Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Galaxy Morphology in CANDELS: Addressing Evolutionary Changes Across $0.2 \leq z \leq 2.4$ with Hybrid Classification Approach

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

Pith's one-line read A hybrid metric-and-CNN classifier finds a flat disk fraction from z=0.2 to z=2.4.

desk verdict A headline claim that would reshape the disk-fraction narrative, but the labeling loop and hand-set thresholds leave it under-supported; still worth a referee's time. read the letter →

arxiv 2412.03778 v2 pith:KBIPPWNZ submitted 2024-12-04 astro-ph.GA

classification astro-ph.GA
keywords galaxymorphologyCANDELShybridclassificationself-organizingmapsconvolutionalneuralnetworksvisualbiasdiskfractionhigh-redshiftgalaxies
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 classify roughly 14,000 CANDELS galaxies between redshift 0.2 and 2.4 into disks, spheroids, and irregulars without any step of visual inspection. Using redshift-bin-specific models trained on labels produced by an unsupervised clustering of non-parametric morphological metrics, it finds that the fraction of disks stays near 60 percent and spheroids near 30 percent across 8.5 billion years of cosmic time. This directly contradicts earlier work reporting that disk galaxies become rare beyond z~1. The paper argues that those earlier results were biased by visual classification, which loses disk features to surface-brightness dimming and shrinking angular size at high redshift, and it reports close agreement with a JWST-based unsupervised study. If the claim holds, the epoch of disk formation is earlier and the morphological evolution of galaxies over this interval is much weaker than previously thought.

What carries the argument

The argument is carried by a hybrid pipeline. First, four non-parametric metrics - the second moment of light $M_{20}$, entropy, Gini coefficient, and gradient pattern asymmetry $G_2$ - are measured in F814W cutouts within a Petrosian ellipse. A Self-Organizing Map clusters these metric vectors, and an IsoData threshold per metric selects 'prominent clusters' that become the disk and spheroid training labels, with no human labeling. For each redshift bin of width 0.2, an ensemble of one hundred convolutional networks is trained on these labels, and the averaged output probability assigns final classes: below 0.1 disk, above 0.9 spheroid, between 0.2 and 0.8 irregular. The FERENGI code is then used to artificially redshift low-redshift galaxies to test how flux dimming and angular-size changes alter the same metrics and classifications.

What would settle it

Take the same CANDELS galaxies and measure the same MEGG metrics on rest-frame near-infrared JWST images; if at z>1 the F814W-based disks are classified as spheroids or irregulars by the rest-frame metrics in large numbers, or if the rest-frame disk fraction falls with redshift, the constant-fraction claim collapses. A simpler check: degrade local disks with FERENGI and have experienced astronomers classify them by eye; if disks remain recognizable, the claim that visual inspection is systematically biased is weakened.

Watch

Extended reading notes

Core claim

The central discovery is that, once human visual judgment is removed from labeling, the global morphology mix of massive galaxies is remarkably stable: disks constitute roughly 60 percent, spheroids roughly 30 percent, and irregulars roughly 10 percent over 0.2<z<2.4, with the fitted slope for disk fraction $m_{\rm disk}=0.08\pm0.03$ consistent with a flat trend. The paper further finds that a single classifier trained on the full redshift range disagrees with bin-dedicated classifiers for about 25 percent of galaxies, mostly above z~1, and that simulated cosmological degradation converts up to about 18 percent of disk galaxies into apparent spheroids or irregulars. It attributes the declining disk fractions reported by visual-classification studies to these effects, and points to the close match with an unsupervised JWST analysis as independent support. In the mass-resolved sample, the fraction of massive spheroids ($M_{\rm stellar}\geq10^{10.5}\,M_\odot$) rises by about 40 percent toward lower redshift while the massive disk fraction falls by about 20 percent, a complementarity the paper reads as evidence that merging massive disks builds spheroids.

Load-bearing premise

The load-bearing premise is that the prominent SOM clusters, built from metrics measured in the single F814W band, are the true disk and spheroid populations at every redshift, and that CNN probabilities between 0.2 and 0.8 faithfully mark irregulars.

Editorial extensions

If this is right

  • Earlier visual-based catalogs that report a decreasing disk fraction beyond z~1 should be re-examined; under this method the disk fraction is flat to z=2.4.
  • Supervised classifiers trained on visually labeled high-redshift samples inherit a bias that this method avoids by deriving labels from metrics alone.
  • Morphology models that cover wide redshift ranges in one shot mislabel about a quarter of galaxies compared to redshift-bin-specific models; future surveys should bin in redshift.
  • Corrections for surface-brightness dimming and angular-size degradation are needed before comparing fractions across redshift; up to 18 percent of disks can be lost to apparent spheroids or irregulars.
  • Massive disk and spheroid fractions change in opposite directions, implying mergers transform massive disks into spheroids, while the overall mix stays constant because low-mass galaxies remain disk-dominated.

Reading between the lines

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

  • If the flat global fractions are real, then the widely reported 'disk decline' is largely a selection and methodology effect; a direct consequence is that disk assembly was already complete by z~2 for galaxies above $10^9\,M_\odot$, which is a stronger statement than the paper explicitly makes.
  • The irregular class, defined as CNN probability between 0.2 and 0.8, is a method-internal category; part of its mild increase at z>1 could be degraded disks rather than genuinely disturbed systems, as the FERENGI experiment hints.
  • A testable extension would be to apply the same metric-plus-SOM pipeline to rest-frame near-infrared JWST images of the same CANDELS fields; agreement would strengthen the claim, while disagreement would reveal a band-dependent bias in the F814W labels.
  • Since the method uses only one band, the agreement with JWST results suggests wavelength-dependent metric variation is modest; this could be quantified by measuring MEGG metrics in multiple bands for the same galaxies at fixed rest-frame wavelength.
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. This paper presents a hybrid unsupervised-supervised classification of ~14,000 CANDELS galaxies at 0.2<z<2.4, using MEGG non-parametric metrics, SOM-based labeling, and Xception CNNs with hundred-model ensembles per redshift bin. The authors define disk, spheroid, and irregular classes via CNN probability thresholds and report a nearly constant disk fraction (~60%) and spheroid fraction (~30%) over the full redshift range, attributing the decline seen in earlier work to visual classification bias. They also use FERENGI to simulate redshifted local galaxies and claim that up to 18% of disks can be misclassified by degradation, and they interpret mass-dependent trends as evidence for merger-driven spheroid formation.

Significance. If the constant-fraction result is correct, it would substantially revise the picture of morphological evolution at z<2.4 and strengthen the case against visual inspection as a ground truth for high-redshift morphology. The paper has real strengths: bin-dedicated training with ensembles, explicit investigation of model transfer across redshift, and a quantitative degradation analysis. However, the central claim rests on a labeling chain in which the same metric distributions used to define labels also define the classes, and the degradation test is circular; without an independent ground truth, the reported fractions cannot be distinguished from a re-expression of the metric thresholds.

major comments (4)
  1. [Sections 3.2 and 3.5] The disk/spheroid labels are generated by selecting 'prominent clusters' of the MEGG metrics (M20, E, Gini, G2) with SOMbrero and IsoData, and the irregular class is defined entirely by the CNN probability interval 0.2-0.8. Consequently the reported fractions are not an independent measurement of morphology but a classification of the same metric distributions used to build the labels. The central claim of a constant ~60% disk fraction requires external validation, e.g., against simulations with known intrinsic morphology or against a separate high-resolution data set, before it can support the conclusion that visual classifications are biased.
  2. [Section 4.2 and Figure 9] The FERENGI degradation test is circular. The bin-dedicated CNNs used to classify the artificially redshifted low-z disks/spheroids were trained on real high-z galaxies whose labels were produced by the same SOM/metric pipeline. High retention rates (92-95%) therefore only show that the pipeline reproduces its own labels under degradation; they do not demonstrate that the constant disk fraction is unbiased. I would like to see the degradation test repeated with models trained on labels from independent sources, such as hydrodynamical simulations, or at least a demonstration that the SOM labels at high z are not dominated by resolution effects.
  3. [Section 4.2, Eq. (8) and surrounding text] The luminosity evolution term is fixed to Mz' = Mz0 - (1 x z') without justification, and the paper states that establishing a more accurate form is beyond scope. Since the claimed 18% disk misclassification depends on this term, the analysis should include a sensitivity test over a plausible range of evolution parameters; otherwise the degradation correction is itself an uncontrolled free parameter.
  4. [Section 5, Figure 9 and Section 6] There is an internal inconsistency in the degradation statistics. The text reports mean retention fractions of 92% (spheroids) and 95% (disks), yet also states 'observed decrease of about 15% on average'; a 92-95% retention corresponds to a 5-8% decrease, not 15%. The abstract's 'up to 18%' also needs to be reconciled with the mean values in Figure 9 and with the actual maximum per redshift bin, and the summary bullet in Section 6 says 'up to 17%' rather than 18%.
minor comments (5)
  1. [Section 4.1 and Figure 6] The quantity called 'F1 score' is defined as the fraction of galaxies that do not change class; this is not the standard F1 metric, which is the harmonic mean of precision and recall. Please rename the quantity to 'agreement fraction' or compute the actual F1 score.
  2. [Section 2.2 and Table 3] The notation M2O and M20 is used interchangeably; the text introduces M2O and Eq. (2) defines M20, and Table 3 uses M20. Please unify the notation throughout.
  3. [Throughout] There are several typos and grammatical slips, including 'we adopt, we adopt' in Section 1, 'It is notable that we we tend to agree' in the Figure 10 caption, 'Irregural Fraction' in Figure 9, and 'Figure 4 exhibit' in Section 3.3. A careful proofread is needed.
  4. [Section 6 and Section 2.1] The paper states that 14,736 galaxies remain after cleaning in Section 2.1 but Section 6 reports classifying 13,988 galaxies; please clarify the relation between these numbers, e.g., removal of 'Unclassifiable' objects and any other cuts.
  5. [Data Availability] The data availability statement says the data are available from the corresponding author upon reasonable request; for reproducibility of the central claim, please release the trained models, the morphological labels, and the code used to generate the fractions.

Circularity Check

3 steps flagged · score 6.0 of 10

Partial circularity: the irregular fraction is defined by classifier uncertainty, and the FERENGI degradation 'validation' uses models trained on labels from the same metric/SOM pipeline, so two supporting results reduce to self-consistency.

  1. self definitional [Section 3.5 (CNN Confidence Thresholding), Figure 5; applied in Section 5 and Table 1]
    "We also define a sample of irregular galaxies, characterized by a probability between 0.2 and 0.8. Nevertheless, the results touching irregular galaxies should be treated with caution, as we here assume that the inability of our method in classifying a system as disk or spheroidal is an indication that the system presents irregular features compatible with irregular galaxies."

    The 'irregular' class is defined as the classifier's indecision region (CNN probability in [0.2,0.8]). The redshift-dependent irregular fraction reported in Section 5 ("our fitting results suggest that the fraction of irregular galaxies increases with redshift, doubling its initial value after a Δz∼1.57") is therefore a property of the thresholded probability distribution, not of an independently defined physical population. The paper's own caveat 'we here assume' makes the result a definitional re-expression of model confidence rather than an empirical morphological measurement.

  2. other [Section 4.2 (Degradation and evolution of disks observed at high redshifts), Figure 9; training labels from Sections 3.2-3.4]
    "We take all galaxies in the redshift bin 0.2 < 𝑧 <0.4 classified as spheroidal or disks and artificially degrade them to simulate observations at higher redshift bins from 0.4to2.4.Wethenclassifythedegradedgalaxiesusingmodelsdedicated to each specific redshift bin, constructed from real data—that is, we use models trained on data from Figure 2."

    The bin-dedicated CNN models were trained on labels produced by the same SOM/IsoData 'prominent cluster' analysis of MEGG metrics (Section 3.2: 'only galaxies consistently assigned to the same supercluster in all runs are preserved and assigned the respective label'; Section 3.4: 'We use the labeled data from SOMbrero as the training set'). A degraded galaxy therefore 'maintains its class' when the CNN agrees with the original metric-cluster label. A systematic bias in the metric/SOM labels at high z is invisible to this test, so the reported ~92-95% retention demonstrates self-consistency of the pipeline, not the unbiasedness of the constant disk/spheroid fractions.

1 more flagged steps
  1. self citation load bearing [Section 4.2, immediately after Figure 9]
    "As presented in K24, we expect an accuracy of approximately 90%. The observed decrease of about 15% on average reinforces how spheroidal and disk galaxies can be misclassified simply due to degradation."

    The acceptability of the degradation retention (92% spheroidal, 95% disk) is calibrated against an accuracy value taken from the authors' own prior paper (K24), which introduced the same SOM+CNN hybrid method. Since K24's accuracy is not an external ground-truth benchmark independent of the metric/SOM labels, using it as the expected baseline makes the FERENGI test a comparison of the pipeline against itself rather than against physical truth.

full rationale

The central claim of an almost constant disk fraction (~60%) and spheroid fraction (~30%) is not fully circular, because the paper benchmarks it against Lee et al. (2024) using independent JWST data and a different unsupervised method; that external comparison provides genuine supporting evidence. However, two supporting results are partially constructed by the method's own definitions. First, the irregular class is defined as CNN probability between 0.2 and 0.8, i.e., as the model's inability to decide between disk and spheroid; the reported increase of the irregular fraction with redshift is therefore a re-expression of how the classifier's confidence distribution shifts, not an independent morphological measurement. Second, the FERENGI degradation test in Section 4.2 classifies artificially redshifted galaxies using bin-dedicated CNNs trained on labels produced by the very same SOM/IsoData metric pipeline they are meant to validate; high retention rates only show that degraded low-redshift galaxies resemble the metric-cluster labels of high-redshift galaxies, not that those labels are physically correct. The additional appeal to K24's 'expected accuracy of approximately 90%' as the interpretation baseline is a self-citation that does not break this circle. These issues affect the irregular-fraction evolution and the quantitative 'up to 18% misclassification' estimate, but do not by themselves invalidate the main constant-fraction conclusion, which has external support. Overall partial circularity: score 6.

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

No new physical entities are introduced. The central claim rests on several domain assumptions about the validity of the unsupervised labels, the definition of irregulars, and the FERENGI degradation simulation. The free parameters (CNN thresholds and the evolution term) directly shape the reported fractions.

free parameters (5)
  • CNN probability thresholds for morphological classes = spheroid: p>0.9; disk: p<0.1; irregular: 0.2<=p<=0.8; unclassifiable: p in (0.1,0.2) or (0.8,0.9)
    Chosen by hand from the CNN output distribution (Figure 5). Directly determines the irregular fraction and affects disk/spheroid fractions.
  • FERENGI luminosity evolution term = -1.0 mag per unit redshift
    Fixed ad hoc in Section 4.2 using the relation M_z' = M_z0 - (1*z'); the paper states it is 'simply a way to show the impact of it' and that a more accurate form is beyond scope.
  • Entropy number of bins (nbins) = optimized via WCSS knee
    Tunable parameter for the Entropy metric, optimized in Section 3.3 by maximizing the WCSS knee.
  • G2 module and phase tolerances (mtol, ptol) = optimized via WCSS knee
    Tunable parameters for the G2 metric, optimized in Section 3.3.
  • SOM grid size = sqrt(N_objects*0.1) with +/-2 variations
    Heuristic from equation 4, varied across runs.
assumptions (5)
  • domain assumption The MEGG metrics measured in the F814W band separate disk from spheroid morphology across 0.2<z<2.4.
    The entire labeling scheme rests on this. The paper cites K24 and Lotz et al. for low redshift, and assumes the separation holds at high z despite degradation (Section 3, Figure 8).
  • domain assumption The SOM prominent clusters correspond to true morphological classes, not just metric extremes.
    Prominent clusters are selected by IsoData thresholds on each metric (Section 3.2); the paper assumes these are representative of disks and spheroids.
  • ad hoc to paper Galaxies with intermediate CNN probability (0.2-0.8) are genuinely irregular systems.
    This definition is introduced in Section 3.5 and is not validated against an independent irregular sample.
  • domain assumption The FERENGI simulation with a fixed luminosity evolution term of -1.0 captures the relevant redshift degradation effects.
    Used in Section 4.2 to estimate misclassification rates; the evolution term is admittedly uncertain.
  • domain assumption Single-band F814W morphology classification corresponds to rest-frame morphology in the same way at all redshifts.
    Lee et al. (2024) vary observed band to match rest-frame; the paper argues agreement implies minimal wavelength dependence, but this is an assumption (Section 5).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Galaxy Morphology in CANDELS: Addressing Evolutionary Changes Across $0.2 \leq z \leq 2.4$ with Hybrid Classification Approach." pith.science (2026). https://pith.science/paper/KBIPPWNZ

@misc{pith2026241203778,
  author       = {Pith},
  title        = {Pith review of: Galaxy Morphology in CANDELS: Addressing Evolutionary Changes Across $0.2 \leq z \leq 2.4$ with Hybrid Classification Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KBIPPWNZ}},
  note         = {Machine review of arXiv:2412.03778}
}
abstract

Morphological classification of galaxies becomes increasingly challenging with redshift. We apply a hybrid supervised-unsupervised method to classify $\sim 14,000$ galaxies in the CANDELS fields at $0.2 \leq z \leq 2.4$ into spheroid, disk, and irregular systems. Unlike previous works, our method is applied to redshift bins of width 0.2. Comparison between models applied to a wide redshift range versus bin-specific models reveals significant differences in galaxy morphology beyond $z \geq 1$ and a consistent $\sim 25\%$ disagreement. This suggests that using a single model across wide redshift ranges may introduce biases due to the large time intervals involved compared to galaxy evolution timescales. Using the FERENGI code to assess the impact of cosmological effects, we find that flux dimming and smaller angular scales may lead to the misclassification of up to $18\%$ of disk galaxies as spheroids or irregulars. Contrary to previous studies, we find an almost constant fraction of disks ($\sim 60\%$) and spheroids ($\sim 30\%$) across redshifts. We attribute discrepancies with earlier works, which suggest a decreasing fraction of disks beyond $z \sim 1$, to the biases introduced by visual classification. Our claim is further strengthened by the striking agreement to the results reported by Lee et al. (2024) using an objective, unsupervised method applied to James Webb Space Telescope data. Exploring mass dependence, we observe a $\sim 40\%$ increase in the fraction of massive ($M_{\rm stellar} \geq 10^{10.5}{\rm M}_{\odot}$) spheroids with decreasing redshift, well balanced with a decrease of $\sim 20\%$ in the fraction of $M_{\rm stellar} \geq 10^{10.5}{\rm M}_{\odot}$ disks, suggesting that merging massive disk galaxies may form spheroidal systems.

Figures

Figures reproduced from arXiv: 2412.03778 by the authors.

Figure 1
Figure 1. Panel (a) displays the median stellar mass distribution across each field, whereas Panel (b) presents the distribution of apparent magnitudes in the H band. Red lines demonstrate cuts we applied to select data for our sample. Panel (c) shows the distribution of stellar masses of CANDELS fields with selected region in black rectangle. disk-like galaxies from 15% at 𝑧 = 0 to 80% at 𝑧 = 2, while the frac￾tion of sphero… view at source ↗
Figure 2
Figure 2. Final galaxy selection for hybrid classification after applying the cuts of Hmag = 24 and stellar mass greater than 109M⊙, followed by cleaning, segmentation of cutouts, and metric measurement. The data is divided into redshift bins, with each bin containing galaxies within a specified redshift range, treated as an independent dataset. corresponding to a galaxy. While it correlates with concentration, it does not ne… view at source ↗
Figure 3
Figure 3. Examples of randomly selected galaxies from out dataset basing on our model classification. This figure is split into 4 segments. Top Left 3x3 segment is showing 9 galaxies classified as spheroidal at 0.2 < z < 0.4 (first bin); Top Right 3x3 segment is showing 9 galaxies classified as spheroidal at 2.2 < z < 2.4 (last bin); Bottom Left 3x3 segment is showing 9 galaxies classified as disk at 0.2 < z < 0.4 (first bin)… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Comparison between metric capacity to split galaxies into multiple classes, at the same time indicating the optimal number of supercluster for each metric (signalized by red segment). It is possible to note that slope of red segment in each panel a is much steeper, ind…
Figure 5
Figure 5. Figure 5: Confidence cuts applied to CNN prediction probabilities based on the return from the Binary Cross Entropy loss function, providing a single numeric value for each galaxy’s prediction. The figure displays the probability distribution from ensemble of one-hundred models …
Figure 6
Figure 6. Figure 6: Showcase of model degradation across bins of z. Panel (a) demonstrates the ability of a model trained on the 0.2 < 𝑧 < 0.4 bin classifies galaxies from all other bins. The figure shows expected results, where the performance falls rapidly as the galaxies change with hi…
Figure 7
Figure 7. Figure 7: The figure presents several examples of the same galaxies that have been artificially redshifted. The left 3x3 mosaic displays the original selected disk galaxies, while the right shows these galaxies redshifted to a redshift of 2.3, corresponding to the last 𝑧 bin. It…
Figure 9
Figure 9. Figure 9: Degradation results for two main classes; spheroidal in panel (a) and disk in panel (b). Each panel contains fractions of disk, spheroidal and irregular galaxies (from the simulated sample only) across our redshift range. It can be noted that that both classes showcase…
Figure 8
Figure 8. Figure 8: Metrics measured on degraded galaxies in comparison with the real disk and real spheroidal galaxies across the entire redshift range based on their metric values. Blue and Red solid lines correspond to median value of main two classes. Blue and Red dotted lines corresp…
Figure 10
Figure 10. Figure 10: Comparison of the fractions obtained in this research with ones presented in the literature over last decade. These fractions are plotted across the redshifts for disk, spheroidal and Irregular galaxies. It is notable that we we tend to agree with most of the findings…
Figure 11
Figure 11. Figure 11: Companion figure to [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. The formation and evolution of Supermassive disks in IllustrisTNG

    astro-ph.GA 2025-06 conditional novelty 6.0 of 10

    In IllustrisTNG, supermassive disk galaxies survive for billions of years and form mainly through quiet, gas-rich merger histories rather than through many major mergers.

Reference graph

Works this paper leans on

90 extracted references · 40 canonical work pages · cited by 1 Pith paper

  1. [1]

    pp 265--283

    Abadi M., et al., 2016, in 12th USENIX symposium on operating systems design and implementation (OSDI 16). pp 265--283

  2. [2]

    G., van den Bergh S., 2001, @doi [Science] 10.1126/science.1060855 , https://ui.adsabs.harvard.edu/abs/2001Sci...293.1273A 293, 1273

    Abraham R. G., van den Bergh S., 2001, @doi [Science] 10.1126/science.1060855 , https://ui.adsabs.harvard.edu/abs/2001Sci...293.1273A 293, 1273

  3. [3]

    G., et al., 2003, The Astrophysical Journal, 588, 218

    Abraham R. G., et al., 2003, The Astrophysical Journal, 588, 218

  4. [4]

    H., et al., 2020, Astronomy and Computing, 30, 100334

    Barchi P. H., et al., 2020, Astronomy and Computing, 30, 100334

  5. [5]

    Barden M., Jahnke K., H \"a u ler B., 2008, The Astrophysical Journal Supplement Series, 175, 105

  6. [6]

    E., Hernquist L., 1992, @doi [ ] 10.1146/annurev.aa.30.090192.003421 , https://ui.adsabs.harvard.edu/abs/1992ARA&A..30..705B 30, 705

    Barnes J. E., Hernquist L., 1992, @doi [ ] 10.1146/annurev.aa.30.090192.003421 , https://ui.adsabs.harvard.edu/abs/1992ARA&A..30..705B 30, 705

  7. [7]

    Barro G., et al., 2019, The Astrophysical Journal Supplement Series, 243, 22

  8. [8]

    V., 2019, in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

    Bello I., Zoph B., Vaswani A., Shlens J., Le Q. V., 2019, in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

Show all 90 references
  1. [9]

    M., 2007, Pattern recognition and machine learning (information science and statistics)

    Bishop C. M., 2007, Pattern recognition and machine learning (information science and statistics)

  2. [11]

    374, From Stars to Galaxies: Building the Pieces to Build Up the Universe

    Bruzual G., 2007, in Vallenari A., Tantalo R., Portinari L., Moretti A., eds, Astronomical Society of the Pacific Conference Series Vol. 374, From Stars to Galaxies: Building the Pieces to Build Up the Universe. p. 303 ( @eprint arXiv astro-ph/0702091 ), @doi 10.48550/arXiv.as...

  3. [12]

    Bruzual G., Charlot S., 2003, @doi [ ] 10.1046/j.1365-8711.2003.06897.x , https://ui.adsabs.harvard.edu/abs/2003MNRAS.344.1000B 344, 1000

  4. [13]

    S., Targett T

    Bundy K., Fukugita M., Ellis R. S., Targett T. A., Belli S., Kodama T., 2009, @doi [ ] 10.1088/0004-637X/697/2/1369 , https://ui.adsabs.harvard.edu/abs/2009ApJ...697.1369B 697, 1369

  5. [14]

    Cassata P., et al., 2005, Monthly Notices of the Royal Astronomical Society, 357, 903

  6. [15]

    Cheng T.-Y., et al., 2021, Monthly Notices of the Royal Astronomical Society, 507, 4425

  7. [16]

    Cheng T.-Y., et al., 2023, Monthly Notices of the Royal Astronomical Society, 518, 2794

  8. [17]

    pp 1251--1258

    Chollet F., 2017, in Proceedings of the IEEE conference on computer vision and pattern recognition. pp 1251--1258

  9. [18]

    Cambridge University Press

    Cimatti A., Fraternali F., Nipoti C., 2019, Introduction to Galaxy Formation and Evolution: From Primordial Gas to Present-Day Galaxies . Cambridge University Press

  10. [19]

    Coe D., et al., 2012, The Astrophysical Journal, 757, 22

  11. [20]

    J., 2014, @doi [ ] 10.1146/annurev-astro-081913-040037 , https://ui.adsabs.harvard.edu/abs/2014ARA&A..52..291C 52, 291

    Conselice C. J., 2014, @doi [ ] 10.1146/annurev-astro-081913-040037 , https://ui.adsabs.harvard.edu/abs/2014ARA&A..52..291C 52, 291

  12. [21]

    Conselice C., Bluck A., Ravindranath S., Mortlock A., Koekemoer A., Buitrago F., Gr \"u tzbauch R., Penny S., 2011, Monthly Notices of the Royal Astronomical Society, 417, 2770

  13. [22]

    Contardo G., Steinmetz M., Fritze-von Alvensleben U., 1998, @doi [ ] 10.1086/306350 , https://ui.adsabs.harvard.edu/abs/1998ApJ...507..497C 507, 497

  14. [23]

    Cottrell M., Letr \'e my P., 2005, Neurocomputing, 63, 193

  15. [24]

    Cottrell M., Ibbou S., Letr \'e my P., 2004, Neural Networks, 17, 1149

  16. [25]

    Dai J., Qi H., Xiong Y., Li Y., Zhang G., Hu H., Wei Y., 2017, in Proceedings of the IEEE International Conference on Computer Vision (ICCV)

  17. [26]

    Delgado-Serrano R., Hammer F., Yang Y., Puech M., Flores H., Rodrigues M., 2010, Astronomy & Astrophysics, 509, A78

  18. [27]

    pp 248--255

    Deng J., Dong W., Socher R., Li L.-J., Li K., Fei-Fei L., 2009, in 2009 IEEE conference on computer vision and pattern recognition. pp 248--255

  19. [28]

    Dom \' nguez S \'a nchez H., Huertas-Company M., Bernardi M., Tuccillo D., Fischer J., 2018, Monthly Notices of the Royal Astronomical Society, 476, 3661

  20. [29]

    Dressler A., 1980, @doi [ ] 10.1086/157753 , https://ui.adsabs.harvard.edu/abs/1980ApJ...236..351D 236, 351

  21. [30]

    Dressler A., et al., 1997, @doi [ ] 10.1086/304890 , https://ui.adsabs.harvard.edu/abs/1997ApJ...490..577D 490, 577

  22. [31]

    F., Kere s D., Chan T

    El-Badry K., Wetzel A., Geha M., Hopkins P. F., Kere s D., Chan T. K., Faucher-Gigu \`e re C.-A., 2016, @doi [ ] 10.3847/0004-637X/820/2/131 , https://ui.adsabs.harvard.edu/abs/2016ApJ...820..131E 820, 131

  23. [32]

    Faber S., 2011, MAST, doi, 10, T94S3X

  24. [33]

    R., Trevisan M., 2015, The Astrophysical Journal, 814, 55

    Ferrari F., de Carvalho R. R., Trevisan M., 2015, The Astrophysical Journal, 814, 55

  25. [34]

    Fioc M., Rocca-Volmerange B., 1997, @doi [ ] 10.48550/arXiv.astro-ph/9707017 , https://ui.adsabs.harvard.edu/abs/1997A&A...326..950F 326, 950

  26. [35]

    A., et al., 2011, The Astrophysical Journal Supplement Series, 197, 35

    Grogin N. A., et al., 2011, The Astrophysical Journal Supplement Series, 197, 35

  27. [36]

    E., 2020, The Astrophysical Journal Supplement Series, 248, 20

    Hausen R., Robertson B. E., 2020, The Astrophysical Journal Supplement Series, 248, 20

  28. [37]

    F., Hernquist L., Cox T

    Hopkins P. F., Hernquist L., Cox T. J., Di Matteo T., Robertson B., Springel V., 2006, @doi [ ] 10.1086/499298 , https://ui.adsabs.harvard.edu/abs/2006ApJS..163....1H 163, 1

  29. [38]

    A., Martin G., Kaviraj S., Laigle C., Devriendt J

    Jackson R. A., Martin G., Kaviraj S., Laigle C., Devriendt J. E. G., Dubois Y., Pichon C., 2019, @doi [ ] 10.1093/mnras/stz2440 , https://ui.adsabs.harvard.edu/abs/2019MNRAS.489.4679J 489, 4679

  30. [39]

    R., 1996, Introductory digital image processing: a remote sensing perspective

    Jensen J. R., 1996, Introductory digital image processing: a remote sensing perspective.. Taylor & Francis

  31. [40]

    S., et al., 2015, The Astrophysical Journal Supplement Series, 221, 11

    Kartaltepe J. S., et al., 2015, The Astrophysical Journal Supplement Series, 221, 11

  32. [41]

    Khalifa N. E. M., Taha M. H. N., Hassanien A. E., Selim I., 2017, arXiv preprint arXiv:1709.02245

  33. [42]

    Khan A., Huerta E., Wang S., Gruendl R., Jennings E., Zheng H., 2019, Physics Letters B, 795, 248

  34. [43]

    Khochfar S., Burkert A., 2003, @doi [ ] 10.1086/379845 , https://ui.adsabs.harvard.edu/abs/2003ApJ...597L.117K 597, L117

  35. [44]

    M., et al., 2011, The Astrophysical Journal Supplement Series, 197, 36

    Koekemoer A. M., et al., 2011, The Astrophysical Journal Supplement Series, 197, 36

  36. [45]

    M., de Carvalho R

    Kolesnikov I., Sampaio V. M., de Carvalho R. R., Conselice C., Rembold S. B., Mendes C. L., Rosa R. R., 2024, @doi [ ] 10.1093/mnras/stad3934 , https://ui.adsabs.harvard.edu/abs/2024MNRAS.528...82K 528, 82

  37. [46]

    H., Park C., Hwang H

    Lee J. H., Park C., Hwang H. S., Kwon M., 2024, The Astrophysical Journal, 966, 113

  38. [47]

    Lintott C., et al., 2011, Monthly Notices of the Royal Astronomical Society, 410, 166

  39. [48]

    Lloyd S., 1982, IEEE transactions on information theory, 28, 129

  40. [49]

    M., Primack J., Madau P., 2004, The Astronomical Journal, 128, 163

    Lotz J. M., Primack J., Madau P., 2004, The Astronomical Journal, 128, 163

  41. [50]

    Madau P., Dickinson M., 2014, @doi [ ] 10.1146/annurev-astro-081811-125615 , https://ui.adsabs.harvard.edu/abs/2014ARA&A..52..415M 52, 415

  42. [51]

    Maraston C., 2005, @doi [ ] 10.1111/j.1365-2966.2005.09270.x , https://ui.adsabs.harvard.edu/abs/2005MNRAS.362..799M 362, 799

  43. [52]

    E., Romanowsky A

    Moody C. E., Romanowsky A. J., Cox T. J., Novak G. S., Primack J. R., 2014, @doi [ ] 10.1093/mnras/stu1444 , https://ui.adsabs.harvard.edu/abs/2014MNRAS.444.1475M 444, 1475

  44. [53]

    Mortlock A., et al., 2013, Monthly Notices of the Royal Astronomical Society, 433, 1185

  45. [54]

    Nayyeri H., et al., 2017, The Astrophysical Journal Supplement Series, 228, 7

  46. [55]

    Peebles P. J. E., 1969, @doi [ ] 10.1086/149876 , https://ui.adsabs.harvard.edu/abs/1969ApJ...155..393P 155, 393

  47. [56]

    Planck Collaboration et al., 2020, @doi [ ] 10.1051/0004-6361/201833910 , https://ui.adsabs.harvard.edu/abs/2020A&A...641A...6P 641, A6

  48. [57]

    M., 1997, @doi [ ] 10.1051/aas:1997142 , https://ui.adsabs.harvard.edu/abs/1997A&AS..122..399P 122, 399

    Poggianti B. M., 1997, @doi [ ] 10.1051/aas:1997142 , https://ui.adsabs.harvard.edu/abs/1997A&AS..122..399P 122, 399

  49. [58]

    J., et al., 2024, RAS Techniques and Instruments, 3, 174

    Popp J. J., et al., 2024, RAS Techniques and Instruments, 3, 174

  50. [59]

    Primack J., et al., 2018, The Astrophysical Journal, 858, 114

  51. [60]

    E., et al., 2023, The Astrophysical Journal Letters, 942, L42

    Robertson B. E., et al., 2023, The Astrophysical Journal Letters, 942, L42

  52. [61]

    Rodriguez-Gomez V., et al., 2015, @doi [ ] 10.1093/mnras/stv264 , https://ui.adsabs.harvard.edu/abs/2015MNRAS.449...49R 449, 49

  53. [62]

    Rodriguez-Gomez V., et al., 2017, @doi [ ] 10.1093/mnras/stx305 , https://ui.adsabs.harvard.edu/abs/2017MNRAS.467.3083R 467, 3083

  54. [63]

    Rosa R., et al., 2018, Monthly Notices of the Royal Astronomical Society: Letters, 477, L101

  55. [64]

    M., de Carvalho R

    Sampaio V. M., de Carvalho R. R., Ferreras I., Arag \'o n-Salamanca A., Parker L. C., 2022, @doi [ ] 10.1093/mnras/stab3018 , https://ui.adsabs.harvard.edu/abs/2022MNRAS.509..567S 509, 567

  56. [65]

    M., Arag \'o n-Salamanca A., Merrifield M

    Sampaio V. M., Arag \'o n-Salamanca A., Merrifield M. R., de Carvalho R. R., Zhou S., Ferreras I., 2023, @doi [ ] 10.1093/mnras/stad2211 , https://ui.adsabs.harvard.edu/abs/2023MNRAS.524.5327S 524, 5327

  57. [66]

    M., de Carvalho R

    Sampaio V. M., de Carvalho R. R., Arag \'o n-Salamanca A., Merrifield M. R., Ferreras I., Cornwell D. J., 2024, @doi [ ] 10.1093/mnras/stae1533 , https://ui.adsabs.harvard.edu/abs/2024MNRAS.532..982S 532, 982

  58. [67]

    Sandage A., 1961, @doi [ ] 10.1086/147218 , https://ui.adsabs.harvard.edu/abs/1961ApJ...134..916S 134, 916

  59. [68]

    Santini P., et al., 2015, The Astrophysical Journal, 801, 97

  60. [69]

    Schawinski K., et al., 2014, @doi [ ] 10.1093/mnras/stu327 , https://ui.adsabs.harvard.edu/abs/2014MNRAS.440..889S 440, 889

  61. [70]

    M., Bryan G

    Simpson C. M., Bryan G. L., Johnston K. V., Smith B. D., Mac Low M.-M., Sharma S., Tumlinson J., 2013, @doi [ ] 10.1093/mnras/stt474 , https://ui.adsabs.harvard.edu/abs/2013MNRAS.432.1989S 432, 1989

  62. [71]

    R., De Roure D., 2014, in Proceedings of the 23rd international conference on world wide web

    Simpson R., Page K. R., De Roure D., 2014, in Proceedings of the 23rd international conference on world wide web. pp 1049--1054

  63. [72]

    Springel V., Hernquist L., 2005, @doi [ ] 10.1086/429486 , https://ui.adsabs.harvard.edu/abs/2005ApJ...622L...9S 622, L9

  64. [73]

    Stefanon M., et al., 2017, The Astrophysical Journal Supplement Series, 229, 32

  65. [74]

    Strateva I., et al., 2001, @doi [ ] 10.1086/323301 , https://ui.adsabs.harvard.edu/abs/2001AJ....122.1861S 122, 1861

  66. [75]

    Talia M., Cimatti A., Mignoli M., Pozzetti L., Renzini A., Kurk J., Halliday C., 2014, Astronomy & Astrophysics, 562, A113

  67. [76]

    A., et al., 2009, Astronomy & Astrophysics, 503, 379

    Tasca L. A., et al., 2009, Astronomy & Astrophysics, 503, 379

  68. [77]

    F., Remus R.-S., Dolag K., Beck A

    Teklu A. F., Remus R.-S., Dolag K., Beck A. M., Burkert A., Schmidt A. S., Schulze F., Steinborn L. K., 2015, @doi [ ] 10.1088/0004-637X/812/1/29 , https://ui.adsabs.harvard.edu/abs/2015ApJ...812...29T 812, 29

  69. [78]

    Tohill C., Ferreira L., Conselice C., Bamford S., Ferrari F., 2021, The Astrophysical Journal, 916, 4

  70. [79]

    Toomre A., Toomre J., 1972, @doi [ ] 10.1086/151823 , https://ui.adsabs.harvard.edu/abs/1972ApJ...178..623T 178, 623

  71. [80]

    Trussler J., Maiolino R., Maraston C., Peng Y., Thomas D., Goddard D., Lian J., 2020, @doi [ ] 10.1093/mnras/stz3286 , https://ui.adsabs.harvard.edu/abs/2020MNRAS.491.5406T 491, 5406

  72. [81]

    R., 1979, Thresholding using the ISODATA clustering algorithm

    Velasco F. R., 1979, Thresholding using the ISODATA clustering algorithm. University of Maryland, Computer Science Center

  73. [82]

    Villa-Vialaneix N., 2017, in 2017 12th International Workshop on Self-Organizing Maps and Learning Vector Quantization, Clustering and Data Visualization (WSOM). pp 1--7

  74. [83]

    Walmsley M., et al., 2022, Monthly Notices of the Royal Astronomical Society, 509, 3966

  75. [84]

    Walmsley M., et al., 2023, Monthly Notices of the Royal Astronomical Society, 526, 4768

  76. [85]

    Wei S., Lu W., Dai W., Liang B., Hao L., Zhang Z., Zhang X., 2023, The Astronomical Journal, 167, 29

  77. [86]

    R., Tinker J

    Wetzel A. R., Tinker J. L., Conroy C., 2012, @doi [ ] 10.1111/j.1365-2966.2012.21188.x , https://ui.adsabs.harvard.edu/abs/2012MNRAS.424..232W 424, 232

  78. [87]

    Whitney A., Ferreira L., Conselice C., Duncan K., 2021, The Astrophysical Journal, 919, 139

  79. [88]

    G., et al., 2000, @doi [ ] 10.1086/301513 , https://ui.adsabs.harvard.edu/abs/2000AJ....120.1579Y 120, 1579

    York D. G., et al., 2000, @doi [ ] 10.1086/301513 , https://ui.adsabs.harvard.edu/abs/2000AJ....120.1579Y 120, 1579

  80. [89]

    Zeng G., Wang L., Gao L., Yang H., 2024, @doi [ ] 10.1093/mnras/stae1651 , https://ui.adsabs.harvard.edu/abs/2024MNRAS.532.2558Z 532, 2558

  81. [90]

    Zhao Z., Alzubaidi L., Zhang J., Duan Y., Gu Y., 2023, Expert Systems with Applications, p. 122807

  82. [91]

    van der Wel A., et al., 2014, The Astrophysical Journal, 788, 28

Pith tools

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