels are reliable enough, one could use them as training labels in a second round of supervised fine-tuning, potentially reducing the remaining gap on detailed features like spiral winding.
  • The requirement that the two surveys share the same physical galaxy population implies the method cannot be blindly applied to fainter or higher-redshift samples; a domain-discriminator check could quantify when the assumption breaks.
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 addresses the problem of applying a galaxy morphology classifier trained on DECaLS images (with Galaxy Zoo DECaLS 5 volunteer labels) to BASS/MzLS (BMz) images, which suffer from a distributional shift. The authors train a source-domain EfficientNet-v2-s model on DECaLS images with Dirichlet-multinomial loss, then fine-tune it on 248,088 unlabelled BMz galaxies using an unsupervised domain adaptation (UDA) method based on spherical K-means pseudo-labels and a temperature-scaled contrastive loss. They evaluate the target model on 3,618 overlap galaxies (835 after vote filtering) between DECaLS and BMz, and also compare their predictions against those of Walmsley et al. (2023). They release a catalogue of detailed morphology classifications for 248,088 BMz galaxies. The central claim is that the UDA-fine-tuned target model significantly improves over directly applying the source model to BMz and reaches performance comparable to the source-domain model on DECaLS.

Significance. If the claim holds, the paper demonstrates a label-free route for transferring detailed morphology classifiers across galaxy surveys and provides a practical catalogue for 248,088 BMz galaxies, complementing the Zoobot series. The source-domain model reproduces W+21-level performance, and the evaluation on held-out volunteer labels excluded from training is a genuine strength. The released catalogue, code availability, and use of well-established GZD-5 labels give the work practical value. However, the load-bearing quantitative evidence for the parity claim is weakened by small per-question samples, absent uncertainty estimates, an uncontrolled cross-sample comparison for the 'comparable to source domain' statement, and an unresolved discrepancy with GZD-8 labels in Appendix A. These issues are addressable with additional analysis, but they currently limit the strength of the central conclusion.

major comments (4)
  1. [§4.2, Tables 2 and 3] The claim that the target model 'reaches a level comparable to that of the source domain' compares the target model on 835 BMz galaxies (Table 3b) with the source model on a different DECaLS validation set (Table 2). This is not an apples-to-apples comparison: the 835-galaxy overlap subset may be unrepresentative (e.g., brighter or higher-signal) of the full DECaLS validation sample. To support the parity claim, the authors should evaluate the source model on the DECaLS images of the same 835 overlap galaxies, or otherwise demonstrate that the overlap subset has the same difficulty distribution as the full validation set. Without this control, the apparent 'parity' could be driven by sample selection rather than successful domain adaptation.
  2. [§4.2, Table 3] Many per-question counts are small (Bar N=133, Spiral winding N=98, Edge-on bulge N=16), and no confidence intervals or significance tests are provided. For example, the Bar accuracy improvement from 75.9% to 83.4% on N=133 is roughly a 1.5–2 sigma effect, and 'How Rounded' actually degrades from 90.5% to 87.7% after adaptation. The statement that the target model 'significantly improves performance' is therefore not quantitatively supported for several questions. The authors should report bootstrap confidence intervals or a statistical test for each metric, and discuss the 'How Rounded' degradation in this context rather than dismissing it as an acceptable trade-off.
  3. [Appendix A, Table A1] The target model's accuracy on 'Spiral winding' with GZD-8 labels is 65.6%, which is more than 12 points below its Table 3 value on the GZD-5 overlap subset (79.5%) and well below the source-domain value in Table 2 (79.1%). The authors attribute this to volunteer bias in GZD-8, but the same bias argument could apply to the Table 3 ground truth, and the consistency check in Table 4 uses predictions derived from GZD-8 labels. This internal tension undermines the robustness of the claimed parity. The authors should either provide a quantitative analysis of the label-bias hypothesis (e.g., comparing volunteer vote distributions on the same galaxies) or temper the parity claim.
  4. [§4.2.1, Table 4] Using Walmsley et al. (2023) predictions as reference labels measures agreement with another model, not accuracy against human labels. The statement that the target model 'does not exhibit significant bias under data shift' is not warranted by agreement with a model whose own accuracy is not established on the same galaxies. Moreover, the comparison in Table 4 again contrasts target-on-BMz with source-on-GZD (Table 2), which confounds domain shift with sample differences. A cleaner check would compare source and target model predictions on the same set of BMz galaxies with the same reference labels, and ideally on a human-labelled subset.
minor comments (5)
  1. [§4.2.1] The paragraph beginning 'Specifically, seven morphology questions...' is repeated verbatim, which should be fixed.
  2. [§3.1, Eq. (1)] The footnote 'assuming a set of concomitant 5 q = 10 questions' is unclear; the authors should clarify what 'concomitant' means here and whether the loss assumes independence or dependence among questions.
  3. [Tables 2 and 3] The tables report counts but not uncertainty intervals; adding bootstrap CIs or at least standard errors for each accuracy value would greatly improve interpretability.
  4. [§2.3] The sentence 'We don’t use the prediction results of 314,000 galaxies from W+21' is a footnote and would be clearer in the main text or omitted entirely.
  5. [Abstract and title] The abstract uses 'domain adaption' while the title and text use 'domain adaptation'; please standardize the spelling.

Circularity Check

1 steps flagged · score 1.0 of 10

No load-bearing circularity: the central UDA claim is evaluated against held-out GZD-5 volunteer labels, while the only self-referential validation (the distribution-match comparison in Sec. 4.3) is a restatement of the UDA objective and is not what carries the paper's conclusion.

  1. self definitional [Section 4.3, 'From DECaLS to BMz: morphology feature probability distribution' (Fig. 3)]
    "a perfect UDA model should predict the same probability distribution of the morphology features on all BMz galaxies as the source domain model on all DECaLS galaxies."

    The UDA loss in Eq. (2) explicitly fine-tunes f_t so that target embeddings align with the fixed source classifier W_s, using pseudo-labels from spherical K-means initialized by W_s. Hence a post-training match between the target and source predicted morphology distributions (Fig. 3) is the algorithm's own optimization objective, not an independent test of morphology accuracy. The paper presents this as 'further show the ability of our UDA model,' so the evidence is self-definitional. It is not load-bearing for the headline claim, which is supported by held-out GZD-5 labels in Table 3.

full rationale

The main derivation chain is self-contained. The source model is supervised by GZD-5 volunteer votes through the Dirichlet loss (Eq. 1); the target model is fine-tuned on unlabeled BMz images using pseudo-labels and the UDA loss (Eq. 2); and the claimed improvement is measured on 3618 overlap galaxies whose GZD-5 labels are explicitly excluded from both the training and validation sets (Sec. 4.2, Table 3). That evaluation is therefore independent of the model's own pseudo-labels and is not circular. The large-sample consistency check in Table 4 uses predictions from Walmsley et al. (2023) as reference labels, and Appendix A compares against GZD-8 volunteer labels; both are external benchmarks, not self-citations. The only coauthor citation relevant to the motivation (Xu et al. 2023) is not load-bearing because the data shift is also demonstrated empirically by the source model's degraded performance on BMz images in Table 3. The single self-referential element is the Sec. 4.3 distribution-overlap argument, which amounts to verifying that the alignment objective was achieved; this is a minor tautology rather than a load-bearing circular step. Statistical concerns such as small per-question test counts, absent error bars, and the use of model predictions as labels for the large-sample check are correctness and robustness issues, not circularity.

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

The central claim rests mainly on the empirical evaluation, but the UDA training itself depends on several hand-chosen hyperparameters and on the unverified assumption that the two surveys sample the same morphology distribution. No new physical entities are introduced.

free parameters (4)
  • UDA temperature tau = 0.05
    Set by hand in Section 3.2 to control the sharpness of the softmax in the domain adaptation loss; no sensitivity analysis is reported.
  • Pseudo-label threshold schedule = 0.9 - 0.02*epoch
    Hand-designed in Section 3.2 to constrain the search region near the source classifier; no evidence this schedule is optimal.
  • UDA learning rate = 1e-6
    Chosen for AdamW in Section 3.2; not reported as tuned.
  • Source-model hyperparameters = LR 1e-2 to 1e-6, batch 256, dropout 0.3
    Training setup in Section 3.1 following typical practice; not systematically optimized in this paper.
assumptions (4)
  • domain assumption DECaLS and BMz MGS galaxies are drawn from the same physical distribution of morphologies
    Stated in Section 6 as a requirement for UDA; if violated, aligning the embeddings would be invalid.
  • domain assumption Spherical K-means pseudo-labels on the latent embeddings correspond to true morphology features
    Implicit in Section 3.2; the quality of pseudo-labels is not independently verified.
  • standard math Volunteer votes follow a Dirichlet-multinomial model
    Used in Section 3.1 (equation 1) following Walmsley et al. (2021); standard Bayesian treatment.
  • domain assumption The source classifier weights Ws remain valid for the target domain after feature alignment
    Assumed in Section 3.2 where Ws is frozen; the UDA loss forces target embeddings to be classified correctly by the source classifier.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Galaxy Zoo DECaLS to BASS/MzLS: detailed galaxy morphology classification with unsupervised domain adaption." pith.science (2026). https://pith.science/paper/W6KUD3RD

@misc{pith2026241215533,
  author       = {Pith},
  title        = {Pith review of: From Galaxy Zoo DECaLS to BASS/MzLS: detailed galaxy morphology classification with unsupervised domain adaption},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W6KUD3RD}},
  note         = {Machine review of arXiv:2412.15533}
}
read the original abstract

The DESI Legacy Imaging Surveys (DESI-LIS) comprise three distinct surveys: the Dark Energy Camera Legacy Survey (DECaLS), the Beijing-Arizona Sky Survey (BASS), and the Mayall z-band Legacy Survey (MzLS). The citizen science project Galaxy Zoo DECaLS 5 (GZD-5) has provided extensive and detailed morphology labels for a sample of 253,287 galaxies within the DECaLS survey. This dataset has been foundational for numerous deep learning-based galaxy morphology classification studies. However, due to differences in signal-to-noise ratios and resolutions between the DECaLS images and those from BASS and MzLS (collectively referred to as BMz), a neural network trained on DECaLS images cannot be directly applied to BMz images due to distributional mismatch. In this study, we explore an unsupervised domain adaptation (UDA) method that fine-tunes a source domain model trained on DECaLS images with GZD-5 labels to BMz images, aiming to reduce bias in galaxy morphology classification within the BMz survey. Our source domain model, used as a starting point for UDA, achieves performance on the DECaLS galaxies' validation set comparable to the results of related works. For BMz galaxies, the fine-tuned target domain model significantly improves performance compared to the direct application of the source domain model, reaching a level comparable to that of the source domain. We also release a catalogue of detailed morphology classifications for 248,088 galaxies within the BMz survey, accompanied by usage recommendations.

Figures

Figures reproduced from arXiv: 2412.15533 by the authors.

Figure 1
Figure 1. Composite images (consisting of 𝑔𝑟 𝑧 bands) of a randomly selected spiral galaxy in DECaLS (left) and BMz (right). Both images are being processed using the same arcsinh stretching method as the DESI Legacy Survey Viewer. resampled to match the pixel scale of DECaLS and MzLS. This op￾eration results in a different background noise distribution of BMz images compared to DECaLS images, as demonstrated in [PITH_FULL_I… view at source ↗
Figure 2
Figure 2. The schematic diagram of target domain training including cutout input 𝑥s from source domain, 𝑥t from target domain, the feature extractor 𝑓s(· ) and 𝑓t(· ), and the classifier 𝑊s. Spherical K-means are used to obtain pseudo-labels. The triangle represents the feature embedding not assigned with a pseudo-label, and the galaxy-like shape represents the feature embedding of assigned morphology. Red colour represents t… view at source ↗
Figure 3
Figure 3. Expected probability 𝜌ˆ 𝑚𝑞 𝑞 of the Dirichlet distribution of the model output visualised by probability simplex for the question ‘Bar’, where the three vertices represent the corresponding three features, namely ‘Weak Bar’, ‘No Bar’, and ‘Strong Bar’. The scatter point is 1% sampling from BMz galaxies. Each data point within a triangle represents the expected probability combinations of the features. To read the pr… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Examples of BMz galaxies are shown with ‘Strong Bar’ (left), ‘Weak Bar’ (middle), and ‘No Bar’ features (right), respectively. All galaxies are selected with 𝜌ˆ 𝑚bar bar > 0.5. The galaxies in the top row have lower variance (the top 15% in 𝜎2 ), while those in the bot…
Figure 5
Figure 5. Figure 5: Example of BMz galaxies selected with ‘Strong Bar’ features. The top row of galaxies are following a decision tree: 𝜌ˆ featured or disc smooth or featured, 𝜌ˆ not edge on edge on , 𝜌ˆ strong bar bar are larger than other features (the top 30% 𝜎2 ), and while the galaxi…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 16 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTION or pop #1...

  2. [2]

    K., Turp M

    Ackermann S., Schawinski K., Zhang C., Weigel A. K., Turp M. D., 2018, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/sty1398 , 479, 415

    doi:10.1093/mnras/sty1398 2018

  3. [3]

    K., Balogh M

    Baldry I. K., Balogh M. L., Bower R., Glazebrook K., Nichol R. C., 2004, in AIP Conference Proceedings . pp 106--119, @doi 10.1063/1.1848322 , http://arxiv.org/abs/astro-ph/0410603

    arXiv 2004

  4. [4]

    R., et al., 2001, @doi [AJ] 10.1086/320405 , 121, 2358

    Blanton M. R., et al., 2001, @doi [AJ] 10.1086/320405 , 121, 2358

    doi:10.1086/320405 2001

  5. [5]

    Measuring detailed galaxy morphologies for Euclid with Machine Learning , http://arxiv.org/abs/2402.10187

    Collaboration E., et al., 2024, Euclid preparation. Measuring detailed galaxy morphologies for Euclid with Machine Learning , http://arxiv.org/abs/2402.10187

    work page arXiv 2024

  6. [6]

    W., et al., 2010, @doi [Monthly Notices of the Royal Astronomical Society] 10.1111/j.1365-2966.2009.15686.x , 401, 1043

    Darg D. W., et al., 2010, @doi [Monthly Notices of the Royal Astronomical Society] 10.1111/j.1365-2966.2009.15686.x , 401, 1043

    arXiv 2010

  7. [7]

    Dey A., et al., 2019, @doi [AJ] 10.3847/1538-3881/ab089d , 157, 168

    doi:10.3847/1538-3881/ab089d 2019

  8. [8]

    W., Dambre J., 2015, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stv632 , 450, 1441

    Dieleman S., Willett K. W., Dambre J., 2015, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stv632 , 450, 1441

    doi:10.1093/mnras/stv632 2015

Show all 60 references
  1. [9]

    L., 2018, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/sty338 , 476, 3661

    Domínguez Sánchez H., Huertas-Company M., Bernardi M., Tuccillo D., Fischer J. L., 2018, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/sty338 , 476, 3661

    2018 doi

  2. [10]

    Domínguez Sánchez H., et al., 2019, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/sty3497 , 484, 93

    2019 doi

  3. [11]

    T., Martínez-Lombilla C., Knapen J

    Dullo B. T., Martínez-Lombilla C., Knapen J. H., 2016, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stw1868 , 462, 3800

    2016 doi

  4. [12]

    Euclid Collaboration et al., 2022, @doi [Astronomy and Astrophysics] 10.1051/0004-6361/202141938 , 662, A112

    2022 doi

  5. [13]

    Gal Y., Ghahramani Z., 2016, Dropout as a Bayesian Approximation : Representing Model Uncertainty in Deep Learning , http://arxiv.org/abs/1506.02142

    2016 arXiv

  6. [14]

    Gong Y., et al., 2019, @doi [The Astrophysical Journal] 10.3847/1538-4357/ab391e , 883, 203

    2019 doi

  7. [15]

    M., et al., 2015, @doi [Astronomy and Astrophysics] 10.1051/0004-6361/201525938 , 581, A103

    González Delgado R. M., et al., 2015, @doi [Astronomy and Astrophysics] 10.1051/0004-6361/201525938 , 581, A103

    2015 doi

  8. [16]

    J., Shlens J., Szegedy C., 2015, Explaining and Harnessing Adversarial Examples , @doi 10.48550/arXiv.1412.6572 , http://arxiv.org/abs/1412.6572

    Goodfellow I. J., Shlens J., Szegedy C., 2015, Explaining and Harnessing Adversarial Examples , @doi 10.48550/arXiv.1412.6572 , http://arxiv.org/abs/1412.6572

    2015 doi arXiv

  9. [17]

    Goto T., et al., 2003, @doi [Publications of the Astronomical Society of Japan] 10.1093/pasj/55.4.739 , 55, 739

    2003 doi

  10. [18]

    J., Lintott C., Kruk S., Masters K

    Géron T., Smethurst R. J., Lintott C., Kruk S., Masters K. L., Simmons B., Stark D. V., 2021, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stab2064 , 507, 4389

    2021 doi

  11. [19]

    stad2238

    Hannon S., et al., 2023, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stad2238 , p. stad2238

    2023 doi

  12. [20]

    E., et al., 2016, @doi [Mon

    Hart R. E., et al., 2016, @doi [Mon. Not. R. Astron. Soc.] 10.1093/mnras/stw1588 , 461, 3663

    2016 doi

  13. [21]

    E., et al., 2017, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stx2137 , 472, 2263

    Hart R. E., et al., 2017, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stx2137 , 472, 2263

    2017 doi

  14. [22]

    He Z., Li N., 2022, @doi [Res. Astron. Astrophys.] 10.1088/1674-4527/ac839b , 22, 095021

    2022 doi

  15. [23]

    He K., Fan H., Wu Y., Xie S., Girshick R., 2020, Momentum Contrast for Unsupervised Visual Representation Learning , @doi 10.48550/arXiv.1911.05722 , http://arxiv.org/abs/1911.05722

    2020 doi arXiv

  16. [24]

    D., Nelson B., Rubinstein B

    Huang L., Joseph A. D., Nelson B., Rubinstein B. I., Tygar J. D., 2011, in Proceedings of the 4th ACM workshop on Security and artificial intelligence. ACM, Chicago Illinois USA, pp 43--58, @doi 10.1145/2046684.2046692 , https://dl.acm.org/doi/10.1145/2046684.2046692

    2011

  17. [25]

    pp 1203--1214

    Huang J., Guan D., Xiao A., Lu S., Shao L., 2022, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp 1203--1214

    2022

  18. [26]

    P., 1926, @doi [ApJ] 10.1086/143018 , 64, 321

    Hubble E. P., 1926, @doi [ApJ] 10.1086/143018 , 64, 321

    1926 doi

  19. [27]

    Ivezić Z., et al., 2019, @doi [The Astrophysical Journal] 10.3847/1538-4357/ab042c , 873, 111

    2019 doi

  20. [29]

    J., et al., 2017, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stx1026 , 469, 3363

    Kruk S. J., et al., 2017, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stx1026 , 469, 3363

    2017 doi

  21. [30]

    Kumar A., 2023, A study of the evolution of bulges and disks of spiral galaxies in interacting and isolated environments, http://arxiv.org/abs/2306.11045

    2023 arXiv

  22. [31]

    K., 2022, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stac1487 , 514, 2497

    Kumar A., Kataria S. K., 2022, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stac1487 , 514, 2497

    2022 doi

  23. [32]

    pp 9641--9650

    Li R., Jiao Q., Cao W., Wong H.-S., Wu S., 2020, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp 9641--9650

    2020

  24. [33]

    Li B., Shi Y., Yu Q., Wang J., 2024, Unsupervised Cross - Domain Image Retrieval via Prototypical Optimal Transport , http://arxiv.org/abs/2402.18411

    2024 arXiv

  25. [34]

    J., et al., 2008, @doi [Monthly Notices of the Royal Astronomical Society] 10.1111/j.1365-2966.2008.13689.x , 389, 1179

    Lintott C. J., et al., 2008, @doi [Monthly Notices of the Royal Astronomical Society] 10.1111/j.1365-2966.2008.13689.x , 389, 1179

    2008

  26. [35]

    Lintott C., et al., 2011, @doi [Monthly Notices of the Royal Astronomical Society] 10.1111/j.1365-2966.2010.17432.x , 410, 166

    2011

  27. [36]

    Loshchilov I., Hutter F., 2019, Decoupled Weight Decay Regularization , http://arxiv.org/abs/1711.05101

    2019 arXiv

  28. [37]

    E., de Carvalho R

    Margoniner V. E., de Carvalho R. R., 2000, @doi [The Astronomical Journal] 10.1086/301318 , 119, 1562

    2000 doi

  29. [38]

    Pooch E. H. P., Ballester P., Barros R. C., 2020, in Petersen J., San José Estépar R., Schmidt-Richberg A., Gerard S., Lassen-Schmidt B., Jacobs C., Beichel R., Mori K., eds, Thoracic Image Analysis . Lecture Notes in Computer Science . Springer International Publishing, Cham,...

    2020 doi

  30. [39]

    Porter-Temple R., et al., 2022, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stac1936 , 515, 3875

    2022 doi

  31. [40]

    Seo E., Kim S., Lee Y., Han S.-I., Kim H.-S., Rey S.-C., Song H., 2023, @doi [Publications of the Astronomical Society of the Pacific] 10.1088/1538-3873/ace851 , 135, 084101

    2023 doi

  32. [41]

    J., White S

    Shen S., Mo H. J., White S. D. M., Blanton M. R., Kauffmann G., Voges W., Brinkmann J., Csabai I., 2003, @doi [Monthly Notices of the Royal Astronomical Society] 10.1046/j.1365-8711.2003.06740.x , 343, 978

    2003

  33. [42]

    N., Topin N., 2018, Super- Convergence : Very Fast Training of Neural Networks Using Large Learning Rates , @doi 10.48550/arXiv.1708.07120 , http://arxiv.org/abs/1708.07120

    Smith L. N., Topin N., 2018, Super- Convergence : Very Fast Training of Neural Networks Using Large Learning Rates , @doi 10.48550/arXiv.1708.07120 , http://arxiv.org/abs/1708.07120

    2018 doi arXiv

  34. [43]

    A., et al., 2002, @doi [The Astronomical Journal] 10.1086/342343 , 124, 1810

    Strauss M. A., et al., 2002, @doi [The Astronomical Journal] 10.1086/342343 , 124, 1810

    2002 doi

  35. [44]

    Tahmasebzadeh B., et al., 2023, Schwarzschild Modeling of Barred S0 Galaxy NGC 4371, @doi 10.48550/arXiv.2310.00497 , https://ui.adsabs.harvard.edu/abs/2023arXiv231000497T

    2023 doi arXiv

  36. [45]

    V., 2021, EfficientNetV2 : Smaller Models and Faster Training , @doi 10.48550/arXiv.2104.00298 , http://arxiv.org/abs/2104.00298

    Tan M., Le Q. V., 2021, EfficientNetV2 : Smaller Models and Faster Training , @doi 10.48550/arXiv.2104.00298 , http://arxiv.org/abs/2104.00298

    2021 doi arXiv

  37. [46]

    Tang H., Scaife A. M. M., Leahy J. P., 2019, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stz1883 , 488, 3358

    2019 doi

  38. [47]

    Tsvetkov N., Moiseev A., Sil'chenko O., Katkov I., Oparin D., Uklein R., Smirnova A., 2024, @doi [Astronomy and Astrophysics] 10.1051/0004-6361/202348437 , 682, L16

    2024 doi

  39. [48]

    Walmsley M., Scaife A. M. M., 2023, Rare Galaxy Classes Identified In Foundation Model Representations , @doi 10.48550/arXiv.2312.02910 , http://arxiv.org/abs/2312.02910

    2023 doi arXiv

  40. [49]

    Walmsley M., et al., 2020, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stz2816 , 491, 1554

    2020 doi

  41. [50]

    Walmsley M., et al., 2021, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stab2093 , 509, 3966

    2021 doi

  42. [51]

    V., Bowles M., Scaife A

    Walmsley M., Slijepcevic I. V., Bowles M., Scaife A. M. M., 2022a, Towards Galaxy Foundation Models with Hybrid Contrastive Learning , @doi 10.48550/arXiv.2206.11927 , http://arxiv.org/abs/2206.11927

    doi arXiv

  43. [52]

    Walmsley M., et al., 2022b, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stac525 , 513, 1581

    doi

  44. [53]

    7M Galaxies in the DESI Legacy Imaging Surveys , @doi 10.48550/arXiv.2309.11425 , http://arxiv.org/abs/2309.11425

    Walmsley M., et al., 2023, Galaxy Zoo DESI : Detailed Morphology Measurements for 8. 7M Galaxies in the DESI Legacy Imaging Surveys , @doi 10.48550/arXiv.2309.11425 , http://arxiv.org/abs/2309.11425

    2023 doi arXiv

  45. [54]

    Multimedia] 10.1109/TMM.2022.3146744 , 25, 1665

    Wang R., Wu Z., Weng Z., Chen J., Qi G.-J., Jiang Y.-G., 2023, @doi [IEEE Trans. Multimedia] 10.1109/TMM.2022.3146744 , 25, 1665

    2023

  46. [55]

    K., et al., 2017, @doi [The Astrophysical Journal] 10.3847/1538-4357/aa8097 , 845, 145

    Weigel A. K., et al., 2017, @doi [The Astrophysical Journal] 10.3847/1538-4357/aa8097 , 845, 145

    2017 doi

  47. [56]

    W., et al., 2013, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stt1458 , 435, 2835

    Willett K. W., et al., 2013, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stt1458 , 435, 2835

    2013 doi

  48. [57]

    Xu Q., et al., 2023, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stad3181 , 526, 6391

    2023 doi

  49. [58]

    J., van den Bosch F

    Yang X., Mo H. J., van den Bosch F. C., Pasquali A., Li C., Barden M., 2007, @doi [The Astrophysical Journal] 10.1086/522027 , 671, 153

    2007 doi

  50. [59]

    J., van den Bosch F

    Zhang Y., Yang X., Wang H., Wang L., Luo W., Mo H. J., van den Bosch F. C., 2015, @doi [The Astrophysical Journal] 10.1088/0004-637X/798/1/17 , 798, 17

    2015 doi

  51. [60]

    Ćiprijanović A., et al., 2022, DeepAdversaries : Examining the Robustness of Deep Learning Models for Galaxy Morphology Classification , http://arxiv.org/abs/2112.14299

    2022 arXiv

  52. [61]

    N., Wild S

    Ćiprijanović A., Lewis A., Pedro K., Madireddy S., Nord B., Perdue G. N., Wild S. M., 2023, DeepAstroUDA : Semi - Supervised Universal Domain Adaptation for Cross - Survey Galaxy Morphology Classification and Anomaly Detection , http://arxiv.org/abs/2302.02005

    2023 arXiv

Pith tools

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