Pith. sign in

REVIEW 3 major objections 5 minor 110 references

Searching for Quiescent Galaxies over $3 < z < 6$ in JWST Surveys Using Manifold Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Manifold learning finds 44 quiescent galaxies at z>3 using a candidate pool five times smaller than color-color selection.

desk verdict A useful and mostly solid application of UMAP to pre-select z>3 quiescent galaxies; the efficiency gain is real, but the validation is partly circular and the number densities need completeness caveats. read the letter →

arxiv 2501.09066 v1 pith:MRVI5HKP submitted 2025-01-15 astro-ph.GA

Alexander de la Vega , Mitchell D. Babcock , Bahram Mobasher , Dominik A. Riemann , Nima Chartab , Shoubaneh Hemmati , Arianna S. Long , Sogol Sanjaripour This is my paper · ORCID
classification astro-ph.GA
keywords QuenchedgalaxiesHigh-redshiftAstronomydatavisualizationquiescentUMAPmanifoldlearningJWST/NIRCamJADES
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

Quiescent galaxies at $3 < z < 6$ are rare and hard to find: they formed and stopped forming stars within about two billion years of the Big Bang, and existing rest-frame color-color diagrams become unreliable at these redshifts. This paper argues that a machine-learning pre-selection step can replace most of that search. The authors train the Uniform Manifold Approximation and Projection algorithm (UMAP) on seven observed-frame NIRCam colors from the JAGUAR mock catalog, producing a two-dimensional map on which galaxies with similar colors sit close together. Projecting 43,926 JADES galaxies onto this map and fitting only those near the quiescent model clusters yields 44 quiescent candidates from a pool of about 2,300 galaxies, roughly five times smaller than the pool needed with observed-frame color-color wedges. Two-thirds of the candidates are recovered from a pool of 247 galaxies, and nearly two-thirds are newly discovered, including galaxies with mass-weighted ages below 300 Myr.

What carries the argument

The central object is UMAP (Uniform Manifold Approximation and Projection), a nonlinear dimensionality-reduction algorithm that builds a weighted nearest-neighbor graph in a high-dimensional space and then embeds it into two dimensions while preserving local and global structure. Trained here on roughly 75,000 JAGUAR mock galaxies described by seven observed-frame NIRCam colors (F115W-F150W, F115W-F277W, F150W-F200W, F150W-F277W, F200W-F277W, F200W-F356W, F277W-F444W), the embedding places galaxies with similar colors near one another. The map is used as a lookup table: observed JADES galaxies are projected onto it with UMAP's transform routine, and proximity to the clustered quiescent models defines the candidate pool. The quiescence of the pre-selected galaxies is then established by fitting their HST+JWST photometry with the bagpipes SED-fitting code and applying an sSFR threshold of 0.2 divided by the age of the Universe at that redshift.

What would settle it

Take a spectroscopically complete sample of galaxies at $3 < z < 6$ in JADES or a similar field, measure their sSFRs, and compare the quiescent ones to the UMAP candidate pools; if any spectroscopically confirmed quiescent galaxy falls outside both the radial-distance and rectangular pre-selection regions, the method is incomplete as a pre-selection and the reported number densities would be lower limits rather than measurements.

Watch

Extended reading notes

Core claim

The paper's central claim is that a pre-selection based on UMAP manifold learning is a more efficient way to find quiescent galaxies over $3 < z < 6$ than observed-frame color-color diagrams. Using seven NIRCam colors designed to isolate high-redshift quiescent galaxies, the trained UMAP transformation maps the 62 model quiescent galaxies at $z \geq 3$ into a tight cluster; all 17 spectroscopically confirmed massive quiescent galaxies from a recent JWST study fall in the same region, validating the map. When the full JADES observational sample is transformed, selecting all galaxies within a radial distance of 1 in UMAP space yields 2,282 candidates, of which 44 are quiescent after SED fitting. An even smaller selection using two rectangular regions recovers 29 of the 44 from only 247 galaxies, about twice as efficient as the comparable color-color wedge. The paper concludes that the method captures young ($< 300$ Myr) quiescent galaxies that color-color criteria tend to miss, and that the derived number densities agree with earlier work at $z < 4$ while tending to be higher, but consistent within errors, at $z > 4$.

Load-bearing premise

The load-bearing premise is that the JAGUAR mock catalog contains the full range of observed colors of real high-redshift quiescent galaxies, particularly the young, blue, recently quenched systems; the authors acknowledge that JAGUAR's quiescent SEDs are built from UVJ-selected, mostly massive galaxies and may not cover the bluest population, and if that diversity is missing, the UMAP map will place such galaxies away from the quiescent model clusters and the pre-selection will miss them.

Editorial extensions

If this is right

  • The same trained UMAP map can be applied to other JWST surveys with NIRCam coverage, so future searches for $z > 3$ quiescent galaxies can avoid SED-fitting tens of thousands of sources.
  • If the higher number densities at $z > 4$ hold, they would strengthen the already serious tension between observed quiescent galaxy abundances and cosmological simulations, which currently underpredict them by 1-2 dex.
  • The recovery of several candidates with mass-weighted ages below 300 Myr implies that a subset of galaxies quench within about 300 Myr of their formation, providing a sharp timescale constraint on quenching mechanisms at early epochs.
  • The roughly fivefold reduction in the candidate pool makes it feasible to obtain spectroscopy for the vast majority of pre-selected candidates, which is needed to confirm their quiescence and measure their ages.

Reading between the lines

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

  • Adding more photometric bands or morphological information as UMAP inputs could shrink the candidate pool further, a direction the paper does not explore.
  • Because JAGUAR's quiescent templates are drawn from UVJ-selected galaxies, the map may be blind to the bluest, youngest quiescent galaxies; injecting such synthetic SEDs into the training set would test whether they land inside the candidate regions.
  • The method's reliance on observed-frame colors means it can be applied without photometric redshifts, so it could be run on raw photometric catalogs before any redshift estimation is performed.
  • If applied to wide-area surveys, the pre-selection would drastically reduce the computational cost of SED fitting, potentially enabling quiescent galaxy searches over hundreds of square degrees.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper applies the Uniform Manifold Approximation and Projection (UMAP) technique to seven observed-frame NIRCam colors of 43,926 galaxies in JADES (GOODS-N and GOODS-S) to pre-select quiescent galaxy candidates at 3<z<6. A UMAP embedding is trained on ~75,000 JAGUAR mock galaxies, and the observed sample is mapped onto this embedding. Two pre-selection strategies are presented: a radial cut of radius 1 around the z>=3 quiescent model cluster, yielding 2,282 galaxies and 44 quiescent candidates, and two rectangular cuts enclosing the main model clusters, yielding 247 galaxies and 29 candidates. The paper claims that the UMAP-based pre-selection is roughly five times more efficient than an equivalent observed-frame color-color wedge and about twice as efficient as the wide Long et al. (2024) wedge, and it reports number densities at 3<z<6 that agree with the literature at z<4 and are generally higher at z>4. The final catalog contains 27 newly discovered candidates, several with mass-weighted ages below 300 Myr.

Significance. If the efficiency claims hold, the paper provides a genuinely useful tool for finding rare high-redshift quiescent galaxies in JWST surveys, and the public code and catalog are a valuable community resource. The efficiency comparison in Figure 6 is well constructed and clearly demonstrates that, for the specific candidate set found here, UMAP pre-selection yields a smaller pool than a color-color wedge containing the same candidates. The recovery of 17 spectroscopically confirmed Baker et al. (2024) quiescent galaxies near the model cluster is encouraging, and the new low-mass, young candidates are of interest. The main weakness is that the pre-selection completeness is anchored to JAGUAR's UVJ-based quiescent locus and to the Baker et al. sample used for calibration, so the selection function is not independently established. This limits the interpretation of the number densities, which are presented without a completeness correction.

major comments (3)
  1. [Section 3.2 and Section 5.6] The number densities in Table 2 and Figure 7 are based on raw counts over the pre-selected pool with no completeness correction, but the pre-selection is trained on JAGUAR quiescent SEDs that, as the paper states in Section 3.2, are drawn from a UVJ-selected, mostly massive parent sample and do not cover the full diversity of z>3 quiescent galaxies, especially bluer, younger populations. If such populations lie outside the model-based clusters in UMAP space, the pre-selection will miss them, and the z>4 and low-mass number densities will be biased low. The statement that the impact is negligible is not demonstrated. Please provide a quantitative completeness test, for example by injecting young, blue quiescent SEDs (from other models or from the observed candidates themselves) into the UMAP embedding and measuring the recovery fraction, or explicitly present the number densities as lower limits with the selection function stated as unknown.
  2. [Sections 4.2, 4.3.1, and 5.1] The validation using the 17 Baker et al. (2024) spectroscopically confirmed galaxies is partly circular. The UMAP hyperparameters (n_neighbors=100, min_dist=0.01) were chosen by sweeping to maximize clustering of the z>=3 quiescent JAGUAR models, and the radial cut of 1 was chosen by doubling the radius of 0.5 that encloses those same 17 galaxies. Consequently, the fact that all 17 lie close to the quiescent model cluster is not an independent confirmation of the technique's completeness. An out-of-sample test, such as applying the pre-trained UMAP to a survey not used in the calibration (e.g., CEERS or UNCOVER) or using a held-out subset of spectroscopically confirmed galaxies, would strengthen the claim that the method recovers quiescent galaxies in general rather than just the specific population used for tuning.
  3. [Section 5.5] The headline efficiency comparison ('about five times fewer galaxies') is constructed retrospectively: the color-color wedge that encloses all 44 candidates is obtained by adjusting the intercepts of the Long et al. (2024) criteria, and the rectangular UMAP cuts are also defined a posteriori to enclose the model clusters. The paper itself notes in Section 5 that the exact pre-selection method is arbitrary. Please add a sensitivity analysis showing how the candidate-pool sizes and the efficiency ratios change with reasonable variations of the UMAP hyperparameters and the radial/rectangular cut positions, so that the reader can assess whether the factor-of-five and factor-of-two efficiency gains are robust or are a product of the specific choices made here.
minor comments (5)
  1. [Section 2.2] In the paragraph on photometric redshift accuracy, 'redshfits' should be 'redshifts'.
  2. [Section 5.2] The rectangular cut for Region C is given as -4.0 <= UMAP1 <= -2.8 and -4.8 <= UMAP2 <= -4.6, which is visibly wider than the small group of five models at approximately (-3.5, -5); please state explicitly how the rectangle boundaries were chosen and whether the result is sensitive to those boundaries.
  3. [Figure 6 caption] The caption states that selecting all 44 candidates using color-color cuts would require the extended wedge shown in magenta, but it should be clearer that this is a custom wedge constructed for this comparison and not the Long et al. (2024) selection, which is shown by the other lines.
  4. [Table 1] Several entries have extremely low log sSFR values with very large asymmetric uncertainties (e.g., ID 13124 and ID 40382); consider flagging these as poorly constrained or excluding them from the number-density computation to avoid giving them equal weight.
  5. [Section 5.6] The number-density computation perturbs photometric redshifts using the 16th and 84th percentiles of the PDF, but the selection criteria in Section 5.1 also require the 16th percentile to satisfy z>=2.5; please clarify whether the Monte Carlo realizations maintain consistency with the sample selection criteria.

Circularity Check

2 steps flagged · score 4.0 of 10

Two calibration loops (UMAP hyperparameters tuned to cluster quiescent models; radial cut tuned to Baker et al. validation galaxies) make the validation and efficiency claims partly self-fulfilling, while the new-candidate and number-density results retain independent content.

  1. self definitional [Section 4.2, 'Two-Dimensional Visualization of the JAGUAR Models Using UMAP' (hyperparameter choice)]
    "We set n to 100 and the minimum distance to 0.01. These values were chosen by sweeping through a wide range of values for each parameter and examining how clustered quiescent galaxy models at 3 < z <6 are within the visualization. The high-redshift quiescent models are maximally clustered assuming these values."

    The UMAP embedding is not an independent map of the color space: the hyperparameters are selected by maximizing the clustering of exactly the z≥3 quiescent models whose compactness is then reported ("55 are clustered tightly...") and used to define Regions A-D. The compactness that motivates small pre-selection regions is therefore partly manufactured by the hyperparameter scan rather than emergent from the data. The observed-galaxy pool sizes remain empirical, so this is a partial, not total, construction.

  2. fitted input called prediction [Section 5.1, 'Pre-Selection Based on Radial Distance From Quiescent Models' (radius choice; cf. Section 4.3.1)]
    "A value of 1 is chosen because the spectroscopically confirmed quiescent galaxies in the observational sample (Figure 3 and Section 4.3.1) lie within a radius of 0.5 and this is doubled to ensure we capture as many quiescent galaxies as possible."

    The radial threshold defining the 2,282-galaxy candidate pool is calibrated on the 17 Baker et al. (2024) spectroscopically confirmed quiescent galaxies. Those same galaxies are then presented as showing that the technique "efficiently captures high-redshift quiescent galaxies," and they fall inside the resulting pool by construction. The factor-of-five efficiency gain over a 10,531-galaxy color-color wedge is therefore conditional on a threshold fit to the validation sample; the recovery of those known galaxies is not a held-out prediction.

full rationale

The paper's main empirical outputs—the 44 quiescent candidates, the 27 new discoveries, and the number-density measurements—are not directly forced by the training set: the bagpipes SED fits use independent data and priors, and the new candidates were not part of the JAGUAR training or the Baker calibration. The UMAP mapping itself is trained only on JAGUAR models, so the observed Baker et al. galaxies mapping near the model locus is partly independent evidence. However, two steps weaken the derivation chain. First, the UMAP hyperparameters are explicitly chosen to maximize the clustering of the z≥3 quiescent models, so the tight clustering used to justify compact selection regions is partly a product of the tuning, not an emergent discovery. Second, the radial cut of 1 is calibrated on the Baker et al. galaxies, making their inclusion in the 2,282-galaxy pool and the subsequent efficiency comparison a calibration exercise rather than a validation prediction. The paper also acknowledges that JAGUAR's quiescent SEDs are UVJ-selected and do not cover the full diversity of young, blue, z>3 quiescent populations; the argument that the impact is negligible because some young (<300 Myr) candidates are found does not close that gap, since those candidates were selected near the existing model locus. The stated arbitrariness of the pre-selection method (Section 5) and the absence of a completeness correction in the number densities (Section 5.6) further condition the efficiency and abundance claims, but they do not reduce the new-candidate sample itself to the input data. Overall, the circularity is partial and localized to the validation/efficiency narrative, not to the core discovery claim.

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

The central claim rests on a handful of modeling choices: the UMAP hyperparameters and selection radii are tuned on the training/validation data, and the SED-fitting results depend on the adopted priors, SFH shape, and JAGUAR representativeness. No new physical entities are introduced.

free parameters (3)
  • UMAP n_neighbors and min_dist = n=100, min_dist=0.01
    Chosen by sweeping values to maximize clustering of z>=3 quiescent JAGUAR models in the UMAP embedding (Section 4.2). These affect which observed galaxies land near quiescent model regions.
  • Radial pre-selection radius = 1.0 in UMAP space
    Chosen as twice the radius (0.5) enclosing the 17 Baker et al. spectroscopically confirmed quiescent galaxies (Section 5.1). This directly sets the size of the 2,282-galaxy candidate pool.
  • Rectangular cut boundaries for Regions A and C = UMAP1 [-3.5,-3.1], UMAP2 [-0.35,-0.1]; UMAP1 [-4.0,-2.8], UMAP2 [-4.8,-4.6]
    Defined to enclose the quiescent-model clusters and used for the 247-galaxy pool (Section 5.2). The exact boundaries appear chosen after inspecting the model and candidate distribution.
assumptions (5)
  • domain assumption JAGUAR mock SEDs are representative of real galaxy colors at 3<z<6
    The UMAP map is trained on JAGUAR colors. If the mock catalog underrepresents young, blue quiescent galaxies, such galaxies will not map near the quiescent-model clusters and will be missed (Section 3.2).
  • domain assumption Double power-law star-formation history adequately models high-z quiescent galaxy SEDs
    The bagpipes fits assume this SFH shape, which controls the derived sSFR, mass-weighted ages, and quiescent classifications (Section 4.4).
  • domain assumption EAZY photometric redshifts with JADES templates are reliable enough for z=3-6 selection
    The redshift cuts for candidates depend on the photometric-redshift PDF from Hainline et al. (2024). The authors validate a bias of -0.003 and NMAD=0.042, but those statistics are not broken down at z>3 (Section 2.2).
  • domain assumption A 5% flux error added to each bandpass accounts for template mismatch
    This error term is added to all photometry before SED fitting and affects the posterior widths of the derived physical properties (Section 2.1).
  • domain assumption Gaussian priors on metallicity and dust attenuation are physically motivated
    The priors (Z/Zsun mean 0.3, sigma 0.5; AV mean 0.3, sigma 1.0) are adopted from de la Vega et al. (2025) and Leja et al. (2019) and influence the sSFR and age estimates for candidates (Section 4.4).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Searching for Quiescent Galaxies over $3 < z < 6$ in JWST Surveys Using Manifold Learning." pith.science (2026). https://pith.science/paper/MRVI5HKP

@misc{pith2026250109066,
  author       = {Pith},
  title        = {Pith review of: Searching for Quiescent Galaxies over $3 < z < 6$ in JWST Surveys Using Manifold Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MRVI5HKP}},
  note         = {Machine review of arXiv:2501.09066}
}
abstract

Quiescent galaxies over $3<z<6$ are rare and puzzling. They formed and quenched within two billion years and simulations routinely struggle to predict their observed abundances. Developing a robust identification technique for these galaxies is crucial for constraining galaxy evolution models. Traditional rest-frame color-color selection techniques for quiescent galaxies are known to break down or require adjustments at $z\gtrsim3$. Recently, observed-frame color-color criteria have been established with JWST/NIRCam colors that efficiently pre-select high-redshift quiescent galaxies using only $\lesssim1\%$ of a given sample. In this work, the Uniform Manifold Approximation and Projection machine-learning technique is applied to pre-select quiescent galaxies over $3<z<6$ using observed NIRCam colors. From a parent sample of 43,926 galaxies in JADES, we ultimately find 44 quiescent candidates from a pool of $\approx2,300$ galaxies. This is about five times fewer galaxies than what would be pre-selected using color-color criteria. Two-thirds of these candidates can be pre-selected from a pool as small as 247, which is about twice as efficient as existing observed-frame color selection techniques. Nearly two-thirds of the candidates are new discoveries and include quiescent galaxies with mass-weighted ages as young as $\lesssim300$ Myr. We obtain number densities in agreement with the literature at $z<4$ and find generally higher abundances at $z>4$, although our measurements are consistent within errors. This technique may be applied to other JWST surveys.

Figures

Figures reproduced from arXiv: 2501.09066 by the authors.

Figure 1
Figure 1. We search for quiescent galaxies by training a machine learning algorithm on seven observed-frame colors of model galaxies: F115W – F150W, F115W – F277W, F150W – F200W, F150W – F277W, F200W – F277W, F200W – F356W, and F277W – F444W. These are chosen to separate high-redshift quiescent galaxies from others. Colors of galaxies in the observational sample (Section 2.3) and model galaxies from the JAGUAR mock catalog (W… view at source ↗
Figure 2
Figure 2. This figure shows a two-dimensional visualization of the JAGUAR models using the UMAP technique. It is trained on the seven NIRCam colors shown in [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. This figure shows the mapping of observed galaxy colors in JADES onto the UMAP visualization on the left. Galaxies are shown as dots colored by F150W-F277W color. The observed colors are largely mapped onto the same regions as the JAGUAR models, though some are mapped onto regions where there are no models. Quiescent models are grouped into four regions labeled A through D. The right panel is centered on region A, w… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Quiescent galaxy candidates are identified by fitting the SEDs of galaxies that are near quiescent galaxy models. This is done in two techniques, shown here. In the first, for each model, shown as an empty square above, all galaxies that lie within a radius of value 1 …
Figure 5
Figure 5. Figure 5: This figure shows false-color rest-frame optical images of the candidates. Each galaxy in the final sample (Section 5) is shown in its own panel, which displays a 6′′x 6′′postage stamp and is labeled with its ID and redshift. The blue, green, and red channels of each i…
Figure 6
Figure 6. Figure 6: We find that it is more efficient to pre-select a sample of high-redshift galaxies using the UMAP technique than it is using a single color-color diagram. This figure shows the results of two pre-selection techniques in the F150W–F277W vs. F277W–F444W color-color diagr…
Figure 7
Figure 7. Figure 7: This figure shows comoving number densities of quiescent galaxies at 3 < z < 6 using the UMAP technique ( [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

110 extracted references · 9 canonical work pages

  1. [1]

    Turp, M. D. 2018, MNRAS, 479, 415, doi: 10.1093/mnras/sty1398

    doi:10.1093/mnras/sty1398 2018

  2. [2]

    C., Helton, J

    Alberts, S., Williams, C. C., Helton, J. M., et al. 2024, ApJ, 975, 85, doi: 10.3847/1538-4357/ad66cc

    doi:10.3847/1538-4357/ad66cc 2024

  3. [3]

    2023, ApJ, 943, 166, doi: 10.3847/1538-4357/aca294 Astropy Collaboration, Robitaille, T

    Antwi-Danso, J., Papovich, C., Leja, J., et al. 2023, ApJ, 943, 166, doi: 10.3847/1538-4357/aca294 Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33, doi: 10.1051/0004-6361/201322068 Astropy Collaboration, Price-Whelan, A. M., Sip˝ ocz, B. M., et al. 2018, AJ, 156, 123, doi: 10.3847/1538-3881/aabc4f Astropy Collaboratio...

    doi:10.3847/1538-4357/aca294 2023

  4. [4]

    B., Finkelstein, S

    Bagley, M. B., Finkelstein, S. L., Koekemoer, A. M., et al. 2023, ApJL, 946, L12, doi: 10.3847/2041-8213/acbb08

    doi:10.3847/2041-8213/acbb08 2023

  5. [5]

    M., Lim, S., D’Eugenio, F., et al

    Baker, W. M., Lim, S., D’Eugenio, F., et al. 2024, arXiv e-prints, arXiv:2410.14773, doi: 10.48550/arXiv.2410.14773

    arXiv doi:10.48550/arxiv.2410.14773 2024

  6. [6]

    M., & Brunner, R

    Ball, N. M., & Brunner, R. J. 2010, International Journal of Modern Physics D, 19, 1049, doi: 10.1142/S0218271810017160

    doi:10.1142/s0218271810017160 2010

  7. [7]

    B., & Ellis, R

    Belli, S., Newman, A. B., & Ellis, R. S. 2019, ApJ, 874, 17, doi: 10.3847/1538-4357/ab07af

    doi:10.3847/1538-4357/ab07af 2019

  8. [8]

    E., et al

    Bezanson, R., Labbe, I., Whitaker, K. E., et al. 2024, ApJ, 974, 92, doi: 10.3847/1538-4357/ad66cf

    doi:10.3847/1538-4357/ad66cf 2024

Show all 110 references
  1. [9]

    2023, astropy/photutils: 1.8.0, 1.8.0, Zenodo, Zenodo, doi: 10.5281/zenodo.7946442

    Bradley, L. 2023, astropy/photutils: 1.8.0, 1.8.0, Zenodo, Zenodo, doi: 10.5281/zenodo.7946442

    2023 doi

  2. [10]

    B., van Dokkum, P

    Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, ApJ, 686, 1503, doi: 10.1086/591786

    2008 doi

  3. [11]

    B., Whitaker, K

    Brammer, G. B., Whitaker, K. E., van Dokkum, P. G., et al. 2011, ApJ, 739, 24, doi: 10.1088/0004-637X/739/1/24

    2011 doi

  4. [12]

    J., Cameron, A

    Bunker, A. J., Cameron, A. J., Curtis-Lake, E., et al. 2024, A&A, 690, A288, doi: 10.1051/0004-6361/202347094

    2024 doi

  5. [13]

    C., et al

    Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682, doi: 10.1086/308692

    2000 doi

  6. [14]

    C., McLure, R

    Carnall, A. C., McLure, R. J., Dunlop, J. S., & Dav´ e, R. 2018, MNRAS, 480, 4379, doi: 10.1093/mnras/sty2169

    2018 doi

  7. [15]

    C., Walker, S., McLure, R

    Carnall, A. C., Walker, S., McLure, R. J., et al. 2020, MNRAS, 496, 695, doi: 10.1093/mnras/staa1535

    2020 doi

  8. [16]

    C., McLeod, D

    Carnall, A. C., McLeod, D. J., McLure, R. J., et al. 2023a, MNRAS, 520, 3974, doi: 10.1093/mnras/stad369

    doi

  9. [17]

    C., McLure, R

    Carnall, A. C., McLure, R. J., Dunlop, J. S., et al. 2023b, Nature, 619, 716, doi: 10.1038/s41586-023-06158-6

    doi

  10. [18]

    2019, ApJL, 880, L14, doi: 10.3847/2041-8213/ab2c80

    Cecchi, R., Bolzonella, M., Cimatti, A., & Girelli, G. 2019, ApJL, 880, L14, doi: 10.3847/2041-8213/ab2c80

    2019 doi

  11. [19]

    Charlot, S., & Fall, S. M. 2000, ApJ, 539, 718, doi: 10.1086/309250

    2000 doi

  12. [20]

    R., et al

    Chartab, N., Mobasher, B., Cooray, A. R., et al. 2023, ApJ, 942, 91, doi: 10.3847/1538-4357/acacf5

    2023 doi

  13. [21]

    2016, MNRAS, 462, 1415, doi: 10.1093/mnras/stw1756

    Chevallard, J., & Charlot, S. 2016, MNRAS, 462, 1415, doi: 10.1093/mnras/stw1756

    2016 doi

  14. [22]

    L., Bandi, B., Philipsborn, S., et al

    Cook, T. L., Bandi, B., Philipsborn, S., et al. 2024, MNRAS, doi: 10.1093/mnras/stae2389

    2024 doi

  15. [23]

    A., Schaye, J., Bower, R

    Crain, R. A., Schaye, J., Bower, R. G., et al. 2015, MNRAS, 450, 1937, doi: 10.1093/mnras/stv725

    2015 doi

  16. [24]

    M., et al

    Dahlen, T., Mobasher, B., Faber, S. M., et al. 2013, ApJ, 775, 93, doi: 10.1088/0004-637X/775/2/93 18

    2013 doi

  17. [25]

    L., et al

    Davidzon, I., Laigle, C., Capak, P. L., et al. 2019, MNRAS, 489, 4817, doi: 10.1093/mnras/stz2486

    2019 doi

  18. [26]

    2022, A&A, 665, A34, doi: 10.1051/0004-6361/202243249 de la Vega, A., Kassin, S

    Davidzon, I., Jegatheesan, K., Ilbert, O., et al. 2022, A&A, 665, A34, doi: 10.1051/0004-6361/202243249 de la Vega, A., Kassin, S. A., Pacifici, C., et al. 2025, arXiv e-prints, arXiv:2501.06297. https://arxiv.org/abs/2501.06297

    2022 arXiv

  19. [27]

    I., Ashby, M

    Deshmukh, S., Caputi, K. I., Ashby, M. L. N., et al. 2018, ApJ, 864, 166, doi: 10.3847/1538-4357/aad9f5 D’Eugenio, C., Daddi, E., Gobat, R., et al. 2020, ApJL, 892, L2, doi: 10.3847/2041-8213/ab7a96 D’Eugenio, F., Cameron, A. J., Scholtz, J., et al. 2024, arXiv e-prints, arXiv...

    2018 doi arXiv

  20. [28]

    S., Abraham, R

    Dunlop, J. S., Abraham, R. G., Ashby, M. L. N., et al. 2021, PRIMER: Public Release IMaging for Extragalactic

    2021

  21. [29]

    J., Willott, C., Alberts, S., et al

    Eisenstein, D. J., Willott, C., Alberts, S., et al. 2023a, arXiv e-prints, arXiv:2306.02465, doi: 10.48550/arXiv.2306.02465

    doi arXiv

  22. [30]

    J., Johnson, B

    Eisenstein, D. J., Johnson, B. D., Robertson, B., et al. 2023b, arXiv e-prints, arXiv:2310.12340, doi: 10.48550/arXiv.2310.12340

    doi

  23. [31]

    J., Stanway, E

    Eldridge, J. J., Stanway, E. R., Xiao, L., et al. 2017, PASA, 34, e058, doi: 10.1017/pasa.2017.51

    2017 doi

  24. [32]

    L., Prakash, A., Capak, P

    Faisst, A. L., Prakash, A., Capak, P. L., & Lee, B. 2019, ApJL, 881, L9, doi: 10.3847/2041-8213/ab3581

    2019 doi

  25. [33]

    J., Chatzikos, M., Guzm´ an, F., et al

    Ferland, G. J., Chatzikos, M., Guzm´ an, F., et al. 2017, RMxAA, 53, 385, doi: 10.48550/arXiv.1705.10877

    2017 doi arXiv

  26. [34]

    P., & Bridges, M

    Feroz, F., Hobson, M. P., & Bridges, M. 2009, MNRAS, 398, 1601, doi: 10.1111/j.1365-2966.2009.14548.x

    2009

  27. [35]

    L., Bagley, M

    Finkelstein, S. L., Bagley, M. B., Arrabal Haro, P., et al. 2022, ApJL, 940, L55, doi: 10.3847/2041-8213/ac966e

    2022 doi

  28. [36]

    L., Bagley, M

    Finkelstein, S. L., Bagley, M. B., Ferguson, H. C., et al. 2023, ApJL, 946, L13, doi: 10.3847/2041-8213/acade4

    2023 doi

  29. [37]

    L., Bagley, M

    Finkelstein, S. L., Bagley, M. B., Arrabal Haro, P., et al. 2025, arXiv e-prints, arXiv:2501.04085, doi: 10.48550/arXiv.2501.04085

    2025 doi arXiv

  30. [38]

    2020, ApJL, 890, L1, doi: 10.3847/2041-8213/ab5b9f

    Forrest, B., Annunziatella, M., Wilson, G., et al. 2020, ApJL, 890, L1, doi: 10.3847/2041-8213/ab5b9f

    2020 doi

  31. [39]

    Kelson, D. D. 2014, ApJ, 788, 72, doi: 10.1088/0004-637X/788/1/72

    2014 doi

  32. [40]

    P., Mather, J

    Gardner, J. P., Mather, J. C., Abbott, R., et al. 2023, PASP, 135, 068001, doi: 10.1088/1538-3873/acd1b5

    2023 doi

  33. [41]

    Geach, J. E. 2012, MNRAS, 419, 2633, doi: 10.1111/j.1365-2966.2011.19913.x

    2012

  34. [42]

    1986, ApJ, 303, 336, doi: 10.1086/164079

    Gehrels, N. 1986, ApJ, 303, 336, doi: 10.1086/164079

    1986 doi

  35. [43]

    C., Koekemoer, A

    Giavalisco, M., Ferguson, H. C., Koekemoer, A. M., et al. 2004, ApJL, 600, L93, doi: 10.1086/379232

    2004 doi

  36. [44]

    2019, A&A, 632, A80, doi: 10.1051/0004-6361/201834547

    Girelli, G., Bolzonella, M., & Cimatti, A. 2019, A&A, 632, A80, doi: 10.1051/0004-6361/201834547

    2019 doi

  37. [45]

    2017, Nature, 544, 71, doi: 10.1038/nature21680

    Glazebrook, K., Schreiber, C., Labb´ e, I., et al. 2017, Nature, 544, 71, doi: 10.1038/nature21680

    2017 doi

  38. [46]

    Gould, K. M. L., Brammer, G., Valentino, F., et al. 2023, AJ, 165, 248, doi: 10.3847/1538-3881/accadc

    2023 doi

  39. [47]

    A., Kocevski, D

    Grogin, N. A., Kocevski, D. D., Faber, S. M., et al. 2011, ApJS, 197, 35, doi: 10.1088/0067-0049/197/2/35

    2011 doi

  40. [48]

    N., Johnson, B

    Hainline, K. N., Johnson, B. D., Robertson, B., et al. 2024, ApJ, 964, 71, doi: 10.3847/1538-4357/ad1ee4

    2024 doi

  41. [49]

    R., Millman, K

    Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357, doi: 10.1038/s41586-020-2649-2

    2020 doi

  42. [50]

    2019, ApJL, 881, L14, doi: 10.3847/2041-8213/ab3418

    Hemmati, S., Capak, P., Pourrahmani, M., et al. 2019, ApJL, 881, L14, doi: 10.3847/2041-8213/ab3418

    2019 doi

  43. [51]

    2023, PASA, 40, e001, doi: 10.1017/pasa.2022.55

    Huertas-Company, M., & Lanusse, F. 2023, PASA, 40, e001, doi: 10.1017/pasa.2022.55

    2023 doi

  44. [52]

    Hunter, J. D. 2007, Computing in Science & Engineering, 9, 90, doi: 10.1109/MCSE.2007.55

    2007 doi

  45. [53]

    B., Magdis, G

    Jin, S., Sillassen, N. B., Magdis, G. E., et al. 2024, A&A, 683, L4, doi: 10.1051/0004-6361/202348540

    2024 doi

  46. [54]

    H., Newman, J

    Kodra, D., Andrews, B. H., Newman, J. A., et al. 2023, ApJ, 942, 36, doi: 10.3847/1538-4357/ac9f12

    2023 doi

  47. [55]

    M., Faber, S

    Koekemoer, A. M., Faber, S. M., Ferguson, H. C., et al. 2011, ApJS, 197, 36, doi: 10.1088/0067-0049/197/2/36

    2011 doi

  48. [56]

    2001, Self-Organizing Maps La Torre, V., Sajina, A., Goulding, A

    Kohonen, T. 2001, Self-Organizing Maps La Torre, V., Sajina, A., Goulding, A. D., et al. 2024, AJ, 167, 261, doi: 10.3847/1538-3881/ad3821 Labb´ e, I., Huang, J., Franx, M., et al. 2005, ApJL, 624, L81, doi: 10.1086/430700

    2001 doi

  49. [57]

    Lagos, C. d. P., Bravo, M., Tobar, R., et al. 2024, MNRAS, 531, 3551, doi: 10.1093/mnras/stae1024

    2024 doi

  50. [58]

    D., Conroy, C., et al

    Leja, J., Johnson, B. D., Conroy, C., et al. 2019, ApJ, 877, 140, doi: 10.3847/1538-4357/ab1d5a

    2019 doi

  51. [59]

    S., Casey, C

    Long, A. S., Casey, C. M., del P. Lagos, C., et al. 2023, ApJ, 953, 11, doi: 10.3847/1538-4357/acddde

    2023 doi

  52. [60]

    S., Antwi-Danso, J., Lambrides, E

    Long, A. S., Antwi-Danso, J., Lambrides, E. L., et al. 2024, ApJ, 970, 68, doi: 10.3847/1538-4357/ad4cea

    2024 doi

  53. [61]

    C., Vijayan, A

    Lovell, C. C., Vijayan, A. P., Thomas, P. A., et al. 2021, MNRAS, 500, 2127, doi: 10.1093/mnras/staa3360

    2021 doi

  54. [62]

    C., Roper, W., Vijayan, A

    Lovell, C. C., Roper, W., Vijayan, A. P., et al. 2023, MNRAS, 525, 5520, doi: 10.1093/mnras/stad2550

    2023 doi

  55. [63]

    2023, MNRAS, 518, 5953, doi: 10.1093/mnras/stac3450

    Lustig, P., Strazzullo, V., Remus, R.-S., et al. 2023, MNRAS, 518, 5953, doi: 10.1093/mnras/stac3450

    2023 doi

  56. [64]

    S., Marchesini, D

    Martis, N. S., Marchesini, D. M., Muzzin, A., et al. 2019, ApJ, 882, 65, doi: 10.3847/1538-4357/ab32f1

    2019 doi

  57. [65]

    2015, ApJ, 813, 53, doi: 10.1088/0004-637X/813/1/53

    Masters, D., Capak, P., Stern, D., et al. 2015, ApJ, 813, 53, doi: 10.1088/0004-637X/813/1/53

    2015 doi

  58. [66]

    2018, arXiv e-prints, arXiv:1802.03426, doi: 10.48550/arXiv.1802.03426 19

    McInnes, L., Healy, J., & Melville, J. 2018, arXiv e-prints, arXiv:1802.03426, doi: 10.48550/arXiv.1802.03426 19

    2018 doi arXiv

  59. [67]

    M., et al

    McKinney, J., Finnerty, L., Casey, C. M., et al. 2023, ApJL, 946, L39, doi: 10.3847/2041-8213/acc322

    2023 doi

  60. [68]

    2018, MNRAS, 473, 2098, doi: 10.1093/mnras/stx2385

    Merlin, E., Fontana, A., Castellano, M., et al. 2018, MNRAS, 473, 2098, doi: 10.1093/mnras/stx2385

    2018 doi

  61. [69]

    2019, MNRAS, 490, 3309, doi: 10.1093/mnras/stz2615

    Merlin, E., Fortuni, F., Torelli, M., et al. 2019, MNRAS, 490, 3309, doi: 10.1093/mnras/stz2615

    2019 doi

  62. [70]

    2024, Scientific Reports, 14, 3724, doi: 10.1038/s41598-024-52585-4

    Nanayakkara, T., Glazebrook, K., Jacobs, C., et al. 2024, Scientific Reports, 14, 3724, doi: 10.1038/s41598-024-52585-4

    2024 doi

  63. [71]

    2014, ApJ, 794, 68, doi: 10.1088/0004-637X/794/1/68

    Nayyeri, H., Mobasher, B., Hemmati, S., et al. 2014, ApJ, 794, 68, doi: 10.1088/0004-637X/794/1/68

    2014 doi

  64. [72]

    B., & Gunn, J

    Oke, J. B., & Gunn, J. E. 1983, ApJ, 266, 713, doi: 10.1086/160817

    1983 doi

  65. [73]

    A., Weiner, B

    Pacifici, C., Kassin, S. A., Weiner, B. J., et al. 2016, ApJ, 832, 79, doi: 10.3847/0004-637X/832/1/79

    2016 doi

  66. [74]

    G., Mobasher, B., et al

    Pacifici, C., Iyer, K. G., Mobasher, B., et al. 2023, ApJ, 944, 141, doi: 10.3847/1538-4357/acacff

    2023 doi

  67. [75]

    2023, ApJ, 953, 119, doi: 10.3847/1538-4357/acd54a

    Park, M., Belli, S., Conroy, C., et al. 2023, ApJ, 953, 119, doi: 10.3847/1538-4357/acd54a

    2023 doi

  68. [76]

    2011, Journal of Machine Learning Research, 12, 2825 Planck Collaboration, Ade, P

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825 Planck Collaboration, Ade, P. A. R., Aghanim, N., et al. 2016, A&A, 594, A13, doi: 10.1051/0004-6361/201525830

    2011 doi

  69. [77]

    J., Robertson, B., Tacchella, S., et al

    Rieke, M. J., Robertson, B., Tacchella, S., et al. 2023, ApJS, 269, 16, doi: 10.3847/1538-4365/acf44d

    2023 doi

  70. [78]

    A., Dobric, N., Adams, N

    Russell, T. A., Dobric, N., Adams, N. J., et al. 2024, arXiv e-prints, arXiv:2412.11861. https://arxiv.org/abs/2412.11861

    2024

  71. [79]

    Salim, S., Boquien, M., & Lee, J. C. 2018, ApJ, 859, 11, doi: 10.3847/1538-4357/aabf3c

    2018 doi

  72. [80]

    2024, arXiv e-prints, arXiv:2410.07354, doi: 10.48550/arXiv.2410.07354

    Sanjaripour, S., Hemmati, S., Mobasher, B., et al. 2024, arXiv e-prints, arXiv:2410.07354, doi: 10.48550/arXiv.2410.07354

    2024 doi arXiv

  73. [81]

    2019, MNRAS, 486, 560, doi: 10.1093/mnras/stz801

    Santini, P., Merlin, E., Fontana, A., et al. 2019, MNRAS, 486, 560, doi: 10.1093/mnras/stz801

    2019 doi

  74. [82]

    A., Bower, R

    Schaye, J., Crain, R. A., Bower, R. G., et al. 2015, MNRAS, 446, 521, doi: 10.1093/mnras/stu2058

    2015 doi

  75. [83]

    2018, A&A, 618, A85, doi: 10.1051/0004-6361/201833070

    Schreiber, C., Glazebrook, K., Nanayakkara, T., et al. 2018, A&A, 618, A85, doi: 10.1051/0004-6361/201833070

    2018 doi

  76. [84]

    J., Khullar, G., Miller, T

    Setton, D. J., Khullar, G., Miller, T. B., et al. 2024, ApJ, 974, 145, doi: 10.3847/1538-4357/ad6a18

    2024 doi

  77. [85]

    A., Lemaux, B., Forrest, B., et al

    Shah, E. A., Lemaux, B., Forrest, B., et al. 2024, MNRAS, 529, 873, doi: 10.1093/mnras/stae519

    2024 doi

  78. [86]

    2020, ApJ, 897, 44, doi: 10.3847/1538-4357/ab96c5

    Shahidi, A., Mobasher, B., Nayyeri, H., et al. 2020, ApJ, 897, 44, doi: 10.3847/1538-4357/ab96c5

    2020 doi

  79. [87]

    2024, arXiv e-prints, arXiv:2408.12713, doi: 10.48550/arXiv.2408.12713

    Shen, Y., Zhuang, M.-Y., Li, J., et al. 2024, arXiv e-prints, arXiv:2408.12713, doi: 10.48550/arXiv.2408.12713

    2024 doi arXiv

  80. [88]

    2024, arXiv e-prints, arXiv:2409.11457, doi: 10.48550/arXiv.2409.11457

    Siegel, J., Setton, D., Greene, J., et al. 2024, arXiv e-prints, arXiv:2409.11457, doi: 10.48550/arXiv.2409.11457

    2024 doi arXiv

  81. [89]

    2021, ApJ, 908, 47, doi: 10.3847/1538-4357/abd179

    Simet, M., Chartab, N., Lu, Y., & Mobasher, B. 2021, ApJ, 908, 47, doi: 10.3847/1538-4357/abd179

    2021 doi

  82. [90]

    E., Whitaker, K

    Skelton, R. E., Whitaker, K. E., Momcheva, I. G., et al. 2014, ApJS, 214, 24, doi: 10.1088/0067-0049/214/2/24

    2014 doi

  83. [91]

    S., Popping, G., & Trager, S

    Somerville, R. S., Popping, G., & Trager, S. C. 2015, MNRAS, 453, 4337, doi: 10.1093/mnras/stv1877

    2015 doi

  84. [92]

    S., Olsen, C., Yung, L

    Somerville, R. S., Olsen, C., Yung, L. Y. A., et al. 2021, MNRAS, 502, 4858, doi: 10.1093/mnras/stab231

    2021 doi

  85. [93]

    R., & Eldridge, J

    Stanway, E. R., & Eldridge, J. J. 2018, MNRAS, 479, 75, doi: 10.1093/mnras/sty1353

    2018 doi

  86. [94]

    L., Capak, P., Masters, D., & Speagle, J

    Steinhardt, C. L., Capak, P., Masters, D., & Speagle, J. S. 2016, ApJ, 824, 21, doi: 10.3847/0004-637X/824/1/21

    2016 doi

  87. [95]

    L., Weaver, J

    Steinhardt, C. L., Weaver, J. R., Maxfield, J., et al. 2020, ApJ, 891, 136, doi: 10.3847/1538-4357/ab76be

    2020 doi

  88. [96]

    L., Finkelstein, S

    Stevans, M. L., Finkelstein, S. L., Kawinwanichakij, L., et al. 2021, ApJ, 921, 58, doi: 10.3847/1538-4357/ac0cf6

    2021 doi

  89. [97]

    Straatman, C. M. S., Spitler, L. R., Quadri, R. F., et al. 2016, ApJ, 830, 51, doi: 10.3847/0004-637X/830/1/51

    2016 doi

  90. [98]

    2014, ApJ, 782, 68, doi: 10.1088/0004-637X/782/2/68

    Toft, S., Smolˇ ci´ c, V., Magnelli, B., et al. 2014, ApJ, 782, 68, doi: 10.1088/0004-637X/782/2/68

    2014 doi

  91. [99]

    2020, ApJ, 889, 93, doi: 10.3847/1538-4357/ab64dc

    Valentino, F., Tanaka, M., Davidzon, I., et al. 2020, ApJ, 889, 93, doi: 10.3847/1538-4357/ab64dc

    2020 doi

  92. [100]

    Valentino, F., Brammer, G., Gould, K. M. L., et al. 2023, ApJ, 947, 20, doi: 10.3847/1538-4357/acbefa van der Maaten, L., & Hinton, G. 2008, J. Mach. Learn. Res.,, 9

    2023 doi

  93. [101]

    2025, MNRAS, 536, 777, doi: 10.1093/mnras/stae2625

    Vani, A., Ayromlou, M., Kauffmann, G., & Springel, V. 2025, MNRAS, 536, 777, doi: 10.1093/mnras/stae2625

    2025 doi

  94. [102]

    P., Lovell, C

    Vijayan, A. P., Lovell, C. C., Wilkins, S. M., et al. 2021, MNRAS, 501, 3289, doi: 10.1093/mnras/staa3715

    2021 doi

  95. [103]

    E., et al

    Virtanen, P., Gommers, R., Oliphant, T. E., et al. 2020, Nature Methods, 17, 261, doi: 10.1038/s41592-019-0686-2

    2020 doi

  96. [104]

    E., Labb´ e, I., van Dokkum, P

    Whitaker, K. E., Labb´ e, I., van Dokkum, P. G., et al. 2011, ApJ, 735, 86, doi: 10.1088/0004-637X/735/2/86

    2011 doi

  97. [105]

    C., Curtis-Lake, E., Hainline, K

    Williams, C. C., Curtis-Lake, E., Hainline, K. N., et al. 2018, ApJS, 236, 33, doi: 10.3847/1538-4365/aabcbb

    2018 doi

  98. [106]

    C., Tacchella, S., Maseda, M

    Williams, C. C., Tacchella, S., Maseda, M. V., et al. 2023, ApJS, 268, 64, doi: 10.3847/1538-4365/acf130

    2023 doi

  99. [107]

    C., Oesch, P

    Williams, C. C., Oesch, P. A., Weibel, A., et al. 2024, arXiv e-prints, arXiv:2410.01875, doi: 10.48550/arXiv.2410.01875

    2024 doi arXiv

  100. [108]

    J., Quadri, R

    Williams, R. J., Quadri, R. F., Franx, M., van Dokkum, P., & Labb´ e, I. 2009, ApJ, 691, 1879, doi: 10.1088/0004-637X/691/2/1879

    2009 doi

  101. [109]

    2019, MNRAS, 483, 2983, doi: 10.1093/mnras/sty3241

    Popping, G., & Dav´ e, R. 2019, MNRAS, 483, 2983, doi: 10.1093/mnras/sty3241

    2019 doi

  102. [110]

    Yung, L. Y. A., Somerville, R. S., Ferguson, H. C., et al. 2022, MNRAS, 515, 5416, doi: 10.1093/mnras/stac2139

    2022 doi

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.