Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Using different sources of ground truths and transfer learning to improve the generalization of photometric redshift estimation

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Mixing photometric and spectroscopic redshifts cuts photo-z bias 5x.

desk verdict A useful new dataset and clean experiments, but the central generalization claim is not established until the authors add a same-framework GalaxiesML-only baseline. read the letter →

arxiv 2411.18054 v1 pith:6L6H2MLZ submitted 2024-11-27 astro-ph.IM astro-ph.GAcs.LG

Jonathan Soriano , Srinath Saikrishnan , Vikram Seenivasan , Bernie Boscoe , Jack Singal , Tuan Do This is my paper · ORCID
classification astro-ph.IMastro-ph.GAcs.LG
keywords photometricredshiftstransferlearninggroundtruthcombinationCOSMOS2020GalaxiesMLneuralnetworkscosmologicalsurveysLSST
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 claims that photometric redshift neural networks can be made to generalize better by training on two complementary sources of real ground truth at once: a broad but imprecise photometric redshift sample (TransferZ, built from the 35-band COSMOS2020 catalog) and a narrower but precise spectroscopic sample (GalaxiesML). Two recipes—transfer learning, where a network pretrained on TransferZ is fine-tuned on GalaxiesML, and joint training on the combined dataset—both improve bias, RMS error, and catastrophic outlier rate on GalaxiesML by roughly 5x, 1.5x, and 1.3x relative to a baseline trained only on TransferZ. The paper argues these gains are enough to meet the redshift accuracy requirements of upcoming surveys like LSST. The cost is a modest worsening of bias and RMS on the TransferZ sample itself, while catastrophic outlier rates on that sample improve.

What carries the argument

The central object is TransferZ, a dataset that pairs five-band HSC grizy photometry from HSC PDR2 with photometric redshift labels taken from the 35-band COSMOS2020 catalog (specifically the LePhare lp_zPDF median of the likelihood). These labels are ~100 times less precise than spectroscopy but cover a much wider and fainter galaxy population. The mechanism that carries the argument is the combination of this broad, imprecise label source with the narrow, precise spectroscopic labels of GalaxiesML, realized in two ways: transfer learning, in which the base network trained on TransferZ is fine-tuned on GalaxiesML with most layers frozen and a learning rate of $5\times10^{-10}$, and joint training, in which a network is trained on the concatenated Combo dataset of 402,408 galaxies. Both recipes force the model to keep the wide coverage learned from photometric redshifts while sharpening its predictions using spectroscopy.

What would settle it

An independent check would be to evaluate NN-TL and NN-Combo on a spectroscopic sample that is distinctly fainter or redder than GalaxiesML (e.g., a deeper spectroscopic survey in the COSMOS field). If the hybrid models show the same improvements over NN-Base on that held-out population, the gains reflect genuine generalization; if the gains shrink or vanish, they are specific to GalaxiesML's color-magnitude range. A second check is to retrain after removing the ~500 galaxies that appear in both TransferZ and GalaxiesML to see whether the reported improvements depend on this overlap.

Watch

Extended reading notes

Core claim

The central discovery is that combining ground truths from different sources—rather than using only the most precise labels available—improves photometric redshift estimation. The authors construct TransferZ, a dataset of 116,335 galaxies with five-band HSC photometry paired with photometric redshifts from the 35-band COSMOS2020 survey (median uncertainty ~0.03), as a source that covers a wider range of galaxy types, magnitudes, and colors than spectroscopic samples. They pair it with GalaxiesML, 286,401 galaxies with spectroscopic redshifts (median uncertainty ~0.0002). Three networks are trained: a base model on TransferZ alone, a transfer-learned model that fine-tunes the base on GalaxiesML, and a combined model trained on both datasets at once. On the GalaxiesML test set both hybrid approaches reduce bias by ~5x, RMS by ~1.5x, and catastrophic outlier rate by ~1.3x compared to the base model, and the paper reports that these results meet cosmological requirements. The combined model slightly outperforms transfer learning on bias and RMS, while transfer learning gives better catastrophic outlier control.

Load-bearing premise

The premise that the photometric redshifts from COSMOS2020 are good enough to act as training truths for the broad galaxy population, despite being about a hundred times less precise than spectroscopy and possibly carrying systematic biases, is load-bearing: if those labels are systematically wrong for some galaxy types, the claimed improvement in generalization could be an illusion.

Editorial extensions

If this is right

  • Photometric redshift models for LSST and similar surveys can be built without waiting for a complete spectroscopic sample: the broad, imprecise photometric labels supply coverage, and a relatively small precise spectroscopic sample anchors accuracy.
  • The two recipes give comparable results, so the choice between them can be driven by the science goal: NN-Combo gives lower bias and RMS on the target sample, while NN-TL gives a lower catastrophic outlier rate.
  • Using photometric redshifts as training labels, despite being ~100x less precise than spectroscopy, improves the model's performance on the spectroscopic test set relative to training on the broad sample alone, indicating that training-set representativeness matters as much as label precision.
  • The released TransferZ dataset lets the community reproduce the hybrid training results and test variations directly.

Reading between the lines

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

  • The reported degradation on the TransferZ test set suggests a trade-off surface between accuracy on the precise narrow sample and accuracy on the broad photometric sample; weighting the two losses or using domain-adaptive training might recover both.
  • The same pretrain-on-photometric, fine-tune-on-spectroscopic recipe could be applied to other surveys where a multi-band photometric catalog with template redshifts overlaps a small spectroscopic calibration sample; the improvement should be tested there.
  • A strong test of the generalization claim is to use the trained models to predict redshifts for galaxies within known clusters: cluster members share a redshift but span many galaxy types, so tight, unbiased predictions would independently validate that the hybrid training generalizes beyond the training color-magnitude range.
  • The assumption that the ~500 overlapping galaxies have negligible impact can be checked directly; if their removal changes the reported factors, part of the apparent improvement is leakage rather than generalization.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript introduces TransferZ, a dataset of 116,335 galaxies with five-band HSC photometry and COSMOS2020 LePhare photometric redshifts, and uses it together with the spectroscopic-redshift GalaxiesML dataset to train photometric redshift networks. Three models are compared: NN-Base trained only on TransferZ, NN-TL initialized on TransferZ and fine-tuned on GalaxiesML, and NN-Combo trained on the combined dataset. On the GalaxiesML test set, both NN-TL and NN-Combo are reported to reduce bias by about 5x, RMS by about 1.5x, and catastrophic outlier rate by about 1.3x relative to NN-Base. The paper also reports that source-side metrics on TransferZ degrade for NN-TL (bias and RMS) and are roughly unchanged for NN-Combo, and concludes that the proposed approaches can meet cosmological requirements for LSST.

Significance. The question addressed is timely and practically relevant: whether broad but less precise photometric-redshift labels can supplement narrow but precise spectroscopic samples for training photo-z models for LSST. The paper's concrete contributions are a publicly released TransferZ dataset (Zenodo DOI) and an evaluation protocol with 100 random initializations, which is a reproducible and honest way to report metric uncertainties. If the generalization claim were supported by a controlled comparison, the result would be valuable for survey preparation. However, the central attribution of the reported gains to the TransferZ source data is not established by the current experimental design, because the target improvements are measured on the same survey used for fine-tuning and no same-framework GalaxiesML-only baseline is provided.

major comments (4)
  1. [Section 4, Table 2] The headline comparison on GalaxiesML compares NN-TL and NN-Combo, both of which are trained or fine-tuned on GalaxiesML spectroscopic redshifts, against NN-Base, which never sees GalaxiesML. The reported 5x bias reduction and 1.5x RMS reduction are therefore expected even if TransferZ contributes nothing, simply from fitting the target labels. A same-architecture, same-loss, same-training-schedule model trained only on GalaxiesML must be added as a baseline before the improvement can be attributed to combining ground truths. Without this control, the 'generalization' claim is unsupported.
  2. [Section 4, Table 2, source rows] The source-side results undercut the generalization narrative: NN-TL on TransferZ shows bias increasing from -0.69e-3 to 7.45e-3 and RMS increasing from 22.6e-3 to 28.5e-3 relative to NN-Base. This indicates catastrophic forgetting of the source features rather than transfer of broadly useful representations. The paper should either provide a mechanism for this degradation, demonstrate generalization on an external survey not used in fine-tuning, or substantially temper the claim that TransferZ improves generalization to the broader galaxy population.
  3. [Section 4, J24 comparison] The text contains a direct contradiction: it first states 'Our transfer learning model performs better than the one from [23]', then states 'The model from J24 achieved a bias of one order magnitude lower than our approach in NN-TL and NN-Combo evaluated on the GalaxiesML.' Please clarify which metric(s) are meant and present a quantitative side-by-side comparison to J24, which is the closest existing target-only baseline and therefore important for interpreting the results.
  4. [Section 2, overlap handling] The paper notes that 500 galaxies are common to TransferZ and GalaxiesML and assumes the impact is negligible. It is not stated explicitly that these overlapping objects are removed from the test sets before splitting. If any of them appear in the GalaxiesML test set and also in the TransferZ training set (or the Combo training set), the target metrics will be optimistically biased. Please state clearly how the overlap was handled in the train/validation/test split, or quantify the effect.
minor comments (5)
  1. [References] References [22] and [23] refer to the same Jones et al. paper and should be consolidated to avoid confusion.
  2. [Figure 3] The label 'Catastrophc Outlier Rate' contains a typo, and the 'LSST Requirements' lines in the figure are not defined in the text; please specify the numerical thresholds used.
  3. [Table 1] The header 'Redshift Median Redshift i-band mag No. Sources 90th percentile Uncertainty 90th percentile Filters' is difficult to parse; please reformat so each column is clearly labeled.
  4. [Section 3] The statement that photometry is 'normalized separately for each training stage' should specify whether validation and test sets are normalized using training-set statistics; otherwise metric comparisons can be affected by a preventable inconsistency.
  5. [Figure 2] Figure 2 shows predictions for the GalaxiesML test set only; given the paper's focus on generalization, showing the analogous TransferZ test panels or explicitly limiting the figure to the target set would improve clarity.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: improvements are measured on held-out test splits and the paper explicitly reports target-supervision costs on source data; minor self-citations are not load-bearing.

full rationale

The core comparison computes bias, RMS, and catastrophic outlier rates on held-out test splits of GalaxiesML and TransferZ after training on disjoint training folds, so the reported improvements are not equivalent by construction to any fitted parameter or label. NN-TL is fine-tuned on GalaxiesML training labels and NN-Combo is trained on the combined training set, and the GalaxiesML test set is a held-out portion of the same survey; this makes the improvement relative to NN-Base (trained only on TransferZ) unsurprising, but it is standard supervised evaluation rather than a circular reduction, and the paper openly reports that NN-TL worsens bias and RMS on TransferZ (Table 2, Section 4), undercutting rather than concealing the generalization story. The main scientific weakness is attribution: without a same-architecture, same-loss GalaxiesML-only baseline, the gain cannot be cleanly attributed to the TransferZ source data, and the paper itself concedes that J24, trained only on GalaxiesML with a different framework, achieves lower bias and that 'additional tests of how well this model generalizes will be important to validate the improvements' (Section 4). This is a limitation in external validity, not circularity. Self-citations to GalaxiesML [15] and Singal et al. [45] are data/method citations and are not load-bearing assumptions that reduce the derivation to themselves. No equation or fitted parameter is defined in terms of the claimed result, so per the stated criteria the paper is self-contained enough for a 0-2 score.

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

The central claim rests on the reliability of external photometric-redshift catalogs (COSMOS2020/LePhare and HSC PDR2) and on hand-chosen hyperparameters and quality cuts. No new physical entities are introduced. The listed free parameters are training choices and data-selection thresholds, not fitted physical constants.

free parameters (5)
  • Transfer learning learning rate = 5e-10
    Chosen by hand for fine-tuning NN-TL; not systematically tuned, affecting the reported target-data improvements.
  • Frozen layer configuration = freeze all layers except input, first, and fifth dense layers
    Hand-selected for transfer learning; no ablation justifying this choice.
  • Loss scale in custom loss = 0.15
    Adopted from Tanaka et al. [47] without re-fitting; controls the loss shape and thus the training dynamics.
  • Quality cut threshold chi^2 < 1 = 1
    Adopted from [45]; defines which galaxies enter TransferZ and thus affects all downstream results.
  • Photo-z agreement threshold |lp_zPDF - lp_zMinChi2| < 0.1 = 0.1
    Hand-chosen quality cut in TransferZ construction; affects the training sample composition.
assumptions (4)
  • domain assumption LePhare photometric redshifts from COSMOS2020 CLASSIC are reliable enough to serve as ground truth for training and evaluation on TransferZ
    The entire TransferZ label set is lp_zPDF from LePhare; if these are systematically biased, the base model learns a biased mapping (Section A.1).
  • domain assumption The overlap of ~500 galaxies between TransferZ and GalaxiesML is negligible
    Authors state this assumption; if false, the target-data improvements could include memorization of shared galaxies (Section 2).
  • domain assumption HSC PDR2 grizy photometry is a sufficient feature set to predict redshifts
    The models use only 5-band photometry, approximating LSST; if the bands lack information, the models cannot generalize (Section 2).
  • standard math Standard i.i.d. train/test split assumption
    Random 80/10/10 splits assume the test distribution matches the training distribution within each dataset; this is standard ML practice (Section 2).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Using different sources of ground truths and transfer learning to improve the generalization of photometric redshift estimation." pith.science (2026). https://pith.science/paper/6L6H2MLZ

@misc{pith2026241118054,
  author       = {Pith},
  title        = {Pith review of: Using different sources of ground truths and transfer learning to improve the generalization of photometric redshift estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6L6H2MLZ}},
  note         = {Machine review of arXiv:2411.18054}
}
abstract

In this work, we explore methods to improve galaxy redshift predictions by combining different ground truths. Traditional machine learning models rely on training sets with known spectroscopic redshifts, which are precise but only represent a limited sample of galaxies. To make redshift models more generalizable to the broader galaxy population, we investigate transfer learning and directly combining ground truth redshifts derived from photometry and spectroscopy. We use the COSMOS2020 survey to create a dataset, TransferZ, which includes photometric redshift estimates derived from up to 35 imaging filters using template fitting. This dataset spans a wider range of galaxy types and colors compared to spectroscopic samples, though its redshift estimates are less accurate. We first train a base neural network on TransferZ and then refine it using transfer learning on a dataset of galaxies with more precise spectroscopic redshifts (GalaxiesML). In addition, we train a neural network on a combined dataset of TransferZ and GalaxiesML. Both methods reduce bias by $\sim$ 5x, RMS error by $\sim$ 1.5x, and catastrophic outlier rates by 1.3x on GalaxiesML, compared to a baseline trained only on TransferZ. However, we also find a reduction in performance for RMS and bias when evaluated on TransferZ data. Overall, our results demonstrate these approaches can meet cosmological requirements.

Figures

Figures reproduced from arXiv: 2411.18054 by the authors.

Figure 1
Figure 1. Two datasets: GalaxiesML [15] with spectroscopic redshift ground truth and TransferZ with [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Comparison of redshift predictions from the three neural network models in this work (NN [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. From left to right, comparison of the bias, outlier and RMS metrics between the baseline [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Flow chart showing the steps used in creating the TransferZ dataset. Green rectangles [PITH_FULL_IMAGE:figures/full_fig_p010_4.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. Enhancing Photometric Redshift Estimation for LSST with a Hybrid LSTM-Mixture Density Network

    astro-ph.GA 2026-07 conditional novelty 5.5 of 10

    LSTM-MDNz improves photometric redshift point estimates and calibrated PDFs over a BNN baseline on HSC GalaxiesML, cutting outliers ~20% and enabling high-purity catalogs via z_conf.

Reference graph

Works this paper leans on

53 extracted references · 15 canonical work pages · cited by 1 Pith paper

  1. [23]

    Improving Photometric Redshift Estimation for Cosmology with LSST Using Bayesian Neural Networks

    Evan Jones et al. “Improving Photometric Redshift Estimation for Cosmology with LSST Using Bayesian Neural Networks”. In: The Astrophysical Journal 964 (Apr. 2024), p. 130. ISSN : 0004-637X. DOI: 10.3847/1538-4357/ad2070

    doi:10.3847/1538-4357/ad2070 2024

  2. [1]

    Second Data Release of the Hyper Suprime-Cam Subaru Strategic Program

    Hiroaki Aihara et al. “Second Data Release of the Hyper Suprime-Cam Subaru Strategic Program”. In: Publications of the Astronomical Society of Japan71.6 (Dec. 2019), p. 114. ISSN : 0004-6264, 2053-051X. DOI: 10.1093/pasj/psz103. arXiv: 1905.12221 [astro-ph]

    arXiv 2019

  3. [2]

    Measuring and Modelling the Redshift Evolution of Clustering: The Hubble Deep Field North

    S. Arnouts et al. “Measuring and Modelling the Redshift Evolution of Clustering: The Hubble Deep Field North”. In: Monthly Notices of the Royal Astronomical Society 310.2 (Dec. 1999), pp. 540–556. ISSN : 0035-8711. DOI: 10.1046/j.1365-8711.1999.02978.x

    arXiv 1999

  4. [3]

    Bayesian Photometric Redshift Estimation

    Narciso Benítez. “Bayesian Photometric Redshift Estimation”. In: The Astrophysical Journal 536 (June 2000), pp. 571–583. ISSN : 0004-637X. DOI: 10.1086/308947

    doi:10.1086/308947 2000

  5. [4]

    Cosmology with Photometric Redshift Surveys

    Chris Blake and Sarah Bridle. “Cosmology with Photometric Redshift Surveys”. In: Monthly Notices of the Royal Astronomical Society363.4 (Nov. 2005), pp. 1329–1348.ISSN : 0035-8711. DOI: 10.1111/j.1365-2966.2005.09526.x

    arXiv 2005

  6. [5]

    Using Neural Networks to Estimate Redshift Distributions. An Applica- tion to CFHTLenS

    Christopher Bonnett. “Using Neural Networks to Estimate Redshift Distributions. An Applica- tion to CFHTLenS”. In: Monthly Notices of the Royal Astronomical Society 449.1 (May 2015), pp. 1043–1056. ISSN : 0035-8711. DOI: 10.1093/mnras/stv230

    doi:10.1093/mnras/stv230 2015

  7. [6]

    High-Velocity Outflows from Young Star-Forming Galaxies in the UKIDSS Ultra-Deep Survey

    E. J. Bradshaw et al. “High-Velocity Outflows from Young Star-Forming Galaxies in the UKIDSS Ultra-Deep Survey”. In: Monthly Notices of the Royal Astronomical Society 433 (July 2013), pp. 194–208. ISSN : 0035-8711. DOI: 10.1093/mnras/stt715

    doi:10.1093/mnras/stt715 2013

  8. [7]

    EAZY: A Fast, Public Photometric Redshift Code

    Gabriel B. Brammer, Pieter G. Van Dokkum, and Paolo Coppi. “EAZY: A Fast, Public Photometric Redshift Code”. en. In: The Astrophysical Journal 686.2 (Oct. 2008), pp. 1503–

    work page 2008

Show all 53 references
  1. [8]

    From Data to Software to Science with the Rubin Observatory LSST

    Katelyn Breivik et al. From Data to Software to Science with the Rubin Observatory LSST. Aug. 2022

    2022

  2. [9]

    Photometric Classification of Quasars from RCS-2 Using Random Forest

    D. Carrasco et al. “Photometric Classification of Quasars from RCS-2 Using Random Forest”. en. In: Astronomy & Astrophysics 584 (Dec. 2015), A44. ISSN : 0004-6361, 1432-0746. DOI: 10.1051/0004-6361/201525752

    2015 doi

  3. [10]

    The PRIsm MUlti-Object Survey (PRIMUS) I: Survey Overview and Characteristics

    Alison L. Coil et al. “The PRIsm MUlti-Object Survey (PRIMUS) I: Survey Overview and Characteristics”. In: The Astrophysical Journal 741.1 (Nov. 2011), p. 8. ISSN : 0004-637X, 1538-4357. DOI: 10.1088/0004-637X/741/1/8. arXiv: 1011.4307 [astro-ph]

    2011 arXiv

  4. [11]

    Euclid Collaboration et al. Euclid. I. Overview of the Euclid Mission . May 2024. DOI: 10. 48550/arXiv.2405.13491. arXiv: 2405.13491 [astro-ph]

    2024 doi

  5. [12]

    ANNz: Estimating Photometric Redshifts Using Artificial Neural Networks

    Adrian A. Collister and Ofer Lahav. “ANNz: Estimating Photometric Redshifts Using Artificial Neural Networks”. en. In: Publications of the Astronomical Society of the Pacific 116.818 (Mar. 2004), p. 345. ISSN : 1538-3873. DOI: 10.1086/383254

    2004 doi

  6. [13]

    The PRIsm MUlti-object Survey (PRIMUS). II. Data Reduction and Redshift Fitting

    Richard J. Cool et al. “The PRIsm MUlti-object Survey (PRIMUS). II. Data Reduction and Redshift Fitting”. In: The Astrophysical Journal 767.2 (Apr. 2013), p. 118. ISSN : 0004-637X, 1538-4357. DOI: 10.1088/0004-637X/767/2/118. arXiv: 1303.2672 [astro-ph]

    2013 arXiv

  7. [14]

    Science Objectives and Early Results of the DEEP2 Redshift Survey

    Marc Davis et al. “Science Objectives and Early Results of the DEEP2 Redshift Survey”. In: 4834 (Feb. 2003), pp. 161–172. DOI: 10.1117/12.457897

    2003 doi

  8. [15]

    GalaxiesML: An Imaging and Photometric Dataset of Galaxies for Machine Learning

    Tuan Do et al. GalaxiesML: An Imaging and Photometric Dataset of Galaxies for Machine Learning. May 2024. DOI: 10.5281/zenodo.11117528

    2024 doi

  9. [16]

    The PAU Survey: Photometric Redshifts Using Transfer Learning from Simulations

    M Eriksen et al. “The PAU Survey: Photometric Redshifts Using Transfer Learning from Simulations”. In: Monthly Notices of the Royal Astronomical Society 497.4 (Oct. 2020), pp. 4565–4579. ISSN : 0035-8711. DOI: 10.1093/mnras/staa2265

    2020 doi

  10. [17]

    The VIMOS Public Extragalactic Survey (VIPERS): First Data Release of 57 204 Spectroscopic Measurements

    B. Garilli et al. “The VIMOS Public Extragalactic Survey (VIPERS): First Data Release of 57 204 Spectroscopic Measurements”. en. In: Astronomy & Astrophysics 562 (Feb. 2014), A23. ISSN : 0004-6361, 1432-0746. DOI: 10.1051/0004-6361/201322790

    2014 doi

  11. [18]

    Photometric Redshifts with the LSST. II. The Impact of Near- infrared and Near-ultraviolet Photometry

    Melissa L. Graham et al. “Photometric Redshifts with the LSST. II. The Impact of Near- infrared and Near-ultraviolet Photometry”. en. In: The Astronomical Journal 159.6 (May 2020), p. 258. ISSN : 1538-3881. DOI: 10.3847/1538-3881/ab8a43

    2020 doi

  12. [19]

    Accurate Photometric Redshifts for the CFHT Legacy Survey Calibrated Using the VIMOS VLT Deep Survey

    O. Ilbert et al. “Accurate Photometric Redshifts for the CFHT Legacy Survey Calibrated Using the VIMOS VLT Deep Survey”. en. In:Astronomy & Astrophysics 457.3 (Oct. 2006), pp. 841–856. ISSN : 0004-6361, 1432-0746. DOI: 10.1051/0004-6361:20065138

    2006 doi

  13. [20]

    COSMOS PHOTOMETRIC REDSHIFTS WITH 30-BANDS FOR 2-Deg2

    O. Ilbert et al. “COSMOS PHOTOMETRIC REDSHIFTS WITH 30-BANDS FOR 2-Deg2”. en. In: The Astrophysical Journal 690.2 (Dec. 2008), p. 1236. ISSN : 0004-637X. DOI: 10. 1088/0004-637X/690/2/1236. 6

    2008

  14. [21]

    LSST: From Science Drivers to Reference Design and Anticipated Data Products

    Željko Ivezi´c et al. “LSST: From Science Drivers to Reference Design and Anticipated Data Products”. en. In: (May 2008). DOI: 10.3847/1538-4357/ab042c

    2008 doi

  15. [24]

    Photometric Redshifts for Cosmology: Improving Accuracy and Uncertainty Estimates Using Bayesian Neural Networks

    Evan Jones et al. Photometric Redshifts for Cosmology: Improving Accuracy and Uncertainty Estimates Using Bayesian Neural Networks. Feb. 2022. DOI: 10.48550/arXiv.2202.07121. arXiv: 2202.07121 [astro-ph]

    2022 doi arXiv

  16. [25]

    THE COSMOS2015 CATALOG: EXPLORING THE 1 < z < 6 UNIVERSE WITH HALF A MILLION GALAXIES

    C. Laigle et al. “THE COSMOS2015 CATALOG: EXPLORING THE 1 < z < 6 UNIVERSE WITH HALF A MILLION GALAXIES”. en. In: The Astrophysical Journal Supplement Series 224.2 (June 2016), p. 24. ISSN : 0067-0049. DOI: 10.3847/0067-0049/224/2/24

    2016 doi

  17. [26]

    Hogg, and David Mykytyn.The Tractor: Probabilistic astronomical source detection and measurement

    Dustin Lang, David W. Hogg, and David Mykytyn.The Tractor: Probabilistic astronomical source detection and measurement. Astrophysics Source Code Library, record ascl:1604.008. Apr. 2016

    2016

  18. [27]

    The VIMOS VLT Deep Survey Final Data Release: A Spectroscopic Sample of 35 016 Galaxies and AGN out to z ~ 6.7 Selected with 17.5 ≤ iAB ≤ 24.75

    O. Le Fèvre et al. “The VIMOS VLT Deep Survey Final Data Release: A Spectroscopic Sample of 35 016 Galaxies and AGN out to z ~ 6.7 Selected with 17.5 ≤ iAB ≤ 24.75”. en. In: Astronomy and Astrophysics 559 (Nov. 2013), A14. ISSN : 0004-6361. DOI: 10.1051/0004- 6361/201322179

    2013 doi

  19. [28]

    Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization

    Lisha Li et al. “Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization”. In: Journal of Machine Learning Research 18.185 (2018), pp. 1–52. ISSN : 1533-7928

    2018

  20. [29]

    The zCOSMOS 10k-Bright Spectroscopic Sample

    Simon J. Lilly et al. “The zCOSMOS 10k-Bright Spectroscopic Sample”. In:The Astrophysical Journal Supplement Series 184 (Oct. 2009), pp. 218–229. ISSN : 0067-0049. DOI: 10.1088/ 0067-0049/184/2/218

    2009

  21. [30]

    Galaxy And Mass Assembly (GAMA): End of Survey Report and Data Release 2

    J. Liske et al. “Galaxy And Mass Assembly (GAMA): End of Survey Report and Data Release 2”. In: Monthly Notices of the Royal Astronomical Society 452 (Sept. 2015), pp. 2087–2126. ISSN : 0035-8711. DOI: 10.1093/mnras/stv1436

    2015 doi

  22. [31]

    UltraVISTA: A New Ultra-Deep near-Infrared Survey in COSMOS

    H. J. McCracken et al. “UltraVISTA: A New Ultra-Deep near-Infrared Survey in COSMOS”. en. In: Astronomy & Astrophysics 544 (Aug. 2012), A156. ISSN : 0004-6361, 1432-0746. DOI: 10.1051/0004-6361/201219507

    2012 doi

  23. [32]

    The Sizes, Masses and Specific Star-Formation Rates of Massive Galaxies at 1.3

    R. J. McLure et al. “The Sizes, Masses and Specific Star-Formation Rates of Massive Galaxies at 1.3”. en. In: (May 2012). DOI: 10.1093/mnras/sts092

    2012 doi

  24. [33]

    On-Sky Characterisation of the VISTA NB118 Narrow-Band Filters at 1.19 µm

    Bo Milvang-Jensen et al. “On-Sky Characterisation of the VISTA NB118 Narrow-Band Filters at 1.19 µm”. en. In: Astronomy & Astrophysics 560 (Dec. 2013), A94. ISSN : 0004-6361, 1432-0746. DOI: 10.1051/0004-6361/201321814

    2013 doi

  25. [34]

    The 3D-HST Survey: Hubble Space Telescope WFC3/G141 Grism Spectra, Redshifts, and Emission Line Measurements for $\sim 100,000$ Galaxies

    Ivelina G. Momcheva et al. “The 3D-HST Survey: Hubble Space Telescope WFC3/G141 Grism Spectra, Redshifts, and Emission Line Measurements for $\sim 100,000$ Galaxies”. In: The Astrophysical Journal Supplement Series 225.2 (Aug. 2016), p. 27. ISSN : 1538-4365. DOI: 10.3847/0067-...

    2016 arXiv

  26. [35]

    Euclid Preparation - XVII. Cosmic Dawn Survey: Spitzer Space Telescope Observations of the Euclid Deep Fields and Calibration Fields

    A. Moneti et al. “Euclid Preparation - XVII. Cosmic Dawn Survey: Spitzer Space Telescope Observations of the Euclid Deep Fields and Calibration Fields”. en. In: Astronomy & As- trophysics 658 (Feb. 2022), A126. ISSN : 0004-6361, 1432-0746. DOI: 10 . 1051 / 0004 - 6361/202142361

    2022

  27. [36]

    Improving Photometric Redshift Estimates with Training Sample Augmentation

    Irene Moskowitz et al. “Improving Photometric Redshift Estimates with Training Sample Augmentation”. In: The Astrophysical Journal 967 (May 2024), p. L6. ISSN : 0004-637X. DOI: 10.3847/2041-8213/ad4039

    2024 doi

  28. [37]

    Photometric Redshifts for Next-Generation Surveys

    Jeffrey A. Newman and Daniel Gruen. “Photometric Redshifts for Next-Generation Surveys”. In: Annual Review of Astronomy and Astrophysics 60.1 (2022), pp. 363–414. DOI: 10.1146/ annurev-astro-032122-014611

    2022

  29. [38]

    The DEEP2 Galaxy Redshift Survey: Design, Observations, Data Reduction, and Redshifts

    Jeffrey A. Newman et al. “The DEEP2 Galaxy Redshift Survey: Design, Observations, Data Reduction, and Redshifts”. In: The Astrophysical Journal Supplement Series 208.1 (Aug. 2013), p. 5. ISSN : 0067-0049, 1538-4365. DOI: 10.1088/0067- 0049/208/1/5 . arXiv: 1203.3192 [astro-ph]. 7

    2013 arXiv

  30. [39]

    A Survey on Transfer Learning

    Sinno Jialin Pan and Qiang Yang. “A Survey on Transfer Learning”. In:IEEE Transactions on Knowledge and Data Engineering 22.10 (Oct. 2010), pp. 1345–1359. ISSN : 1558-2191. DOI: 10.1109/TKDE.2009.191

    2010 doi

  31. [40]

    Photometric Redshifts from SDSS Images Using a Convolutional Neural Network

    Johanna Pasquet et al. “Photometric Redshifts from SDSS Images Using a Convolutional Neural Network”. In: Astronomy & Astrophysics 621 (Jan. 2019), A26. ISSN : 0004-6361, 1432-0746. DOI: 10.1051/0004-6361/201833617

    2019 doi

  32. [41]

    The Euclid Mission Design

    Giuseppe D. Racca et al. “The Euclid Mission Design”. In: July 2016, 99040O. DOI: 10. 1117/12.2230762. arXiv: 1610.05508 [astro-ph]

    2016 arXiv

  33. [42]

    S-COSMOS: The Spitzer Legacy Survey of the Hubble Space Telescope ACS 2 Deg2 COSMOS Field I: Survey Strategy and First Analysis*

    D. B. Sanders et al. “S-COSMOS: The Spitzer Legacy Survey of the Hubble Space Telescope ACS 2 Deg2 COSMOS Field I: Survey Strategy and First Analysis*”. en. In:The Astrophysical Journal Supplement Series 172.1 (Sept. 2007), p. 86. ISSN : 0067-0049. DOI: 10 . 1086 / 517885

    2007

  34. [43]

    The CFHT Large Area U-band Deep Survey (CLAUDS)

    Marcin Sawicki et al. “The CFHT Large Area U-band Deep Survey (CLAUDS)”. In: Monthly Notices of the Royal Astronomical Society489.4 (Nov. 2019), pp. 5202–5217.ISSN : 0035-8711. DOI: 10.1093/mnras/stz2522

    2019 doi

  35. [44]

    The Cosmic Evolution Survey (COSMOS): Overview

    N. Scoville et al. “The Cosmic Evolution Survey (COSMOS): Overview”. In:The Astrophysical Journal Supplement Series 172 (Sept. 2007), pp. 1–8. ISSN : 0067-0049. DOI: 10 . 1086 / 516585

    2007

  36. [45]

    Machine Learning Classification to Identify Catastrophic Outlier Photometric Redshift Estimates

    J. Singal et al. “Machine Learning Classification to Identify Catastrophic Outlier Photometric Redshift Estimates”. en. In: The Astrophysical Journal 928.1 (Mar. 2022), p. 6. ISSN : 0004- 637X. DOI: 10.3847/1538-4357/ac53b5

    2022 doi

  37. [46]

    3D-HST WFC3-selected Photometric Catalogs in the Five CANDELS/3D-HST Fields: Photometry, Photometric Redshifts, and Stellar Masses

    Rosalind E. Skelton et al. “3D-HST WFC3-selected Photometric Catalogs in the Five CANDELS/3D-HST Fields: Photometry, Photometric Redshifts, and Stellar Masses”. In: The Astrophysical Journal Supplement Series 214 (Oct. 2014), p. 24. ISSN : 0067-0049. DOI: 10.1088/0067-0049/214/2/24

    2014 doi

  38. [47]

    Photometric Redshifts for Hyper Suprime-Cam Subaru Strategic Program Data Release 1

    Masayuki Tanaka et al. “Photometric Redshifts for Hyper Suprime-Cam Subaru Strategic Program Data Release 1”. In: Publications of the Astronomical Society of Japan 70.SP1 (Jan. 2018), S9. ISSN : 0004-6264. DOI: 10.1093/pasj/psx077

    2018 doi

  39. [48]

    The Cosmic Evolution Survey (COSMOS): Subaru Observations of the HST Cosmos Field*

    Y . Taniguchi et al. “The Cosmic Evolution Survey (COSMOS): Subaru Observations of the HST Cosmos Field*”. en. In: The Astrophysical Journal Supplement Series 172.1 (Sept. 2007), p. 9. ISSN : 0067-0049. DOI: 10.1086/516596

    2007 doi

  40. [49]

    The Subaru COSMOS 20: Subaru Optical Imaging of the HST COSMOS Field with 20 Filters*

    Yoshiaki Taniguchi et al. “The Subaru COSMOS 20: Subaru Optical Imaging of the HST COSMOS Field with 20 Filters*”. In: Publications of the Astronomical Society of Japan 67.6 (Dec. 2015), p. 104. ISSN : 0004-6264. DOI: 10.1093/pasj/psv106

    2015 doi

  41. [50]

    CNN Photometric Redshifts in the SDSS at r≤ 20

    M. Treyer et al. “CNN Photometric Redshifts in the SDSS at r≤ 20”. In: Monthly Notices of the Royal Astronomical Society (Oct. 2023). ISSN : 0035-8711. DOI: 10.1093/mnras/stad3171

    2023 doi

  42. [51]

    COSMOS2020: A Panchromatic View of the Universe to z 10 from Two Complementary Catalogs

    J. R. Weaver et al. “COSMOS2020: A Panchromatic View of the Universe to z 10 from Two Complementary Catalogs”. In: The Astrophysical Journal Supplement Series 258 (Jan. 2022), p. 11. ISSN : 0067-0049. DOI: 10.3847/1538-4365/ac3078

    2022 doi

  43. [52]

    A Survey of Transfer Learning

    Karl Weiss, Taghi M. Khoshgoftaar, and DingDing Wang. “A Survey of Transfer Learning”. In: Journal of Big Data 3.1 (May 2016), p. 9. ISSN : 2196-1115. DOI: 10.1186/s40537-016- 0043-6

    2016 doi

  44. [53]

    Deep GALEX Imaging of the COSMOS HST Field: A First Look at the Morphology of z ~ 0.7 Star-forming Galaxies

    M. A. Zamojski et al. “Deep GALEX Imaging of the COSMOS HST Field: A First Look at the Morphology of z ~ 0.7 Star-forming Galaxies”. en. In: The Astrophysical Journal Supplement Series 172.1 (Sept. 2007), p. 468. ISSN : 0067-0049. DOI: 10.1086/516593. A Data Creation A.1 COSMO...

    2007 doi

  45. [1513]

    DOI: 10.1086/591786

    ISSN : 0004-637X, 1538-4357. DOI: 10.1086/591786

    doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.