Pith. sign in

REVIEW 2 major objections 5 minor 2 cited by

Selection of optically variable active galactic nuclei via a random forest algorithm

T0 review · 2 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that reducing a random forest from 162 features to eight—chosen to separate obscured AGN from inactive galaxies—raises the recall of known obscured AGN to about 68%, nearly doubling the earlier result.

desk verdict Useful g-band extension of the VST-COSMOS RF pipeline, but the headline 68% obscured-AGN recall is likely inflated by feature selection on the full labeled set before LOOCV. read the letter →

arxiv 2505.15819 v1 pith:MUBOVK6B submitted 2025-05-21 astro-ph.GA astro-ph.IM

Demetra De Cicco , Gaetano Zazzaro , Stefano Cavuoti , Maurizio Paolillo , Giuseppe Longo , Vincenzo Petrecca , Ivano Saccheo , Paula Sánchez-Sáez This is my paper · ORCID
classification astro-ph.GAastro-ph.IM
keywords activegalacticnucleirandomforestopticalvariabilityobscuredAGNmultibandphotometryfeatureselectionCOSMOSfieldVLTSurveyTelescope
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

This paper asks whether optical variability alone—the stochastic brightness changes that set active galactic nuclei (AGN) apart from quiet stars and galaxies—can be turned into a more complete census of AGN, especially the obscured ones whose light is dimmed and reddened by dust. The authors train random forest classifiers on 2,543 labeled sources from the COSMOS field observed with the VLT Survey Telescope in $r$ and $g$ bands, starting from 162 variability, color, and morphology features. They find that most of those features are unnecessary: a classifier built from only eight features, chosen specifically to separate obscured AGN from inactive galaxies, recovers ($68.1\pm1.2$)% of known obscured AGN, nearly doubling the recall of the earlier version of the same pipeline, while still finding about 99% of unobscured AGN. The paper presents this as a step toward variability-based AGN selection in upcoming wide-field surveys, where completeness for obscured AGN has been a known weakness.

What carries the argument

The engine of the result is a two-stage pruning procedure. Stage one is a Kolmogorov-Smirnov screen: for each of the 162 features, compare the distribution over obscured AGN with the distribution over inactive galaxies, and keep the 25 with distance $D>0.25$. Stage two is iterative elimination: train a random forest with class weights balanced, rank features by impurity-based importance, delete the least important, retrain, and repeat down to seven features. The random forest itself is standard, with leave-one-out cross-validation used to score each setting; the load-bearing element is the choice of the final eight features, which the paper argues separate obscured AGN from inactive galaxies better than the original 162-feature set. The K-S screen also shows which features change most when synthetic visits replace real ones, linking the feature selection to the cadence tests.

What would settle it

Take the ks8 feature set to an independent sample of spectroscopically classified AGN from a different survey and measure the obscured-AGN recall; if it falls to roughly 50 percent, the level of the single-band classifiers here, the 68 percent gain is mostly selection on the same data used for scoring rather than a portable advantage.

Watch

Extended reading notes

Core claim

The paper's central claim is that the difficulty of finding obscured AGN through optical variability is primarily a feature-selection problem, not a data problem. By measuring, for every feature, the Kolmogorov-Smirnov distance between the feature distributions of obscured AGN and inactive galaxies, the authors isolate 25 features worth keeping and then iteratively strip the least important ones until eight remain: the mid-infrared color ch21, three optical colors ($u-B$, $r-i$, $i-z$), the HST stellarity index, and three $r$-band variability features (ExcessVar, GP_DRW_sigma, GP_DRW_tau). On the same labeled set, the eight-feature random forest returns ($68.1\pm1.2$)% recall for obscured AGN and 99.1% recall for unobscured AGN, at the cost of lower precision than feature-rich classifiers. The paper also reports that bivariate features combining $r$ and $g$ light curves add no benefit, that $g$-band features vanish from the final selection, and that replacing up to half of a light curve's visits with linearly interpolated synthetic points does not significantly change the results. These findings are presented as evidence that variability selection can be made competitive for obscured AGN and ready for application to LSST-era data.

Load-bearing premise

The performance numbers assume that choosing the best 25 and then eight features on the full labeled set and evaluating those same sources with leave-one-out cross-validation yields an unbiased estimate of classifier performance on new data.

Editorial extensions

If this is right

  • Obscured AGN completeness is not locked to survey depth: better feature selection on the same light curves can raise recall from roughly 50% to about 68%.
  • The final eight-feature set—two variability timescales, an excess-variance measure, four colors, and morphology—is a concrete recipe to test on other surveys.
  • For LSST-like monitoring, the result implies that a modest number of well-chosen features should be used when completeness for obscured AGN is the goal, rather than the full feature zoo.
  • Because synthetic imputed visits did not degrade the classifiers, observing strategies that do not sample two bands simultaneously can still feed the same selection method.
  • The higher obscured-AGN recall is traded against higher contamination, so samples built with the eight-feature classifier will need follow-up to keep purity.

Reading between the lines

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

  • Editor's inference: because the 25-then-8 feature choice was made on the same labeled set later used for leave-one-out evaluation, the 68.1% recall figure is likely an optimistic in-sample estimate; an independent transfer test could settle the true performance.
  • Editor's inference: the dominance of $r$-band variability features in the final set, despite the $g$-band being bluer and generally expected to show larger AGN variability, hints that the advantage comes more from the $r$-band's denser real sampling than from physics; a redder-band version of the experiment would test this.
  • Editor's inference: if the eight-feature set transfers, a practical extension would be to use it as a prior for anomaly detection in streaming alert streams from wide-field surveys, flagging obscured-AGN candidates before spectra are taken.
  • Editor's inference: the K-S distance table provides a feature ranking that could be repurposed as a cheap pre-filter to remove inactive galaxies, reducing the labeled-set size needed for training on new surveys.
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

2 major / 5 minor

Summary. The paper presents a random forest classification pipeline for selecting AGN from VST-COSMOS optical light curves, using 162 variability, color, and morphology features measured in the r and g bands, with imputation to 33 visits per band. The labeled set contains 380 AGN (217 Type I, 104 Type II, 211 MIR-selected) and 2163 non-AGN. Using leave-one-out cross-validation (LOOCV), the authors compare several feature sets and then, in Sect. 4.2, use K-S feature selection and iterative elimination to build a compact eight-feature classifier (ks8) that reaches 68.1% recall for spectroscopically confirmed obscured AGN, described as almost doubling the previous result from De Cicco et al. (2021). The paper also investigates the effect of replacing real visits with synthetic imputed visits and finds no large impact on the main metrics.

Significance. If the 68.1% obscured-AGN recall were an unbiased estimate, the result would be genuinely useful for time-domain AGN surveys, since completeness for obscured AGN is a known weakness of optical variability selection. The study also contributes a broad feature set, careful class-imbalance handling, and a transparent comparison across classifiers. However, the headline performance is not yet supported because feature selection and model selection are performed on the full labeled set before LOOCV; the reported recall is an in-sample selection metric. With a nested cross-validation or external test, the method could be valuable for LSST-era applications.

major comments (2)
  1. [Sect. 4.2 and Table 4] The reported 68.1% obscured-AGN recall is likely an optimistic in-sample figure because feature selection and model selection use the full labeled set before LOOCV. Specifically, the K-S test between obscured AGN and inactive galaxies selects 25 of 162 features, the feature-importance ranking is computed from an RF trained on the same 2543 sources, and the iterative elimination from 25 to 8 features is driven by the recall values shown in Table 4. LOOCV is then applied only to the already-determined ks8 feature set, so every held-out source has contributed to the choice of features and to the number of features retained. Because the K-S pre-filter was explicitly optimized to separate obscured AGN from inactive galaxies, the bias is likely largest for the obscured-AGN recall. The authors should re-estimate performance with nested cross-validation, with feature selection inside each training loop, or on an external validation sample.
  2. [Sect. 2.2 and Fig. 4] The most important feature in ks8, and in all classifiers tested, is the MIR color ch21, which is the same color entering the Donley et al. (2012) criterion used to build part of the AGN labeled set (211 MIR-selected sources). This creates a risk that the classifier is partly reproducing its own label definition rather than learning a new selection rule. The obscured-AGN recall itself is less directly affected because the 104 Type II AGN are spectroscopically classified, but the general AGN-selection numbers and the feature-importance interpretation are not. I ask for a concrete test: recompute the classification and feature importances either after removing MIR-selected AGN from the positive class or after dropping ch21 from the feature set, and quantify how much of the reported performance depends on this potentially circular feature.
minor comments (5)
  1. [Sect. 4.2] The claim that the 68.1% recall 'almost doubles' the De Cicco et al. (2021) value is not numerically correct: the Introduction quotes 21% recall for Type II AGN, and 68.1/21 ≈ 3.2; please correct the comparison (e.g., 'more than triples') or use an appropriate baseline.
  2. [Sects. 3.1 and 4.1] The synthetic-visit impact test is carried out on r-band data only; because the g-band light curves contain up to 16 synthetic points out of 33, a direct check of the g-band feature distributions would be useful, especially for the g-band features that appear in the K-S preselection (Autocor_length_g, Q31_g, MedianAbsDev_g).
  3. [Table 4 and Appendix A] The reported uncertainties are standard deviations across ten simulations with different random seeds; they do not include the sampling uncertainty of the LOOCV estimate or the variability introduced by feature selection. The text should state this limitation explicitly.
  4. [Sect. 2.2] The degree of overlap between the MIR-selected AGN and the spectroscopic Type I/II samples is not quantified; explicit counts or a Venn diagram would help the reader assess the ch21 circularity concern and interpret the per-subclass recalls.
  5. [Sect. 4.2] The text should be careful that ks8 is chosen for maximum obscured-AGN recall, not because it is uniformly the best classifier; ks9 has higher precision, accuracy, and F1, so the choice is a scientifically motivated trade-off rather than a global optimum.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline 68.1% obscured-AGN recall is a fitted figure of merit: feature selection and model choice use the full labeled set before LOOCV, making the reported performance optimistically biased rather than a prediction for new sources.

  1. fitted input called prediction [Section 4.2, "Selection of obscured AGN"; validation protocol in Section 3.3; Table 4, bottom section.]
    "Based on the results of the K-S test, we selected the features where the distance between the two distributions in a pair is large and the corresponding probability to get by chance a larger distance is small. We identified a possible threshold for distances D>0.25, which equals keeping 25 of the initial set of 162 features, and we built a classifier using only the selected features, which are reported in Table 6."

    LOOCV (Sect. 3.3) leaves each source out only from tree fitting, but the features used by those trees were chosen earlier from the same 2,543-source LS: the K-S filter with D>0.25 separates obscured AGN from inactive galaxies using labels, and backward elimination drops the least important feature based on importance rankings computed on those same data; ks8 is then selected because it returns the highest obscured-AGN recall in Table 4. Thus each left-out source's label has already influenced which features the model uses, so (68.1±1.2)% is not an unbiased estimate for unseen sources; it is a fit-selected, optimistically biased figure of merit presented as predictive performance.

full rationale

The headline value (68.1±1.2)% obscured-AGN recall from classifier ks8 (Sect. 4.2, Table 4) is the paper's central quantitative claim. The chain that produces it is: (i) a K-S test compares obscured AGN vs inactive galaxies on the full 2,543-source LS and keeps 25 of 162 features with D>0.25; (ii) a feature-importance ranking computed from the same labeled data drives backward elimination from 25 down to seven features; (iii) ks8 is singled out because it gives the highest obscured-AGN recall; (iv) LOOCV is then run only on tree fitting for the already-chosen feature set. Because every left-out source's label participated in steps (i)-(iii), the LOOCV estimate is not independent: the 68.1% value is a fit-selected, optimistically biased in-sample figure, not an unbiased prediction for new sources. This matches the pattern of a fitted input called a prediction. The ch21/MIR-label overlap is not separately definitional: the obscured-AGN labels come from optical spectroscopy (Marchesi et al. 2016), while the Donley et al. (2012) MIR criterion uses F(8.0)/F(4.5) and F(5.8)/F(3.6), not the ch21 = m(4.5)-m(3.6) color; at most this is a feature-label correlation. Self-citations to De Cicco et al. (2021) establish the LOOCV protocol and feature definitions but are not load-bearing for the new ks8 result. A separate numerical inconsistency is that 68.1% versus the quoted 21% Type II recall is more than triple, not "almost doubles"; that is an accuracy issue rather than circularity. The qualitative band and cadence comparisons in Table 4 are less affected, but the central obscured-AGN improvement claim rests on the biased estimate, so partial circularity is present.

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

The central results rest on several hand-chosen thresholds (imputation window, K-S feature-selection threshold, stopping criterion) and on the assumption that feature selection plus LOOCV on the same labeled set yields unbiased performance. The latter is the most consequential choice in the paper, as it directly affects the headline 68% obscured-AGN recall.

free parameters (4)
  • 15-day imputation window = 15 days
    Chosen threshold for linear interpolation of missing visits (Sect. 3.1).
  • K-S feature selection threshold = D > 0.25
    Threshold used to reduce 162 features to 25 based on distribution distances (Sect. 4.2).
  • Minimum feature set size = 7 features
    Iterative elimination stopped at seven features because no further improvement was expected (Sect. 4.2).
  • RF hyperparameters = n_estimators 100-500, max_depth 10-20, min_samples_split 2-10, min_samples_leaf 4, max_features sqrt/log2
    Tuned via grid search on the same labeled set used for evaluation (Sect. 4, Appendix A).
assumptions (4)
  • domain assumption Linear interpolation between visits separated by up to 15 days approximates the true light curve well enough for variability features.
    Sect. 3.1; this underlies all g-band features and bivariate features, and the synthetic-visit experiments.
  • domain assumption The literature-defined labels (Marchesi 2016 spectroscopic, Donley 2012 MIR, COSMOS2015 inactive galaxies) are correct and complete.
    Sect. 2.2; the entire training and evaluation depends on these labels.
  • ad hoc to paper Leave-one-out cross-validation after feature selection on the same data yields unbiased performance estimates.
    Sect. 4.2; violated by standard practice, since feature selection must be nested inside the CV loop to avoid leakage.
  • domain assumption Bivariate similarity measures computed on magnitude vectors are meaningful for AGN variability.
    Sect. 3.2 and Table 3; several formulas assume non-negative intensity-like vectors, while magnitudes can be near zero or negative in principle.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Selection of optically variable active galactic nuclei via a random forest algorithm." pith.science (2026). https://pith.science/paper/MUBOVK6B

@misc{pith2026250515819,
  author       = {Pith},
  title        = {Pith review of: Selection of optically variable active galactic nuclei via a random forest algorithm},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MUBOVK6B}},
  note         = {Machine review of arXiv:2505.15819}
}
read the original abstract

Context. A defining characteristic of active galactic nuclei (AGN) that distinguishes them from other astronomical sources is their stochastic variability, which is observable across the entire electromagnetic spectrum. Upcoming optical wide-field surveys, such as the Vera C. Rubin Observatory's Legacy Survey of Space and Time, are set to transform astronomy by delivering unprecedented volumes of data for time domain studies. This data influx will require the development of the expertise and methodologies necessary to manage and analyze it effectively. Aims. This project focuses on optimizing AGN selection through optical variability in wide-field surveys and aims to reduce the bias against obscured AGN. We tested a random forest (RF) algorithm trained on various feature sets to select AGN. The initial dataset consisted of 54 observations in the r-band and 25 in the g-band of the COSMOS field, captured with the VLT Survey Telescope over a 3.3-year baseline. Methods. Our analysis relies on feature sets derived separately from either band plus a set of features combining data from both bands, mostly characterizing AGN on the basis of their variability properties and obtained from their light curves. We trained multiple RF classifiers using different subsets of selected features and assessed their performance via targeted metrics. Results. Our tests provide valuable insights into the use of multiband and multivisit data for AGN identification. We compared our findings with previous studies and dedicated part of the analysis to potential enhancements in selecting obscured AGN. The expertise gained and the methodologies developed here are readily applicable to datasets from other ground- and space-based missions.

Figures

Figures reproduced from arXiv: 2505.15819 by the authors.

Figure 1
Figure 1. Average r-band magnitude (upper-left panel) and redshift (upper-right panel) for the two classes (AGN and non-AGN) in our LS. Each histogram has been normalized to the total number of sources in the corresponding class. Average r-band magnitude (lower-left panel) and redshift (lower-right panel) for the AGN LS and the three subsamples of unobscured, obscured, and MIR AGN. work, we also used the same set of color fea… view at source ↗
Figure 2
Figure 2. Comparison of the results obtained from the five tests where real visits were progressively replaced by synthetic visits for the vari￾ous metrics used in this work. The total number of visits is always 33, and we replaced part of them, four by four, with synthetic visits, up to a maximum of 16. The only error bars large enough to be visible corre￾spond to the recall for obscured AGN. suitable to identify obscured AG… view at source ↗
Figure 3
Figure 3. Importance ranking for the top 25 features that, based on the results of the K-S test, allow a better separation between obscured AGN and inactive galaxies. ch21 u-B class_star_hst ExcessVar_r GP_DRW_σ_r r-i GP_DRW_τ_r i-z 0 0.08 0.16 0.24 0.32 0.4 Feature Importance [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Importance ranking for the eight features that were used to build the RF classifier identifying the highest fraction (i.e., returning the high￾est recall) of obscured AGN. selection techniques that rely on optical variability to be able to return a completeness for obs…
Figure 5
Figure 5. Figure 5: Redshift as a function of the average r-band magnitude for the AGN correctly classified (TPs, dark green dots) and the misclassified AGN (FNs, magenta squares) from the ks8 classifier. using VST-COSMOS data (De Cicco et al. 2015, 2019, 2021), we confirm the well-known …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. VAR-PZnn: A machine-learning framework for AGN photometric redshifts using color and variability-based features

    astro-ph.GA 2026-07 conditional novelty 6.0 of 10

    VAR-PZnn combines ZTF g-band variability features with optical, MIR and NIR photometry in a mixture-density network to estimate AGN photometric redshifts with σ_NMAD=0.058 and an 8.2% outlier fraction on a 72,728-sour...

  2. Continuum optical-UV and X-ray variability of AGN: current results and future challenges

    astro-ph.HE 2025-06 unverdicted novelty 1.0 of 10

    A comprehensive review of AGN continuum variability from optical/UV to X-rays, with no new data.

Reference graph

Works this paper leans on

50 extracted references · 32 canonical work pages · cited by 2 Pith papers

  1. [1]

    2013, ApJ, 771, 9

    Allevato, V ., Paolillo, M., Papadakis, I., & Pinto, C. 2013, ApJ, 771, 9

    work page 2013

  2. [2]

    2016, IEEE Access, 4, 7940

    Amin, A., Anwar, S., Adnan, A., et al. 2016, IEEE Access, 4, 7940

    work page 2016

  3. [3]

    1993, ARA&A, 31, 473

    Antonucci, R. 1993, ARA&A, 31, 473

    1993

  4. [4]

    T., Cappellaro, E., Greggio, L., et al

    Botticella, M. T., Cappellaro, E., Greggio, L., et al. 2017, A&A, 598, A50

    work page 2017

  5. [5]

    2001, Machine Learning, 45, 5, cited By 34434

    Breiman, L. 2001, Machine Learning, 45, 5, cited By 34434

    work page 2001

  6. [6]

    2010, ApJ, 716, 348

    Brusa, M., Civano, F., Comastri, A., et al. 2010, ApJ, 716, 348

    work page 2010

  7. [7]

    & Charlot, S

    Bruzual, G. & Charlot, S. 2003, Monthly Notices of the Royal Astronomical Society, 344, 1000

    2003

  8. [8]

    R., Burbidge, E

    Burbidge, G. R., Burbidge, E. M., & Sandage, A. R. 1963, Reviews of Modern Physics, 35, 947

    work page 1963

Show all 50 references
  1. [9]

    & Schipani, P

    Capaccioli, M. & Schipani, P. 2011, The Messenger, 146, 2

    2011

  2. [10]

    T., Pignata, G., et al

    Cappellaro, E., Botticella, M. T., Pignata, G., et al. 2015, A&A, 584, A62

    2015

  3. [11]

    2024, A&A, 687, A246

    Cavuoti, S., De Cicco, D., Doorenbos, L., et al. 2024, A&A, 687, A246

    2024

  4. [12]

    2007, Int

    Cha, S.-H. 2007, Int. J. Math. Model. Meth. Appl. Sci., 1 De Cicco, D., Bauer, F. E., Paolillo, M., et al. 2021, A&A, 645, A103 De Cicco, D., Bauer, F. E., Paolillo, M., et al. 2022, A&A, 664, A117 De Cicco, D., Paolillo, M., Covone, G., et al. 2015, A&A, 574, A112 De Cicco, D...

    2007

  5. [13]

    L., Koekemoer, A

    Donley, J. L., Koekemoer, A. M., Brusa, M., et al. 2012, ApJ, 748, 142

    2012

  6. [14]

    A., Alexander, T., Crenshaw, D

    Edelson, R. A., Alexander, T., Crenshaw, D. M., et al. 1996, ApJ, 470, 364 Euclid Collaboration, Mellier, Y ., Abdurro’uf, et al. 2024, A&A, in prep., arXiv:2405.13491

    1996

  7. [15]

    2018, MNRAS, 481, 4311

    Eyheramendy, S., Elorrieta, F., & Palma, W. 2018, MNRAS, 481, 4311

    2018

  8. [16]

    2015, A&A, 579, A115

    Falocco, S., Paolillo, M., Covone, G., et al. 2015, A&A, 579, A115

    2015

  9. [17]

    2018, MNRAS, 479, 3858

    Fu, L., Liu, D., Radovich, M., et al. 2018, MNRAS, 479, 3858

    2018

  10. [18]

    J., Djorgovski, S

    Graham, M. J., Djorgovski, S. G., Drake, A. J., et al. 2017, MNRAS, 470, 4112

    2017

  11. [19]

    J., Pulgarin-Duque, L., Anderson, S

    Green, P. J., Pulgarin-Duque, L., Anderson, S. F., et al. 2022, ApJ, 933, 180

    2022

  12. [20]

    P., Cuillandre, J.-C., Bachelet, E., et al

    Guy, L. P., Cuillandre, J.-C., Bachelet, E., et al. 2022, in Zenodo id. 5836022, V ol. 58, 5836022

    2022

  13. [21]

    A., Förster, F., et al

    Huijse, P., Estévez, P. A., Förster, F., et al. 2018, ApJS, 236, 12 Ivezi´c, Ž., Kahn, S. M., Tyson, J. A., et al. 2019, ApJ, 873, 111

    2018

  14. [22]

    2013, in Proceedings - 2013 Humaine Asso- ciation Conference on Affective Computing and Intelligent Interaction, ACII 2013, V ol

    Jeni, L., Cohn, J., & De la Torre, F. 2013, in Proceedings - 2013 Humaine Asso- ciation Conference on Affective Computing and Intelligent Interaction, ACII 2013, V ol. 2013

    2013

  15. [23]

    Kim, D.-W., Protopapas, P., Bailer-Jones, C. A. L., et al. 2014, A&A, 566, A43

    2014

  16. [24]

    2011, ApJ, 735, 68

    Kim, D.-W., Protopapas, P., Byun, Y .-I., et al. 2011, ApJ, 735, 68

    2011

  17. [25]

    & Sarajedini, V

    Klesman, A. & Sarajedini, V . 2007, ApJ, 665, 225

    2007

  18. [26]

    M., Aussel, H., Calzetti, D., et al

    Koekemoer, A. M., Aussel, H., Calzetti, D., et al. 2007, ApJS, 172, 196

    2007

  19. [27]

    & Richstone, D

    Kormendy, J. & Richstone, D. 1995, ARA&A, 33, 581

    1995

  20. [28]

    J., Ilbert, O., et al

    Laigle, C., McCracken, H. J., Ilbert, O., et al. 2016, ApJS, 224, 24

    2016

  21. [29]

    M., Cales, S., Moran, E

    LaMassa, S. M., Cales, S., Moran, E. C., et al. 2015, ApJ, 800, 144

    2015

  22. [30]

    2020, MNRAS, 493, 3825

    Liu, D., Deng, W., Fan, Z., et al. 2020, MNRAS, 493, 3825

    2020

  23. [31]

    2018, MNRAS, 478, 2388 LSST Science Collaboration, Abell, P

    Liu, D., Fu, L., Liu, X., et al. 2018, MNRAS, 478, 2388 LSST Science Collaboration, Abell, P. A., Allison, J., et al. 2009, ArXiv e-prints, arXiv:0912.0201

    2018 arXiv

  24. [32]

    L., Ross, N

    MacLeod, C. L., Ross, N. P., Lawrence, A., et al. 2016, MNRAS, 457, 389

    2016

  25. [33]

    2016, ApJ, 817, 34

    Marchesi, S., Civano, F., Elvis, M., et al. 2016, ApJ, 817, 34

    2016

  26. [34]

    A., Mattox, J

    McLaughlin, M. A., Mattox, J. R., Cordes, J. M., & Thompson, D. J. 1996, ApJ, 473, 763

    1996

  27. [35]

    P., Andreon, S., et al

    Nakos, T., Willis, J. P., Andreon, S., et al. 2009, A&A, 494, 579

    2009

  28. [36]

    2015, arXiv e-prints, arXiv:1506.00010

    Nun, I., Protopapas, P., Sim, B., et al. 2015, arXiv e-prints, arXiv:1506.00010

    2015 arXiv

  29. [37]

    E., Paolillo, M., De Cicco, D., & Bauer, F

    Petrecca, V ., Papadakis, I. E., Paolillo, M., De Cicco, D., & Bauer, F. E. 2024, A&A, 686, A286

    2024

  30. [38]

    2020, A&A, 634, A50

    Poulain, M., Paolillo, M., De Cicco, D., et al. 2020, A&A, 634, A50

    2020

  31. [39]

    Rees, M. J. 1984, ARA&A, 22, 471

    1984

  32. [40]

    W., Starr, D

    Richards, J. W., Starr, D. L., Butler, N. R., et al. 2011, ApJ, 733, 10

    2011

  33. [41]

    Salpeter, E. E. 1964, ApJ, 140, 796

    1964

  34. [42]

    & Webb, G

    Sammut, C. & Webb, G. I., eds. 2010, Leave-One-Out Cross-Validation (Boston, MA: Springer US), 600–601 Sánchez-Sáez, P., Arredondo, J., Bayo, A., et al. 2023, A&A, 675, A195 Sánchez-Sáez, P., Lira, P., Cartier, R., et al. 2019, ApJS, 242, 10 Sánchez-Sáez, P., Reyes, I., Valenz...

    2010

  35. [43]

    L., Koo, D

    Sarajedini, V . L., Koo, D. C., Klesman, A. J., et al. 2011, ApJ, 731, 97 Savi´c, Ð. V ., Jankov, I., Yu, W., et al. 2023, ApJ, 953, 138

    2011

  36. [44]

    B., Marshall, P

    Schmidt, K. B., Marshall, P. J., Rix, H.-W., et al. 2010, ApJ, 714, 1194

    2010

  37. [45]

    2008, A&A, 488, 73

    Trevese, D., Boutsia, K., Vagnetti, F., Cappellaro, E., & Puccetti, S. 2008, A&A, 488, 73

    2008

  38. [46]

    G., Majewski, S

    Trevese, D., Kron, R. G., Majewski, S. R., Bershady, M. A., & Koo, D. C. 1994, ApJ, 433, 494

    1994

  39. [47]

    G., Koo, D

    Trevese, D., Pittella, G., Kron, R. G., Koo, D. C., & Bershady, M. 1989, AJ, 98, 108

    1989

  40. [48]

    & Hernandez-Rivera, E

    Tschopp, M. & Hernandez-Rivera, E. 2017, Quantifying Similarity and Distance Measures for Vector-Based Datasets: Histograms, Signals, and Probability Distribution Functions

    2017

  41. [49]

    H., Courvoisier, T

    Ulrich, M. H., Courvoisier, T. J. L., & Wamsteker, W. 1993, ApJ, 411, 125

    1993

  42. [50]

    Urry, C. M. & Padovani, P. 1995, PASP, 107, 803 Vanden Berk, D. E., Wilhite, B. C., Kron, R. G., et al. 2004, ApJ, 601, 692 Appendix A: Best hyperparameters obtained per classifier Table A.1.Set of best values obtained from a grid search-based opti- mization of the five hyperp...

    1995

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.