Pith. sign in

REVIEW 3 major objections 6 minor 104 references

DES to HSC: Detecting low surface brightness galaxies in the Abell 194 cluster using transfer learning

T0 review · 3 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

Pith's one-line read Transformer models trained on DES imaging find 171 low-surface-brightness galaxies in deeper HSC data, with 93% recall and no fine-tuning.

desk verdict New cross-survey transfer test for LSBG detection, but the 93% TPR is partly a training-set artifact; still worth refereeing. read the letter →

arxiv 2502.03142 v1 pith:XJWXYP5C submitted 2025-02-05 astro-ph.GA

classification astro-ph.GA
keywords lowsurfacebrightnessgalaxiestransferlearningtransformersAbell194ultra-diffuseHyperSuprime-CamDarkEnergySurveygalaxyclusters
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 asks whether a machine-learning model trained on a shallower survey can be trusted to find low-surface-brightness galaxies in a deeper survey. It takes transformer models trained on Dark Energy Survey images and runs them, without fine-tuning, on Hyper Suprime-Cam images of the Abell 194 cluster, which go about two magnitudes deeper. After converting both surveys' images to common surface-brightness units, the transformer ensembles recover 159 of 171 LSBGs, a 93% true positive rate, with 12 more found by visual inspection. The resulting catalogue, the largest for Abell 194 to date, contains 28 ultra-diffuse galaxies, and its UDG counts and radial distribution support a near-linear scaling with cluster mass. The paper concludes that transfer learning across survey depth is feasible with proper normalization, opening the door to applying existing trained models to upcoming deep surveys.

What carries the argument

The load-bearing mechanism is the pixel-level surface-brightness normalization: DES and HSC images are both converted to µJy arcsec$^{-2}$ before being resized into 64$\times$64 pixel cutouts, so the models see equivalent pixel statistics despite different zero points, pixel scales, and PSFs. The second piece is the transformer ensemble itself: four LSBG Detection Transformers and four LSBG Vision Transformers, each tuned with different hyperparameters, whose averaged probabilities classify each cutout as LSBG or contaminant at a 0.5 threshold. The pipeline closes with GALFIT single-component Sérsic fits on masked, locally sky-subtracted images, which re-derive $r_{\rm eff,g}$ and $\mu_{\rm eff,g}$ to enforce the sample definition and reject poor fits and contaminants.

What would settle it

A blind, independent census of the same HSC field—carried out by human annotators or a separately trained model without access to the DES catalogue—that counts a substantially different number of LSBGs would falsify the 93% claim. Spectroscopic redshifts for all 171 candidates, showing many are not at the cluster redshift, would falsify the cluster-member assumption and the UDG scaling conclusion.

Watch

Extended reading notes

Core claim

The central discovery is that transformer models trained exclusively on DES imaging can be transferred to HSC data of a different depth and resolution, achieving a true positive rate of 93% in identifying LSBGs without any fine-tuning, provided the inputs are standardized to pixel-level surface brightness. The paper further claims that this process yields a sample of 171 LSBGs in the Abell 194 cluster (87 new), of which 28 meet the UDG criteria. The UDG abundance and mass-normalized radial density profile agree with a near-linear log-scale relation between UDG count and cluster halo mass, and the UDGs lie at the extended diffuse end of the dwarf-galaxy size-luminosity plane, suggesting they are part of a continuous dwarf population rather than a distinct class.

Load-bearing premise

The reported 93% true positive rate is measured against a final 171-galaxy sample that the authors themselves assembled using the same fitting and inspection pipeline, and which they treat as complete without confirming membership for every galaxy.

Editorial extensions

If this is right

  • If the transfer result holds, models trained on existing DES data can be applied directly to deeper surveys such as LSST and Euclid, avoiding the need to assemble new large training sets for each survey.
  • The 171-object catalogue roughly doubles the known LSBG population of Abell 194 and provides a larger statistical base for studying how cluster environment shapes faint galaxies.
  • The measured UDG abundance in Abell 194 sits on the literature $N_{\rm UDG}$–$M_{200}$ relation, suggesting that UDG counts can serve as a halo-mass tracer once selection effects are controlled.
  • The overlap of UDGs with dwarf galaxies in the size-luminosity plane argues against UDGs being a discrete population, which should guide simulations of dwarf galaxy evolution.
  • The identified failure modes—underrepresentation of faint LSBGs in training and confusion near bright galaxies—point to specific data-augmentation and sample-balancing strategies for future models.

Reading between the lines

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

  • A natural next test is a fully blind independent search of the same HSC field; if the human or algorithmic census differs by more than the reported 12 missed objects, the 93% TPR would need revising.
  • The paper does not recalibrate the UDG–halo-mass relation despite its Abell 194 point lying above the Karunakaran & Zaritsky prediction; a homogeneous reanalysis across many clusters with matched UDG selection would determine whether the offset is real.
  • Surface-brightness normalization alone does not correct for PSF or depth differences; applying this transfer approach to surveys with substantially different seeing (e.g., space vs ground) may require an additional PSF-homogenisation step.
  • The central FUV$-$NUV colour gradient is based on only 15$-$20 detections; stacking or deeper far-UV imaging would test whether the quenching trend is a physical signal or a selection artefact.
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 / 6 minor

Summary. The paper trains LSBG DETR and LSBG ViT ensembles on DES DR1 data from the Tanoglidis et al. (2021b) and Thuruthipilly et al. (2024b) catalogues, standardizes DES and HSC images to surface-brightness units, applies the ensembles without fine-tuning to SExtractor-selected candidates in deep HSC observations of Abell 194, and then uses GALFIT Sérsic fits and visual inspection to produce 171 LSBGs, including 87 new discoveries and 28 UDGs. The paper reports a 93% true positive rate on HSC without fine-tuning, compares the new sample with previous DES and Zaritsky catalogues, and uses the UDG counts and radial densities to argue for a cluster-mass scaling and for UDGs as an extended dwarf-galaxy population.

Significance. If the transfer claim survives a clean evaluation, the paper would be a useful demonstration that surface-brightness-standardized transformer models can be applied across surveys without fine-tuning, with direct implications for LSST and Euclid. The compiled sample of 171 LSBGs (28 UDGs) with GALFIT parameters and GALEX photometry is a useful resource, and the comparison with previous DES catalogues highlights concrete improvements in masking and local sky subtraction. The code and training data are public, which makes the evaluation issue checkable and fixable. The astrophysical conclusions about UDG abundance and radial density are plausible but are contingent on cluster membership and on the transfer evaluation being clean.

major comments (3)
  1. [§2.1, §4.1, §5.1] The training set in §2.1 is built from the combined Tanoglidis et al. (2021b) and Thuruthipilly et al. (2024b) DES LSBG catalogue without any stated exclusion of Abell 194 sources, and §5.1 reports that 84 objects in the final HSC sample come from exactly that catalogue. Since 18,532 of 27,873 LSBGs were randomly drawn for training, roughly two-thirds of those 84 objects are expected to have been in the training set. The 93% TPR quoted in §4.1 is therefore largely a measure of cross-survey consistency for galaxies the model has already seen in DES, not of generalization to unseen LSBGs. I request that the authors retrain with all Abell 194 sources excluded, or otherwise demonstrate that the overlap is negligible, and that they report the TPR separately for the 87 HSC-only sources.
  2. [§4.1, §4.2] The TPR of 159/171 is evaluated against a 'true' sample whose 87 new objects are labelled by the authors' own visual inspection, and no FPR or confusion matrix for the HSC field is reported. From the numbers in §4.1, roughly 113 of the 272 ML-selected candidates were rejected as non-LSBGs, which implies a non-negligible false-positive rate that should be stated explicitly; a high TPR alone is not sufficient to establish successful transfer because it can be achieved by classifying everything as an LSBG. I request the full 2x2 confusion matrix on the HSC candidate set, with TPR and FPR, or an independent validation set for the HSC-only sources.
  3. [§4.1, §5.2.1, §5.2.2] Section 4.1 treats all 171 LSBGs as Abell 194 members even though 12 lie beyond R200 and none of the 87 new sources has a spectroscopic redshift, and this assumption is carried into the UDG abundance comparison with Karunakaran & Zaritsky (2023) in §5.2.1 and the mass-normalized radial density profile in §5.2.2. Background or foreground interlopers would inflate the UDG count and alter the radial distribution, so the support for a linear N_UDG-M200 relation and for the normalized density profile is not established by these data. I request a membership-restricted analysis, using available redshifts and a statistical background correction, or at least an explicit quantitative assessment of how plausible interloper fractions change the conclusions.
minor comments (6)
  1. [Figures 10 and 12] The x-axis labels in Figure 10 ('10 1 100') and the axis labels in Figure 12 ('102 103 104') appear to be missing superscript formatting and should be typeset as powers of ten.
  2. [Figure 14 and §5.3.1] The labels and running-median legend use 'LSBS' in several places; this should be 'LSBG'.
  3. [Figure 13] The caption cites 'Zaritsky +21' while the text and reference list use 2023; please make the citation consistent.
  4. [§5.3.2] There is a typo 'Fig, 16' and an odd spacing in the section header 'T rends in color'; these should be corrected in proofreading.
  5. [Table B.1] The catalogue table would be easier to use if it included a flag for sources beyond R200 and for the 12 DES-sample objects reclassified as non-LSBGs, since the text discusses these subsets.
  6. [§3.6] The visual inspection is described as independent and performed by two authors, but no inter-rater agreement statistic or explicit rule for resolving disagreements is given; reporting the disagreement rate would improve reproducibility.

Circularity Check

1 steps flagged · score 6.0 of 10

The 93% transfer TPR is partly circular: the HSC evaluation sample includes 84 DES-catalogue LSBGs, about two-thirds of which are expected to be in the training set, so the headline metric largely measures recognition of already-seen galaxies rather than generalization to the 87 new LSBGs.

  1. other [Sec. 2.1 (training data) and Secs. 4.1/5.1 (TPR and comparison sample)]
    "… we used the labelled dataset of LSBGs and contaminants identified from DES by Tanoglidis et al. (2021b) and extended by Thuruthipilly et al. (2024b). … we randomly selected 18 532 LSBGs (comparable to the number of contaminants) from the extended sample of 27 873 LSBGs … In total, Tanoglidis et al. (2021b) and Thuruthipilly et al. (2024b) identified 96 LSBGs in Abell 194, which we call the LSBGs-DES sample. … 95 out of 96 from the LSBGs-DES sample were re-identified by the ensemble model. … the final catalogue contains only 84 LSBGs from the LSBGs-DES sample."

    The evaluation set for the headline 93% TPR is not independent of the training set. The training positives are randomly drawn from the same extended DES catalogue that defines the LSBGs-DES sample, and no Abell 194 exclusion is stated; with 18,532 of 27,873 objects selected, about two-thirds of the 84 DES-known LSBGs in the final HSC sample are expected to be training objects. The aggregate TPR = 159/171 = (83 DES-known + 76 new)/171 ≈ 93% is a weighted average in which the DES-known subset (TPR ≈ 98.8%) measures recognition of galaxies whose DES cutouts the model was trained on, while the 87 genuinely new HSC LSBGs—the actual transfer test—are only 76/87 ≈ 87% and are not reported separately.

full rationale

This is not an equation-level derivation; the circularity is in the evaluation of the central transfer-learning claim. Section 2.1 draws 18,532 training LSBGs from the 27,873-object DES catalogue of Tanoglidis et al. (2021b) plus Thuruthipilly et al. (2024b), with no stated exclusion of Abell 194. Section 5.1 treats that same catalogue as the LSBGs-DES sample, reports 95/96 re-identified in HSC, and retains 84 in the final sample. Because the training draw is ~66.5%, about 56 of those 84 objects are expected to be training positives. The headline TPR of 159/171 = 93% is a weighted average of ~98.8% on the DES-known subset and ~87.4% on the 87 genuinely new HSC LSBGs; the new-only rate is not given. Recognizing a galaxy whose DES cutout was in the training set when it reappears in deeper HSC data is a cross-survey consistency check, not a demonstration that the model generalizes to unseen LSBGs. Other elements—surface-brightness standardization, Sérsic fitting, and comparison with literature UDG scaling relations—are independent of this issue. The cluster-membership assumption and lack of spectroscopic confirmation are completeness/correctness risks rather than circularity. Score 6 reflects a partial but central circularity in the reported transfer metric, not in the astrophysical catalogue itself.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

No new physical entities are introduced. The free parameters are hand-set selection thresholds rather than fitted physical constants. The assumptions concern cluster membership, transferability after normalization, visual ground truth, external stellar-mass calibration, and Sersic modeling. The most consequential unresolved point is whether the test sources overlap the training catalogue.

free parameters (3)
  • Transformer classification threshold = 0.5
    Chosen by the authors, not tuned on HSC; changing it alters TPR and purity, and no HSC ROC curve is provided.
  • SExtractor preselection cuts = mu_eff,g 24-31; reff,g 2-20 arcsec; q 0.3-1.0
    Hand-chosen following Greco et al. (2018) and Tanoglidis et al. (2021b); these cuts define the candidate pool.
  • GALFIT fitting and rejection bounds = n 0.2-5.0; q 0.3-1.0; reduced chi2 > 3 rejected
    Hand-chosen ranges and cutoffs affect which model-selected objects remain in the final catalogue.
assumptions (5)
  • domain assumption All retained LSBGs are at the Abell 194 redshift (z=0.0178) except sources removed by SDSS cross-match.
    Sec 4.1 removes 14 SDSS interlopers but treats the remaining sources, including 12 outside R200, as cluster members; physical sizes, UDG counts, and masses depend on this.
  • domain assumption Converting pixel counts to surface brightness units makes the DES-trained model input distribution close enough to HSC for transfer; PSF differences are ignored.
    Sec 3.2 notes the standardization does not address PSF differences; if PSF mismatch matters, detection rates and TPR change.
  • domain assumption Visual inspection by two authors and GALFIT convergence provide accurate labels for LSBG versus contaminant.
    Sec 3.6; no spectroscopy; TPR is measured against this visual ground truth.
  • domain assumption Stellar masses and stellar mass surface densities are derived from the Du et al. (2020) mass-to-light versus g-r color relation, appropriate for LSBGs.
    Sec 4.1; log Mstar and log Sigma_star depend on this external calibration.
  • standard math Single-component Sersic profiles and the Graham and Driver (2005) relations in Eqs. (2)-(3) accurately convert fitted m, reff, q, n into surface brightness.
    Sec 3.5; the paper itself notes Sersic fitting may not capture complete galaxy light.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DES to HSC: Detecting low surface brightness galaxies in the Abell 194 cluster using transfer learning." pith.science (2026). https://pith.science/paper/XJWXYP5C

@misc{pith2026250203142,
  author       = {Pith},
  title        = {Pith review of: DES to HSC: Detecting low surface brightness galaxies in the Abell 194 cluster using transfer learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XJWXYP5C}},
  note         = {Machine review of arXiv:2502.03142}
}
abstract

Low surface brightness galaxies (LSBGs) are important for understanding galaxy evolution and cosmological models. The upcoming large-scale surveys are expected to uncover a large number of LSBGs, requiring accurate automated or machine learning-based methods for their detection. We study the scope of transfer learning for the identification of LSBGs. We use transformer models divided into two categories: LSBG Detection Transformer (LSBG DETR) and LSBG Vision Transformer (LSBG ViT), trained on Dark Energy Survey (DES) data, to identify LSBGs from dedicated Hyper Suprime-Cam (HSC) observations of the Abell 194 cluster, which are two magnitudes deeper than DES. The data from DES and HSC were standardized based on pixel-level surface brightness. We used two transformer ensembles to detect LSBGs. This was followed by a single-component S\'ersic model fit and a final visual inspection to filter out potential false positives and improve sample purity. We present a sample of 171 low surface brightness galaxies (LSBGs) in the Abell 194 cluster using HSC data, including 87 new discoveries. Of these, 159 were identified using transformer models, and 12 additional LSBGs were found through visual inspection. The transformer model achieved a true positive rate (TPR) of 93% in HSC data without any fine-tuning. Among the LSBGs, 28 were classified as ultra-diffuse galaxies (UDGs). The number of UDGs and the radial UDG number density suggest a linear relationship between UDG numbers and cluster mass on a log scale. UDGs share similar S\'ersic parameters with dwarf galaxies and occupy the extended end of the $R_{\mathrm{eff}}-M_g$ plane, suggesting they might be an extended subpopulation of dwarf galaxies. We have demonstrated that transformer models trained on shallower surveys can be successfully applied to deeper surveys with appropriate data normalization.

Figures

Figures reproduced from arXiv: 2502.03142 by the authors.

Figure 1
Figure 1. g-band cutouts of four examples of LSBGs (1a) and con￾taminants (1b) used in the training data. Each cutout corresponds to a 40′′ × 40′′ (152 × 152 pixels) region of the sky centred on the LSBG or artefact. Article number, page 3 of 19 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Image of an LSBG (RA = 21.82879◦ , DEC=-1.81381◦ ) as observed in HSC in the g-band, the image with all the objects other than LSBG masked, corresponding Sérsic model fitted by GALFIT and the residual are shown respectively from left to right. centre was set at X = 119 pixels and Y = 119 pixels, which was allowed to vary ±5 pixels (0.84 ′′) in both directions. These con￾ditions for the Sérsic fit were based on Greco… view at source ↗
Figure 3
Figure 3. Schematic diagram showing the sequential selection steps [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: The top panel (4a) shows six examples of LSBGs identified from our study. The bottom panel (4b) shows six examples of UDGs identified from our study. Each cutout of the LSBG and UDG corresponds to a 40′′ × 40′′ (152 × 152 pixels) region of the sky centred around the LS…
Figure 5
Figure 5. Figure 5: The normalized distributions of the g-band magnitude and Sérsic index for the LSBGs and UDGs identified from HSC data in this study are shown in the left (5a) and right (5b) panels, respectively. The long black arrow shows the median of the UDG distribution, and the sh…
Figure 7
Figure 7. Figure 7: Example g-band images of LSBGs missed by the ensem￾ble models. The left panel shows an LSBG, which is fainter than the training sample, and the right panel shows an LSBG near bright galaxies. A total of 101 objects initially classified as LSBGs were re￾classified as no…
Figure 6
Figure 6. Figure 6: Normalized distribution of g-band magnitudes for the LS￾BGs used to train the ensemble models, LSBGs identified by the ensemble models in the HSC data of the Abell 194 cluster, and LSBGs missed by the ensemble models in the HSC data of the Abell 194 cluster. Similarly,…
Figure 9
Figure 9. Figure 9: The number of UDGs as a function of halo mass in [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: The radial surface density profile of cluster UDGs, nor [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 12
Figure 12. Figure 12: The size-luminosity relation for local dwarf ellipticals, [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]
Figure 14
Figure 14. Figure 14: Morphological and physical properties of LSBS and [PITH_FULL_IMAGE:figures/full_fig_p014_14.png]
Figure 15
Figure 15. Figure 15: FUV − NUV (top panel), NUV − r (middle panel) and g−r (bottom panel) colors of the LSBGs presented in this works as a function of cluster centric distance. The number of LSBGs used is also shown in each subplot since not all the LSBGs had 3σ detection in NUV and FUV b…
Figure 16
Figure 16. Figure 16: The NUV − r color is plotted against the absolute mag￾nitude in the r-band. The red region denotes the space occupied by the red-quiescent galaxy population, while the blue region represents the blue galaxy population (Singh et al. 2019). Ma￾genta dots represent LSBGs…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

104 extracted references · 58 canonical work pages

  1. [1]

    Abbott, T. M. C., Abdalla, F. B., Allam, S., et al. 2018, ApJS, 239, 18

  2. [2]

    Abbott, T. M. C., Adamów, M., Aguena, M., et al. 2021, ApJS, 255, 20

  3. [3]

    K., & Turp, M

    Ackermann, S., Schawinski, K., Zhang, C., Weigel, A. K., & Turp, M. D. 2018, MNRAS, 479, 415

  4. [4]

    Aguerri, J. A. L. & González-García, A. C. 2009, A&A, 494, 891

  5. [5]

    2020, ApJS, 249, 3

    Ahumada, R., Allende Prieto, C., Almeida, A., et al. 2020, ApJS, 249, 3

  6. [6]

    2018, PASJ, 70, S8

    Aihara, H., Armstrong, R., Bickerton, S., et al. 2018, PASJ, 70, S8

  7. [7]

    Amorisco, N. C. & Loeb, A. 2016, MNRAS, 459, L51

  8. [8]

    Bautista, J. M. G., Koda, J., Yagi, M., Komiyama, Y ., & Yamanoi, H. 2023, ApJS, 267, 10

Show all 104 references
  1. [9]

    M., Abbott, T

    Bernstein, G. M., Abbott, T. M. C., Armstrong, R., et al. 2018, PASP, 130, 054501

  2. [10]

    2013, PSFEx: Point Spread Function Extractor, Astrophysics Source Code Library, record ascl:1301.001

    Bertin, E. 2013, PSFEx: Point Spread Function Extractor, Astrophysics Source Code Library, record ascl:1301.001

  3. [11]

    & Arnouts, S

    Bertin, E. & Arnouts, S. 1996, A&AS, 117, 393

  4. [12]

    W., Wilkinson, S., Ferreira, L., et al

    Bickley, R. W., Wilkinson, S., Ferreira, L., et al. 2024, The effect of image quality on galaxy merger identification with deep learning

  5. [13]

    1997, PASP, 109, 745

    Bothun, G., Impey, C., & McGaugh, S. 1997, PASP, 109, 745

  6. [14]

    Bullock, J. S. & Boylan-Kolchin, M. 2017, ARA&A, 55, 343

  7. [15]

    S., Kolatt, T

    Bullock, J. S., Kolatt, T. S., Sigad, Y ., et al. 2001, MNRAS, 321, 559

  8. [16]

    2001, AJ, 122, 2318

    Burkholder, V ., Impey, C., & Sprayberry, D. 2001, AJ, 122, 2318

  9. [17]

    A., & Maureira, J.-C

    Cabrera-Vives, G., Reyes, I., Förster, F., Estévez, P. A., & Maureira, J.-C. 2017, ApJ, 836, 97

  10. [18]

    Chamba, N., Trujillo, I., & Knapen, J. H. 2022, A&A, 667, A87

  11. [19]

    C., Magnier, E

    Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2016, arXiv e-prints, arXiv:1612.05560

  12. [20]

    2016, in 4th International Confer- ence on Learning Representations, ICLR 2016, San Juan, Puerto Rico, May 2-4, 2016, Conference Track Proceedings, ed

    Clevert, D., Unterthiner, T., & Hochreiter, S. 2016, in 4th International Confer- ence on Learning Representations, ICLR 2016, San Juan, Puerto Rico, May 2-4, 2016, Conference Track Proceedings, ed. Y . Bengio & Y . LeCun

  13. [21]

    Conselice, C. J. 2018, Research Notes of the American Astronomical Society, 2, 43

  14. [22]

    J., Gallagher, John S., I., & Wyse, R

    Conselice, C. J., Gallagher, John S., I., & Wyse, R. F. G. 2003, AJ, 125, 66 Da Costa, G. S. 1992, in Astronomical Society of the Pacific Conference Series, V ol. 23, Astronomical CCD Observing and Reduction Techniques, ed. S. B. Howell, 90

  15. [23]

    J., Spergel, D

    Dalcanton, J. J., Spergel, D. N., Gunn, J. E., Schmidt, M., & Schneider, D. P. 1997, AJ, 114, 635 de Blok, W. J. G., McGaugh, S. S., Bosma, A., & Rubin, V . C. 2001, ApJ, 552, L23

  16. [24]

    2009, in 2009 IEEE Conference on Com- puter Vision and Pattern Recognition, 248–255

    Deng, J., Dong, W., Socher, R., et al. 2009, in 2009 IEEE Conference on Com- puter Vision and Pattern Recognition, 248–255

  17. [25]

    J., Lang, D., et al

    Dey, A., Schlegel, D. J., Lang, D., et al. 2019, AJ, 157, 168 Di Cintio, A., Brook, C. B., Dutton, A. A., et al. 2017, MNRAS, 466, L1 Domínguez Sánchez, H., Huertas-Company, M., Bernardi, M., et al. 2019, MN- RAS, 484, 93

  18. [26]

    Driver, S. P. 1999, ApJ, 526, L69

  19. [27]

    2020, The Astronomical Journal, 159, 138

    Du, W., Cheng, C., Zheng, Z., & Wu, H. 2020, The Astronomical Journal, 159, 138

  20. [28]

    H., Taylor, M

    Eigenthaler, P., Puzia, T. H., Taylor, M. A., et al. 2018, ApJ, 855, 142

  21. [29]

    A., et al

    Ferrarese, L., Côté, P., MacArthur, L. A., et al. 2020, ApJ, 890, 128

  22. [30]

    Fitzpatrick, E. L. 1999, PASP, 111, 63

  23. [31]

    A., Dullo, B

    Forbes, D. A., Dullo, B. T., Gannon, J., et al. 2020, MNRAS, 494, 5293

  24. [32]

    2018, PASJ, 70, S3

    Furusawa, H., Koike, M., Takata, T., et al. 2018, PASJ, 70, S3

  25. [33]

    1998, ApJ, 505, 74

    Girardi, M., Giuricin, G., Mardirossian, F., Mezzetti, M., & Boschin, W. 1998, ApJ, 505, 74

  26. [34]

    & Bengio, Y

    Glorot, X. & Bengio, Y . 2010, in JMLR Proceedings, V ol. 9, Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, AISTATS 2010, Chia Laguna Resort, Sardinia, Italy, May 13-15, 2010, ed. Y . W. Teh & D. M. Titterington (JMLR.org), 249–256

  27. [35]

    Graham, A. W. & Driver, S. P. 2005, PASA, 22, 118

  28. [36]

    P., Greene, J

    Greco, J. P., Greene, J. E., Strauss, M. A., et al. 2018, ApJ, 857, 104

  29. [37]

    2024, A&A, 688, A34

    Grespan, M., Thuruthipilly, H., Pollo, A., et al. 2024, A&A, 688, A34

  30. [38]

    J., & Dettmar, R

    Haberzettl, L., Bomans, D. J., & Dettmar, R. J. 2007, A&A, 471, 787

  31. [39]

    C., Lee, J

    Hannon, S., Whitmore, B. C., Lee, J. C., et al. 2023, MNRAS, 526, 2991

  32. [40]

    2013, ApJS, 208, 19

    Hinshaw, G., Larson, D., Komatsu, E., et al. 2013, ApJS, 208, 19

  33. [41]

    C.-F., Chang, P.-W., et al

    Huang, K.-W., Chen, G. C.-F., Chang, P.-W., et al. 2023, in Computer Vision – ECCV 2022 Workshops, ed. L. Karlinsky, T. Michaeli, & K. Nishino (Cham: Springer Nature Switzerland), 143–153

  34. [42]

    Y ., Sabiu, C

    Hwang, S. Y ., Sabiu, C. G., Park, I., & Hong, S. E. 2023, J. Cosmology Astropart. Phys., 2023, 075

  35. [43]

    N., et al

    Ivezic, Z., Axelrod, T., Brandt, W. N., et al. 2008, Serbian Astronomical Journal, 176, 1 Ivezi´c, Ž., Kahn, S. M., Tyson, J. A., et al. 2019, ApJ, 873, 111 Article number, page 16 of 19 H. Thuruthipilly et al.: DES to HSC with transfer learning

  36. [44]

    R., Abraham, R., Brodie, J., Forbes, D

    Janssens, S. R., Abraham, R., Brodie, J., Forbes, D. A., & Romanowsky, A. J. 2019, ApJ, 887, 92

  37. [45]

    2023, AJ, 165, 26

    Jia, P., Sun, R., Li, N., et al. 2023, AJ, 165, 26

  38. [46]

    2019, MNRAS, 487, 5272

    Jiang, F., Dekel, A., Freundlich, J., et al. 2019, MNRAS, 487, 5272

  39. [47]

    2022, A&A, 667, A76

    Junais, Boissier, S., Boselli, A., et al. 2022, A&A, 667, A76

  40. [48]

    M., Epinat, B., et al

    Junais, Weilbacher, P. M., Epinat, B., et al. 2024, A&A, 681, A100 Juri´c, M., Kantor, J., Lim, K. T., et al. 2017, in Astronomical Society of the Pacific Conference Series, V ol. 512, Astronomical Data Analysis Software and Systems XXV , ed. N. P. F. Lorente, K. Shortridge, &...

  41. [49]

    & Zaritsky, D

    Karunakaran, A. & Zaritsky, D. 2023, MNRAS, 519, 884

  42. [50]

    2018, PASJ, 70, 66

    Kawanomoto, S., Uraguchi, F., Komiyama, Y ., et al. 2018, PASJ, 70, 66

  43. [51]

    Kingma, D. P. & Ba, J. 2015, in 3rd International Conference on Learning Rep- resentations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, ed. Y . Bengio & Y . LeCun

  44. [52]

    2015, ApJ, 807, L2

    Koda, J., Yagi, M., Yamanoi, H., & Komiyama, Y . 2015, ApJ, 807, L2

  45. [53]

    2018, PASJ, 70, S2

    Komiyama, Y ., Obuchi, Y ., Nakaya, H., et al. 2018, PASJ, 70, S2

  46. [54]

    1985, ApJ, 295, 73

    Kormendy, J. 1985, ApJ, 295, 73

  47. [55]

    B., Cornell, M

    Kormendy, J., Fisher, D. B., Cornell, M. E., & Bender, R. 2009, ApJS, 182, 216 La Marca, A., Iodice, E., Cantiello, M., et al. 2022a, A&A, 665, A105 La Marca, A., Iodice, E., Cantiello, M., et al. 2022b, A&A, 665, A105

  48. [56]

    & Salzano, V

    Laudato, E. & Salzano, V . 2023, European Physical Journal C, 83, 402

  49. [57]

    2011, arXiv e-prints, arXiv:1110.3193

    Laureijs, R., Amiaux, J., Arduini, S., et al. 2011, arXiv e-prints, arXiv:1110.3193

  50. [58]

    E., et al

    Lazar, I., Kaviraj, S., Watkins, A. E., et al. 2024, MNRAS, 529, 499

  51. [59]

    H., Hodges-Kluck, E., & Gallo, E

    Lee, C. H., Hodges-Kluck, E., & Gallo, E. 2020, MNRAS, 497, 2759

  52. [60]

    W., et al

    Lim, S., Côté, P., Peng, E. W., et al. 2020, ApJ, 899, 69

  53. [61]

    A., Schlafly, E., Finkbeiner, D., et al

    Magnier, E. A., Schlafly, E., Finkbeiner, D., et al. 2013, ApJS, 205, 20 Mancera Piña, P. E., Aguerri, J. A. L., Peletier, R. F., et al. 2019, MNRAS, 485, 1036 Mancera Piña, P. E., Peletier, R. F., Aguerri, J. A. L., et al. 2018, MNRAS, 481, 4381

  54. [62]

    R., Habas, R., Poulain, M., et al

    Marleau, F. R., Habas, R., Poulain, M., et al. 2021, A&A, 654, A105

  55. [63]

    C., Fanson, J., Schiminovich, D., et al

    Martin, D. C., Fanson, J., Schiminovich, D., et al. 2005, ApJ, 619, L1

  56. [64]

    2019, MNRAS, 485, 796

    Martin, G., Kaviraj, S., Laigle, C., et al. 2019, MNRAS, 485, 796

  57. [65]

    McGaugh, S. S. 1996, MNRAS, 280, 337

  58. [66]

    McGaugh, S. S. & Bothun, G. D. 1994, AJ, 107, 530

  59. [67]

    2016, ApJ, 833, 168

    Merritt, A., van Dokkum, P., Danieli, S., et al. 2016, ApJ, 833, 168

  60. [68]

    C., Durrell, P

    Mihos, J. C., Durrell, P. R., Ferrarese, L., et al. 2015, ApJ, 809, L21

  61. [69]

    F., Disney, M

    Minchin, R. F., Disney, M. J., Parker, Q. A., et al. 2004, MNRAS, 355, 1303

  62. [70]

    2018, PASJ, 70, S1

    Miyazaki, S., Komiyama, Y ., Kawanomoto, S., et al. 2018, PASJ, 70, S1

  63. [71]

    2024, A&A, 681, A15

    Montes, M., Trujillo, I., Karunakaran, A., et al. 2024, A&A, 681, A15

  64. [72]

    1999, ApJ, 524, L19

    Moore, B., Ghigna, S., Governato, F., et al. 1999, ApJ, 524, L19

  65. [73]

    1996, Nature, 379, 613

    Moore, B., Katz, N., Lake, G., Dressler, A., & Oemler, A. 1996, Nature, 379, 613

  66. [74]

    A., Menanteau, F., et al

    Morganson, E., Gruendl, R. A., Menanteau, F., et al. 2018, PASP, 130, 074501

  67. [75]

    A., et al

    Morrissey, P., Conrow, T., Barlow, T. A., et al. 2007, ApJS, 173, 682 Müller, O., Jerjen, H., & Binggeli, B. 2018, A&A, 615, A105 O’Neil, K. & Bothun, G. 2000, ApJ, 529, 811

  68. [76]

    2023, ApJS, 265, 57

    Paudel, S., Yoon, S.-J., Yoo, J., et al. 2023, ApJS, 265, 57

  69. [77]

    J., Suelves, L

    Pearson, W. J., Suelves, L. E., Ho, S. C. C., et al. 2022, A&A, 661, A52

  70. [78]

    Y ., Ho, L

    Peng, C. Y ., Ho, L. C., Impey, C. D., & Rix, H.-W. 2010, AJ, 139, 2097 Pérez-Carrasco, M., Cabrera-Vives, G., Martinez-Marin, M., et al. 2019, PASP, 131, 108002

  71. [79]

    R., Habas, R., et al

    Poulain, M., Marleau, F. R., Habas, R., et al. 2021, MNRAS, 506, 5494

  72. [80]

    J., Kurtz, M

    Rines, K., Geller, M. J., Kurtz, M. J., & Diaferio, A. 2003, AJ, 126, 2152 Román, J. & Trujillo, I. 2017, MNRAS, 468, 4039 Román, J., Trujillo, I., & Montes, M. 2020, A&A, 644, A42

  73. [81]

    S., Chilingarian, I

    Saburova, A. S., Chilingarian, I. V ., Kasparova, A. V ., et al. 2021, MNRAS, 503, 830

  74. [82]

    S., Chilingarian, I

    Saburova, A. S., Chilingarian, I. V ., Kulier, A., et al. 2023, MNRAS, 520, L85

  75. [83]

    V ., Navarro, J

    Sales, L. V ., Navarro, J. F., Peñafiel, L., et al. 2020, MNRAS, 494, 1848

  76. [84]

    & Binggeli, B

    Sandage, A. & Binggeli, B. 1984, AJ, 89, 919

  77. [85]

    Schlafly, E. F. & Finkbeiner, D. P. 2011, ApJ, 737, 103

  78. [86]

    J., Finkbeiner, D

    Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525

  79. [87]

    R., Zaritsky, D., Donnerstein, R., & Spekkens, K

    Singh, P. R., Zaritsky, D., Donnerstein, R., & Spekkens, K. 2019, AJ, 157, 212

  80. [88]

    J., Greene, J

    Somalwar, J. J., Greene, J. E., Greco, J. P., et al. 2020, ApJ, 902, 45

  81. [89]

    D., Bothun, G

    Sprayberry, D., Impey, C. D., Bothun, G. D., & Irwin, M. J. 1995, AJ, 109, 558

  82. [90]

    2024, MNRAS, 528, 873

    Su, H., Yi, Z., Liang, Z., et al. 2024, MNRAS, 528, 873

  83. [91]

    2016, A&A, 588, A14

    Tempel, E., Kipper, R., Tamm, A., et al. 2016, A&A, 588, A14

  84. [92]

    2022, A&A, 664, A4 van der Burg, R

    Thuruthipilly, H., Zadrozny, A., Pollo, A., & Biesiada, M. 2022, A&A, 664, A4 van der Burg, R. F. J., Hoekstra, H., Muzzin, A., et al. 2017, A&A, 607, A79 van der Burg, R. F. J., Muzzin, A., & Hoekstra, H. 2016, A&A, 590, A20 van Dokkum, P. G., Abraham, R., Merritt, A., et al....

  85. [93]

    Vaswani, A., Shazeer, N., Parmar, N., et al. 2017, in Advances in Neural In- formation Processing Systems 30: Annual Conference on Neural Informa- tion Processing Systems 2017, December 4-9, 2017, Long Beach, CA, USA, 5998–6008

  86. [94]

    2018, A&A, 620, A165

    Venhola, A., Peletier, R., Laurikainen, E., et al. 2018, A&A, 620, A165

  87. [95]

    2019, A&A, 625, A143

    Venhola, A., Peletier, R., Laurikainen, E., et al. 2019, A&A, 625, A143

  88. [96]

    F., Salo, H., et al

    Venhola, A., Peletier, R. F., Salo, H., et al. 2022, A&A, 662, A43

  89. [97]

    W., Liu, C., et al

    Wang, K., Peng, E. W., Liu, C., et al. 2023, Nature, 623, 296

  90. [98]

    A., Whitmore, B

    Wei, W., Huerta, E. A., Whitmore, B. C., et al. 2020, MNRAS, 493, 3178

  91. [99]

    2023, ApJS, 269, 59

    Xing, Y ., Yi, Z., Liang, Z., et al. 2023, ApJS, 269, 59

  92. [100]

    2016, ApJS, 225, 11

    Yagi, M., Koda, J., Komiyama, Y ., & Yamanoi, H. 2016, ApJS, 225, 11

  93. [101]

    2022, MNRAS, 513, 3972

    Yi, Z., Li, J., Du, W., et al. 2022, MNRAS, 513, 3972

  94. [102]

    2014, in Advances in Neural Information Processing Systems, ed

    Yosinski, J., Clune, J., Bengio, Y ., & Lipson, H. 2014, in Advances in Neural Information Processing Systems, ed. Z. Ghahramani, M. Welling, C. Cortes, N. Lawrence, & K. Weinberger, V ol. 27 (Curran Associates, Inc.)

  95. [103]

    2023, ApJS, 267, 27

    Zaritsky, D., Donnerstein, R., Dey, A., et al. 2023, ApJS, 267, 27

  96. [104]

    H., Liang, Y

    Zhong, G. H., Liang, Y . C., Liu, F. S., et al. 2008, MNRAS, 391, 986 Zöller, R., Kluge, M., Staiger, B., & Bender, R. 2024, ApJS, 271, 52 Article number, page 17 of 19 A&A proofs: manuscript no. aanda Appendix A: Transformer models Appendix A.1: Training All of the LSBG DETR ...

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.