Pith. sign in

REVIEW 2 major objections 6 minor 1 cited by

A full-spectrum autoencoder plus nearest-neighbor search finds AGN that standard line-ratio diagrams miss, even in noisy DESI spectra.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-10 19:25 UTC pith:WLVGBELX

load-bearing objection Solid, well-validated application of SPENDER + k-d tree to DESI AGN selection; metrics are lower bounds and the recovery claim holds up. the 2 major comments →

arxiv 2607.07329 v1 pith:WLVGBELX submitted 2026-07-08 astro-ph.GA

Beyond traditional emission-line diagnostics: using autoencoders to uncover active galactic nuclei in DESI spectra

J. A. Alcolea , M. Siudek , M. Eriksen , M. Mezcua , R. Pucha , S. Juneau , S. Gontcho A Gontcho , S. Panda
show 44 more authors
J. Aguilar S. Ahlen D. Bianchi A. Brodzeller D. Brooks F. J. Castander T. Claybaugh A. Cuceu A. de la Macorra B. Dey P. Doel S. Ferraro A. Font-Ribera J. E. Forero-Romero E. Gazta\~naga G. Gutierrez C. Hahn H. K. Herrera-Alcantar D. Joyce R. Kehoe D. Kirkby T. Kisner A. Kremin O. Lahav C. Lamman M. Landriau L. Le Guillou A. Meisner R. Miquel J. Moustakas S. Nadathur W. Percival F. Prada I. P\'erez-R\`afols G. Rossi E. Sanchez E. Schlafly D. Schlegel M. Schubnell J. Silber D. Sprayberry G. Tarl\'e B. A. Weaver H. Zou
This is my paper
classification astro-ph.GA
keywords active galactic nucleiDESI spectraautoencoderslatent spacesemi-supervised classificationemission-line diagnosticsbroad-line AGNgalaxy classification
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Modern surveys deliver millions of galaxy spectra, yet the usual way to flag active galactic nuclei still relies on a handful of emission-line ratios that fail when lines are weak or buried. This paper shows that an unsupervised autoencoder can compress entire DESI optical spectra into a compact latent space whose geometry already groups galaxies by physical type. A simple nearest-neighbor vote in that space then recovers AGN and broad-line AGN with accuracies of 0.952 and 0.965, including many objects that classical BPT or equivalent-width cuts never label. Stacked spectra and an independent multi-diagnostic catalog confirm that most of those “missed” objects are genuine AGN, so the reported numbers are lower bounds. The same latent space also tracks stellar mass and star-formation rate, proving that the representation is astrophysically meaningful rather than a pure black box. The result is a scalable path to more complete AGN catalogs without re-fitting every spectrum by hand.

Core claim

Unsupervised spectral representation learning with the SPENDER autoencoder, followed by k-d tree nearest-neighbor classification in the ten-dimensional latent space, identifies AGN and broad-line AGN in DESI Guadalupe spectra (z ≤ 0.5) with accuracies of 0.952 and 0.965 respectively, recovers sources missed by single-diagnostic emission-line methods even at low signal-to-noise, and yields performance metrics that are conservative lower limits of the model’s true capability.

What carries the argument

The SPENDER autoencoder that compresses observed-frame DESI spectra into a redshift-aware, ten-dimensional rest-frame latent space; classification is then performed by majority vote among the ten nearest neighbors in that space, using FastSpecFit-derived labels only as external reference, not as training targets for the representation itself.

Load-bearing premise

The emission-line labels taken from FastSpecFit (BPT cuts, Hα equivalent-width bins, and broad-component signal-to-noise thresholds) are treated as a reliable enough reference set both to assign classes and to score accuracy, even though many of the recovered AGN fail those same cuts.

What would settle it

Apply the identical trained latent-space classifier to a large, multi-wavelength AGN sample (X-ray or mid-infrared selected) that is deliberately free of optical line-ratio pre-selection; if the recovery rate of confirmed AGN drops well below the reported 0.95 accuracy, or if the stacked spectra of “new” AGN no longer show AGN-like line ratios, the central claim fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 6 minor

Summary. The paper adapts the SPENDER autoencoder to compress ~50k DESI Guadalupe (DR1) galaxy spectra (z≤0.5) into a 10-dimensional rest-frame latent space, then assigns classes via k-d tree nearest-neighbor majority vote using FastSpecFit emission-line labels (BPT, EW(Hα), broad-component AON/S/N). It reports high accuracies for AGN (0.952) and broad-line sources (0.965), shows that many sources predicted as AGN but labeled non-AGN by single-diagnostic criteria are confirmed by stacking (BPT/WHAN) and by the independent DESI AGN/Galaxy VAC (93.5% of 1121 discrepant cases), and argues that the reported metrics are therefore lower bounds. The latent space is further shown to correlate with stellar mass, SFR, and NUV–r, supporting a physical interpretation of the representation.

Significance. If the recovery of low-S/N and single-diagnostic-missed AGN holds under broader validation, the work offers a practical, scalable route to more complete AGN catalogs for DESI-scale and future surveys (DR2, 4MOST, PFS, WEAVE). Strengths include: fully unsupervised representation learning (labels enter only at the k-d tree/metric stage), explicit framing of metrics as lower bounds, independent multi-diagnostic VAC cross-match and stacked-spectrum evidence, public data and code (Zenodo DOIs), and a latent space that encodes astrophysically meaningful structure beyond the discrete labels. The semi-supervised design also allows alternative label schemes to be projected without retraining the encoder.

major comments (2)
  1. The absolute accuracy numbers (Tables 2–3: AGN 0.952, BL 0.965) are evaluated against FastSpecFit-derived labels that the paper itself shows are incomplete (§2.4, Table 1; §4.4.1, Appendix C: ~50% of the 286 Other/passive o AGN cases fail S/N>3 on BPT lines; §4.4.2, Table 4: 93.5% of 1121 discrepant AGN predictions confirmed only by external VAC criteria). The authors correctly treat the metrics as lower bounds, and the stacking + VAC evidence supports genuine recovery. For the central claim to be fully load-bearing, the main text should more prominently quantify how much of the reported accuracy is driven by the incomplete reference set versus true spectral similarity (e.g., a sensitivity test re-labeling the VAC-confirmed subset and recomputing metrics, or reporting precision/recall against a multi-diagnostic gold set).
  2. Sample construction retains all NL/BL AGN and randomly subsamples other classes in magnitude–redshift bins (§2.5), producing an AGN fraction of ~24% (Table 1) that is substantially higher than the field. While this is appropriate for balanced training and for demonstrating recovery, the paper should clarify whether the quoted accuracies and the k-d tree neighbor statistics are expected to hold on an un-rebalanced DESI-like prior, or provide a reweighted/held-out evaluation that reflects the true class frequencies. Without that, the scalability claim for full DR1/DR2 catalogs remains only partially demonstrated.
minor comments (6)
  1. Abstract and §1 state 50,222 / 50,241 inconsistently before/after quality cuts; align the numbers throughout.
  2. Figure 2 vs Figure 4: the three-class and seven-class UMAP embeddings appear to use different random seeds or slightly different embeddings; state whether the same projection is reused or regenerated, and note that quantitative metrics are computed in 10-D only (§3.2 already says this, but a one-sentence reminder in the figure captions would help).
  3. Table 3: the Other class has recall 0.327 / F1 0.400; a short discussion of whether this is expected (heterogeneous residual class) or whether a different residual definition would improve interpretability would strengthen §4.2/§5.2.
  4. Appendix E: the PyQSOFit decomposition of the stacked spectrum is used only for illustration; make explicit that no black-hole mass or precise FWHM is claimed from the stack (the text already cautions this, but a stronger statement in the figure caption would avoid misreading).
  5. Minor typography: “Guadalupedataset” / missing spaces in a few places; “star-forming” capitalization is inconsistent across tables and text; “Redrock” vs “Redrock” template naming.
  6. Data/code availability: the Zenodo DOIs are welcome; ensure the released latent vectors and k-d tree labels match the exact 50,222-object sample used for Tables 2–3 so that the metrics are fully reproducible.

Circularity Check

1 steps flagged

No significant circularity: unsupervised SPENDER representation is independent of labels; metrics against incomplete FastSpecFit labels are explicitly framed as lower bounds and externally validated.

specific steps
  1. fitted input called prediction [§2.4, Table 1; §3.3–3.4; Tables 2–3; §4.1]
    "We validate the performance using labels derived from FastSpecFit’s emission line measurements defining seven galaxy classes... The model achieves high performance in identifying AGN and BL sources, with accuracies of 0.952 and 0.965, respectively... Our classification metrics are benchmarked against traditional diagnostics and we show they represent lower limits of the model’s true performance."

    The primary accuracy/precision/recall numbers are computed by comparing k-d tree majority votes (which use the FastSpecFit labels as neighbor votes) against the same FastSpecFit labels. This is a mild form of evaluating against the input labeling scheme. It is not a true circular derivation of a physical prediction, because (a) the latent representation itself is learned without those labels, (b) the paper explicitly frames the numbers as lower bounds, and (c) independent stacking + multi-diagnostic VAC evidence is supplied to break pure self-consistency. Hence only a minor residual dependence, not a load-bearing circular step.

full rationale

The derivation chain is: (1) SPENDER autoencoder trained solely on reconstruction/similarity/consistency losses with no class labels (§3.1); (2) 10-D latent space obtained; (3) k-d tree majority vote using FastSpecFit-derived reference labels for class assignment and metric evaluation (§3.3–3.4, Tables 2–3); (4) stacking of off-diagonal populations and cross-match to the independent multi-diagnostic AGN/Galaxy VAC to show that many ‘false positives’ relative to FastSpecFit are genuine AGN (§4.4, Table 4, Appendices C–F). Step (1) is fully unsupervised and label-agnostic. Step (3) inherits the known incompleteness of classical emission-line labels, but the paper never claims the reported accuracies (0.952/0.965) are absolute truth; it repeatedly states they are lower bounds and supplies independent evidence (stacked BPT/WHAN locations, VAC multi-diagnostic confirmation of 93.5 % of discrepant AGN predictions). There is no self-definitional loop, no fitted parameter re-labeled as a prediction, no uniqueness theorem imported from the authors, and no ansatz smuggled via self-citation that forces the central recovery claim. Residual dependence of the numerical accuracy figures on the same incomplete labels used for voting is a soft label-bias caveat already treated by the authors, not circularity of the derivation. Score 1.5 reflects only this minor residual dependence of the primary metrics on the reference labels.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central claim rests on standard ML architecture choices, domain conventions for optical AGN labeling, and a deliberately balanced training sample. No new physical entities are postulated. Free parameters are the usual architectural and classifier hyper-parameters; they are not fitted to force the AGN accuracy numbers.

free parameters (4)
  • latent_space_dimensionality = 10
    Fixed at 10 following Melchior et al. (2023); small ablation mentioned but not exhaustive.
  • k_nearest_neighbors = 10
    Majority vote among 10 nearest neighbors; leaf size and distance metric lightly explored.
  • learning_rate_schedule = 1e-3 → 1e-5
    OneCycleLR with peak LR stepped from 1e-3 to 1e-5 over 60 epochs; chosen for loss convergence.
  • sample_balancing = N_AGN kept full; others down-sampled
    All NL/BL AGN retained; other classes randomly sub-sampled in magnitude–redshift bins to produce ~50k balanced set.
axioms (4)
  • domain assumption FastSpecFit emission-line measurements plus Kewley/Kauffmann BPT and EW(Hα) cuts define reference galaxy classes for both neighbor voting and accuracy evaluation.
    Stated in §2.4 and Table 1; the paper later shows these labels are incomplete, yet still uses them as the primary benchmark.
  • domain assumption Proximity in the 10-D SPENDER latent space implies spectral similarity sufficient for majority-vote class transfer.
    Core of the k-d tree step (§3.3); standard nearest-neighbor assumption, not re-derived.
  • domain assumption Inverse-variance weighted MSE + similarity + consistency losses produce a redshift-robust rest-frame latent representation.
    Inherited from Liang et al. (2023b) and Melchior et al. (2023); adopted without re-proof.
  • domain assumption z ≤ 0.5 keeps Hα inside the DESI wavelength range and avoids Galactic contamination.
    Sample cut in §2.2; standard observational boundary.

pith-pipeline@v1.1.0-grok45 · 33203 in / 3081 out tokens · 55062 ms · 2026-07-10T19:25:14.772820+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Beyond traditional emission-line diagnostics: using autoencoders to uncover active galactic nuclei in DESI spectra." pith.science (2026). https://pith.science/paper/WLVGBELX

@misc{pith2026260707329,
  author       = {Pith},
  title        = {Pith review of: Beyond traditional emission-line diagnostics: using autoencoders to uncover active galactic nuclei in DESI spectra},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WLVGBELX}},
  note         = {Machine review of arXiv:2607.07329}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

The growing volume of spectroscopic data in modern surveys motivates data-driven approaches that complement traditional emission-line diagnostics for active galactic nuclei (AGN) identification. We present a machine learning framework that exploits the full optical spectrum using unsupervised representation learning within a semi-supervised classification scheme. We use the SPENDER autoencoder to compress DESI galaxy spectra into a low-dimensional latent space and classify sources through a k-d tree nearest-neighbor search. The model is trained on 50,222 DESI Main Survey spectra from the Guadalupe dataset and released as part of Data Release 1 (DR1), restricted to z <= 0.5. We validate the performance using labels derived from FastSpecFit's emission line measurements defining seven galaxy classes: AGN, broad-line (BL), composite, star-forming, passive, retired, and Other. The method achieves high accuracies for AGN (0.952) and broad-line AGN (0.965), reliably identifying these sources even in low signal-to-noise spectra and recovering AGN missed by standard single-diagnostic methods. Our classification metrics are benchmarked against traditional diagnostics, and we show they represent lower limits of the model's true performance. We also find that the learned latent space correlates with key galaxy properties such as stellar mass and star-formation rate, demonstrating that it captures physically meaningful information. These results show that unsupervised spectral representation learning, implemented within a semi-supervised classification framework, provides a scalable and effective approach for constructing more complete AGN catalogues for current and future spectroscopic surveys.

Figures

Figures reproduced from arXiv: 2607.07329 by A. Brodzeller, A. Cuceu, A. de la Macorra, A. Font-Ribera, A. Kremin, A. Meisner, B. A. Weaver, B. Dey, C. Hahn, C. Lamman, D. Bianchi, D. Brooks, D. Joyce, D. Kirkby, D. Schlegel, D. Sprayberry, E. Gazta\~naga, E. Sanchez, E. Schlafly, F. J. Castander, F. Prada, G. Gutierrez, G. Rossi, G. Tarl\'e, H. K. Herrera-Alcantar, H. Zou, I. P\'erez-R\`afols, J. A. Alcolea, J. Aguilar, J. E. Forero-Romero, J. Moustakas, J. Silber, L. Le Guillou, M. Eriksen, M. Landriau, M. Mezcua, M. Schubnell, M. Siudek, O. Lahav, P. Doel, R. Kehoe, R. Miquel, R. Pucha, S. Ahlen, S. Ferraro, S. Gontcho A Gontcho, S. Juneau, S. Nadathur, S. Panda, T. Claybaugh, T. Kisner, W. Percival.

Figure 2
Figure 2. Figure 2: UMAP projection of the 10-dimensional latent space into [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: UMAP embedding of the SPENDER model’s 10 latent features. The points are colored a posteriori according to seven galaxy classes: NL AGN, BL, composite, star-forming, passive, retired, and Other. terms to discourage the encoding of redshift information, red￾shift trends can still appear in the latent representation. As discussed in Liang et al. (2023b), this effect arises primarily from selection effects in… view at source ↗
Figure 5
Figure 5. Figure 5: UMAP embedding of the 10-dimensional latent space, color-coded by redshifts (top left), stellar masses (top right), specific [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Location of the stacked spectra corresponding to sources [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

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

  1. Unsupervised selection and characterisation of Little Red Dots in JWST surveys with manifold learning

    astro-ph.GA 2026-07 conditional novelty 7.0

    An unsupervised UMAP embedding of JWST photometry isolates Little Red Dots in two compact redshift-split regions, reaching ~0.78 purity at ~0.82 completeness and yielding ~100 new candidates.

Reference graph

Works this paper leans on

161 extracted references · 161 canonical work pages · cited by 1 Pith paper · 40 internal anchors

  1. [1]

    The Astronomical Journal , year =

    Peter Melchior and Yan Liang and ChangHoon Hahn and Andy Goulding , title =. The Astronomical Journal , year =

    work page

  2. [3]

    AGN Emission Line Diagnostics and the Mass-Metallicity Relation up to Redshift z~2: the Impact of Selection Effects and Evolution

    Active Galactic Nuclei Emission Line Diagnostics and the Mass-Metallicity Relation up to Redshift z -0.5ex 2: The Impact of Selection Effects and Evolution. , keywords =. doi:10.1088/0004-637X/788/1/88 , archivePrefix =. 1403.6832 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/788/1/88

  3. [4]

    2023 , month =

    Yan Liang and Peter Melchior and Sicong Lu and Andy Goulding and Charlotte Ward , title =. 2023 , month =. doi:10.3847/1538-3881/ace100 , url =

    work page doi:10.3847/1538-3881/ace100 2023

  4. [5]

    2023 , month =

    Liang, Yan and Melchior, Peter and Hahn, ChangHoon and Shen, Jeff and Goulding, Andy and Ward, Charlotte , title =. 2023 , month =. doi:10.3847/2041-8213/acfa03 , url =

    work page doi:10.3847/2041-8213/acfa03 2023

  5. [6]

    , keywords =

    The Spitzer-WISE Survey of the Ecliptic Poles. , keywords =. doi:10.1088/0004-637X/735/2/112 , adsurl =

    work page doi:10.1088/0004-637x/735/2/112

  6. [7]

    MNRAS , author =

    Using the Bright Ultrahard XMM-Newton survey to define an IR selection of luminous AGN based on WISE colours. , keywords =. doi:10.1111/j.1365-2966.2012.21843.x , archivePrefix =. 1208.2530 , primaryClass =

    work page doi:10.1111/j.1365-2966.2012.21843.x 2012

  7. [8]

    Mid-infrared Selection of Active Galactic Nuclei with the Wide-Field Infrared Survey Explorer. I. Characterizing WISE-selected Active Galactic Nuclei in COSMOS. , keywords =. doi:10.1088/0004-637X/753/1/30 , archivePrefix =. 1205.0811 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/753/1/30

  8. [9]

    The WISE AGN Catalog

    The WISE AGN Catalog. , keywords =. doi:10.3847/1538-4365/aaa00a , archivePrefix =. 1706.09901 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4365/aaa00a

  9. [10]

    Galaxy and Mass Assembly (GAMA): A $\textit{WISE}$ study of the activity of emission-line systems in G23

    Galaxy and Mass Assembly (GAMA): A WISE Study of the Activity of Emission-line Systems in G23. , keywords =. doi:10.3847/1538-4357/abba1a , archivePrefix =. 2009.05981 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4357/abba1a 2009

  10. [11]

    Monthly Notices of the Royal Astronomical Society , volume =

    Carvalho, S P and Dors, O L and Cardaci, M V and Hägele, G F and Krabbe, A C and Pérez-Montero, E and Monteiro, A F and Armah, M and Freitas-Lemes, P , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2020 , month =. doi:10.1093/mnras/staa193 , url =

    work page doi:10.1093/mnras/staa193 2020

  11. [12]

    doi:10.1111/j.1365-2966.2004.07942.x , journal=

    The luminosity-metallicity relation in the local Universe from the 2dF Galaxy Redshift Survey. , keywords =. doi:10.1111/j.1365-2966.2004.07697.x , archivePrefix =. astro-ph/0401615 , primaryClass =

    work page doi:10.1111/j.1365-2966.2004.07697.x 2004

  12. [13]

    Spectral classification of emission-line galaxies from the Sloan Digital Sky Survey. I. An improved classification for high-redshift galaxies. , keywords =. doi:10.1051/0004-6361/200913168 , archivePrefix =. 0910.4814 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/200913168

  13. [14]

    The ROSAT deep survey. II. Optical identification, photometry and spectra of X-ray sources in the Lockman field. , keywords =. doi:10.48550/arXiv.astro-ph/9709144 , archivePrefix =. astro-ph/9709144 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.astro-ph/9709144

  14. [15]

    , keywords =

    A New Diagnostic Diagram of Ionization Sources for High-redshift Emission Line Galaxies. , keywords =. doi:10.3847/1538-4357/aab207 , adsurl =

    work page doi:10.3847/1538-4357/aab207

  15. [16]

    MNRAS , author =

    Strongly star forming galaxies in the local Universe with nebular He II 4686 emission. , keywords =. doi:10.1111/j.1365-2966.2012.20439.x , archivePrefix =. 1201.1290 , primaryClass =

    work page doi:10.1111/j.1365-2966.2012.20439.x 2012

  16. [17]

    SDSS-IV MaNGA: Refining Strong Line Diagnostic Classifications Using Spatially Resolved Gas Dynamics

    SDSS-IV MaNGA: Refining Strong Line Diagnostic Classifications Using Spatially Resolved Gas Dynamics. , keywords =. doi:10.3847/1538-4357/abfe0a , archivePrefix =. 2011.06012 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4357/abfe0a 2011

  17. [18]

    Storchi-Bergmann and R

    T. Storchi-Bergmann and R. A. Schmitt and H. R. Schmitt and O. L. Dors , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2006 , doi =

    work page 2006

  18. [19]

    2018 , eprint=

    QuasarNET: Human-level spectral classification and redshifting with Deep Neural Networks , author=. 2018 , eprint=

    work page 2018

  19. [20]

    and Thomas, D

    Schawinski, K. and Thomas, D. and Sarzi, M. and Maraston, C. and Kaviraj, S. and Joo, S.-J. and Yi, S. K. and Silk, J. , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2007 , doi =

    work page 2007

  20. [21]

    A New Infrared Criterion for Selecting Active Galactic Nuclei to Lower Luminosities

    A New Infrared Criterion for Selecting Active Galactic Nuclei to Lower Luminosities. , keywords =. doi:10.3847/1538-3881/ac5e33 , archivePrefix =. 2203.11217 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-3881/ac5e33

  21. [22]

    Padovani and D.M

    P. Padovani and D.M. Alexander and R.J. Assef and et al. , title =. The Astronomy and Astrophysics Review , year =

    work page

  22. [23]

    Data Mining and Machine Learning in Astronomy

    Data Mining and Machine Learning in Astronomy. International Journal of Modern Physics D , keywords =. doi:10.1142/S0218271810017160 , archivePrefix =. 0906.2173 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1142/s0218271810017160

  23. [24]

    , title =

    Baron, D. , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2019 , doi =

    work page 2019

  24. [25]

    2024 , series =

    The Multi-AGN Catalog: The First Literature-Complete Accounting of Multi-AGN Systems , booktitle =. 2024 , series =

    work page 2024

  25. [26]

    Shrader , publisher =

    Volker Beckmann and Chris R. Shrader , publisher =. The Central Engine , booktitle =. doi:https://doi.org/10.1002/9783527666829.ch3 , url =. https://onlinelibrary.wiley.com/doi/pdf/10.1002/9783527666829.ch3 , year =

    work page doi:10.1002/9783527666829.ch3

  26. [27]

    Siudek, M. and. The VIMOS Public Extragalactic Redshift Survey (VIPERS). Unsupervised classification with photometric redshifts: a method to accurately classify large galaxy samples without spectroscopic information. arXiv e-prints , keywords =. doi:10.48550/arXiv.1805.09905 , archivePrefix =. 1805.09905 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1805.09905

  27. [28]

    10.1051/0004-6361/201832784

    Siudek, M. and. The VIMOS Public Extragalactic Redshift Survey (VIPERS) - The complexity of galaxy populations at 0.4 < z < 1.3 revealed with unsupervised machine-learning algorithms , DOI= "10.1051/0004-6361/201832784", url= "https://doi.org/10.1051/0004-6361/201832784", journal =

    work page doi:10.1051/0004-6361/201832784

  28. [29]

    2025 , eprint=

    The optical and infrared are connected , author=. 2025 , eprint=

    work page 2025

  29. [30]

    Reconstructing Quasar Spectra and Measuring the Ly$\alpha$ Forest with ${\rm S{\scriptsize pender}Q}$

    ChangHoon Hahn and Satya Gontcho A Gontcho and Peter Melchior and Hiram K. Herrera-Alcantar and Jessica Nicole Aguilar and Steven Ahlen and Davide Bianchi and David Brooks and Todd Claybaugh and Axel de la Macorra and Arjun Dey and Peter Doel and Jaime E. Forero-Romero and Gaston Gutierrez and Mustapha Ishak and Stephanie Juneau and David Kirkby and Theod...

    work page internal anchor Pith review Pith/arXiv arXiv

  30. [31]

    Data Release 1 of the Dark Energy Spectroscopic Instrument

    Data Release 1 of the Dark Energy Spectroscopic Instrument. arXiv e-prints , keywords =. doi:10.48550/arXiv.2503.14745 , archivePrefix =. 2503.14745 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2503.14745

  31. [32]

    Ashley and Martínez-Galarza, Rafael and Dillmann, Steven , title =

    Song, Yanke and Villar, V. Ashley and Martínez-Galarza, Rafael and Dillmann, Steven , title =. 2025 , month =. doi:10.3847/1538-4357/add72e , url =

    work page doi:10.3847/1538-4357/add72e 2025

  32. [33]

    Classification, regression, and accretion signatures of Chandra sources , author=

    Extracting latent representations from X-ray spectra. Classification, regression, and accretion signatures of Chandra sources , author=. 2025 , eprint=

    work page 2025

  33. [34]

    and Hainline, Kevin N

    Hviding, Raphael E. and Hainline, Kevin N. and Goulding, Andy D. and Greene, Jenny E. , title =. The Astronomical Journal , volume =. 2024 , doi =

    work page 2024

  34. [35]

    Monthly Notices of the Royal Astronomical Society , volume =

    Pérez-Díaz, Víctor Samuel and Martínez-Galarza, Juan Rafael and Caicedo, Alexander and D’Abrusco, Raffaele , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2024 , month =. doi:10.1093/mnras/stae260 , url =

    work page doi:10.1093/mnras/stae260 2024

  35. [36]

    The first catalogue of spectroscopically confirmed red nuggets at z~0.7 from the VIPERS survey. Linking high-z red nuggets and local relics

    The first catalogue of spectroscopically confirmed red nuggets at z 0.7 from the VIPERS survey. Linking high-z red nuggets and local relics. , keywords =. doi:10.1051/0004-6361/202243616 , archivePrefix =. 2208.04601 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/202243616

  36. [37]

    Monthly Notices of the Royal Astronomical Society , volume =

    Siudek, M and Mezcua, M and Krywult, J , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2022 , month =. doi:10.1093/mnras/stac3092 , url =

    work page doi:10.1093/mnras/stac3092 2022

  37. [38]

    Overmassive Black Holes in Dwarf Galaxies Out to z ∼ 0.9 in the VIPERS Survey , volume=

    Mezcua, Mar and Siudek, Malgorzata and Suh, Hyewon and Valiante, Rosa and Spinoso, Daniele and Bonoli, Silvia , year=. Overmassive Black Holes in Dwarf Galaxies Out to z ∼ 0.9 in the VIPERS Survey , volume=. The Astrophysical Journal Letters , publisher=. doi:10.3847/2041-8213/acae25 , number=

    work page doi:10.3847/2041-8213/acae25 2041

  38. [39]

    Unsupervised classification reveals new evolutionary pathways

    Unsupervised classification reveals new evolutionary pathways. arXiv e-prints , keywords =. doi:10.48550/arXiv.2211.11792 , archivePrefix =. 2211.11792 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2211.11792

  39. [40]

    2019 , month =

    Shen, Yue and Hwang, Hsiang-Chih and Zakamska, Nadia and Liu, Xin , title =. 2019 , month =. doi:10.3847/2041-8213/ab4b54 , url =

    work page doi:10.3847/2041-8213/ab4b54 2019

  40. [41]

    The Journal of Open Source Software , keywords =

    corner.py: Scatterplot matrices in Python. The Journal of Open Source Software , keywords =

    work page

  41. [42]

    Revisiting the Unified Model of Active Galactic Nuclei

    Revisiting the Unified Model of Active Galactic Nuclei , journal =. doi:10.1146/annurev-astro-082214-122302 , archivePrefix =. 1505.00811 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1146/annurev-astro-082214-122302

  42. [43]

    2022 , eprint=

    Reconstructing and Classifying SDSS DR16 Galaxy Spectra with Machine-Learning and Dimensionality Reduction Algorithms , author=. 2022 , eprint=

    work page 2022

  43. [44]

    Portillo, Stephen K. N. and Parejko, John K. and Vergara, Jorge R. and Connolly, Andrew J. , year=. Dimensionality Reduction of SDSS Spectra with Variational Autoencoders , volume=. The Astronomical Journal , publisher=. doi:10.3847/1538-3881/ab9644 , number=

    work page doi:10.3847/1538-3881/ab9644

  44. [45]

    Teimoorinia, Hossen and Archinuk, Finn and Woo, Joanna and Shishehchi, Sara and Bluck, Asa F. L. , year=. Mapping the Diversity of Galaxy Spectra with Deep Unsupervised Machine Learning , volume=. The Astronomical Journal , publisher=. doi:10.3847/1538-3881/ac4039 , number=

    work page doi:10.3847/1538-3881/ac4039

  45. [46]

    2020 , eprint=

    UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction , author=. 2020 , eprint=

    work page 2020

  46. [47]

    The Journal of Open Source Software , volume=

    UMAP: Uniform Manifold Approximation and Projection , author=. The Journal of Open Source Software , volume=

    work page

  47. [48]

    and Varoquaux, G

    Pedregosa, F. and Varoquaux, G. and Gramfort, A. and Michel, V. and Thirion, B. and Grisel, O. and Blondel, M. and Prettenhofer, P. and Weiss, R. and Dubourg, V. and Vanderplas, J. and Passos, A. and Cournapeau, D. and Brucher, M. and Perrot, M. and Duchesnay, E. , journal=. Scikit-learn: Machine Learning in

    work page

  48. [49]

    2023 , month = aug, eid =

    FastSpecFit: Fast spectral synthesis and emission-line fitting of DESI spectra. 2023 , month = aug, eid =

    work page 2023

  49. [50]

    Johnson and Charlie Conroy , title =

    Justin Alsing and Hiranya Peiris and Joel Leja and ChangHoon Hahn and Rita Tojeiro and Daniel Mortlock and Boris Leistedt and Benjamin D. Johnson and Charlie Conroy , title =. 2020 , month =. doi:10.3847/1538-4365/ab917f , url =

    work page doi:10.3847/1538-4365/ab917f 2020

  50. [51]

    2023 , institution =

    Suárez Pérez, John Fredy , title =. 2023 , institution =

    work page 2023

  51. [52]

    and Bailey, S

    Guy, J. and Bailey, S. and Kremin, A. and Alam, Shadab and Alexander, D. M. and Allende Prieto, C. and BenZvi, S. and Bolton, A. S. and Brooks, D. and Chaussidon, E. and Cooper, A. P. and Dawson, K. and de la Macorra, A. and Dey, A. and Dey, Biprateep and Dhungana, G. and Eisenstein, D. J. and Font-Ribera, A. and Forero-Romero, J. E. and Gaztañaga, E. and...

    work page doi:10.3847/1538-3881/acb212

  52. [53]

    Deep Learning , author=

    work page

  53. [54]

    The wide-field, multiplexed, spectroscopic facility WEAVE: Survey design, overview, and simulated implementation

    The wide-field, multiplexed, spectroscopic facility WEAVE: Survey design, overview, and simulated implementation. , keywords =. doi:10.1093/mnras/stad557 , archivePrefix =. 2212.03981 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stad557

  54. [55]

    and Yang, Jinyi and Ahlen, Steven and Bailey, Stephen and Brooks, David and Cai, Zheng and Chabanier, Solène and Davis, Tamara M

    Chaussidon, Edmond and Yèche, Christophe and Palanque-Delabrouille, Nathalie and Alexander, David M. and Yang, Jinyi and Ahlen, Steven and Bailey, Stephen and Brooks, David and Cai, Zheng and Chabanier, Solène and Davis, Tamara M. and Dawson, Kyle and de laMacorra, Axel and Dey, Arjun and Dey, Biprateep and Eftekharzadeh, Sarah and Eisenstein, Daniel J. a...

    work page doi:10.3847/1538-4357/acb3c2 2023

  55. [56]

    The Early Data Release of the Dark Energy Spectroscopic Instrument

    The Early Data Release of the Dark Energy Spectroscopic Instrument. , keywords =. doi:10.3847/1538-3881/ad3217 , archivePrefix =. 2306.06308 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-3881/ad3217

  56. [57]

    Communications of the ACM , volume=

    Multidimensional binary search trees used for associative searching , author=. Communications of the ACM , volume=. 1975 , publisher=

    work page 1975

  57. [58]

    Proceedings of ICML Workshop on Unsupervised and Transfer Learning , pages =

    Autoencoders, Unsupervised Learning, and Deep Architectures , author =. Proceedings of ICML Workshop on Unsupervised and Transfer Learning , pages =. 2012 , volume =

    work page 2012

  58. [59]

    Architecture and Automated Redshift Measurement , author=

    SpecPT (Spectroscopy Pre-trained Transformer) Model for Extragalactic Spectroscopy: I. Architecture and Automated Redshift Measurement , author=. 2025 , eprint=

    work page 2025

  59. [60]

    Galaxy Spectra neural Network (GaSNet)

    Zhong, Fucheng and Napolitano, Nicola R and Heneka, Caroline and Li, Rui and Bauer, Franz Erik and Bouche, Nicolas and Comparat, Johan and Kim, Young-Lo and Krogager, Jens-Kristian and Longhetti, Marcella and Loveday, Jonathan and Roukema, Boudewijn F and Rouse, Benedict L and Salvato, Mara and Tortora, Crescenzo and Assef, Roberto J and Cassarà, Letizia ...

    work page doi:10.1093/mnras/stae1461

  60. [61]

    AstroCLIP: a cross-modal foundation model for galaxies , volume=

    Parker, Liam and Lanusse, Francois and Golkar, Siavash and Sarra, Leopoldo and Cranmer, Miles and Bietti, Alberto and Eickenberg, Michael and Krawezik, Geraud and McCabe, Michael and Morel, Rudy and Ohana, Ruben and Pettee, Mariel and Régaldo-Saint Blancard, Bruno and Cho, Kyunghyun and Ho, Shirley , year=. AstroCLIP: a cross-modal foundation model for ga...

    work page doi:10.1093/mnras/stae1450

  61. [62]

    2011 , eprint=

    Euclid Definition Study Report , author=. 2011 , eprint=

    work page 2011

  62. [63]

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

    LSST: From Science Drivers to Reference Design and Anticipated Data Products. , keywords =. doi:10.3847/1538-4357/ab042c , archivePrefix =. 0805.2366 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4357/ab042c

  63. [64]

    and Dey, Arjun and Siudek, M

    Pucha, Ragadeepika and Juneau, S. and Dey, Arjun and Siudek, M. and Mezcua, M. and Moustakas, J. and BenZvi, S. and Hainline, K. and Hviding, R. and Mao, Yao-Yuan and Alexander, D. and Alfarsy, R. and Circosta, Chiara and Guo, Wei-Jian and Manwadkar, Viraj and Martini, P. and Weaver, B. and Aguilar, Janette and Ahlen, S. and Zou, H. , year =. Tripling the...

    work page

  64. [65]

    and Mezcua, M

    Siudek, M. and Mezcua, M. and Circosta, C. and Maraston, C. and Moustakas, J. and Zou, H. and Aguilar, J. and Ahlen, S. and Bianchi, D. and Brooks, D. and Claybaugh, T. and Dawson, K. S. and de la Macorra, A. and Dey, A. and Doel, P. and Forero-Romero, J. E. and Gaztañaga, E. and Gontcho A Gontcho, S. and Gutierrez, G. and Ishak, M. and Juneau, S. and Kir...

    work page doi:10.1051/0004-6361/202555463

  65. [66]

    Iron-corrected Single-epoch Black Hole Masses of DESI Quasars at low redshift

    Iron-corrected Single-epoch Black Hole Masses of DESI Quasars at Low Redshift. , keywords =. doi:10.3847/1538-4357/add7dd , archivePrefix =. 2502.03684 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4357/add7dd

  66. [67]

    arXiv e-prints , keywords =

    Identifying Anomalous DESI Galaxy Spectra with a Variational Autoencoder. arXiv e-prints , keywords =. doi:10.48550/arXiv.2506.17376 , archivePrefix =. 2506.17376 , primaryClass =

    work page doi:10.48550/arxiv.2506.17376

  67. [68]

    No significant evolution of relations between Black hole mass and Galaxy total stellar mass up to z~2.5

    No Significant Evolution of Relations between Black Hole Mass and Galaxy Total Stellar Mass Up to z 2.5. , keywords =. doi:10.3847/1538-4357/ab5f5f , archivePrefix =. 1912.02824 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4357/ab5f5f 1912

  68. [69]

    Black Holes at the Centers of Nearby Dwarf Galaxies

    Black Holes At the Centers of Nearby Dwarf Galaxies. , keywords =. doi:10.1088/0004-6256/148/6/136 , archivePrefix =. 1408.4451 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-6256/148/6/136

  69. [70]

    and Doel, Peter and Gutierrez, Gaston and Besuner, Robert and Brooks, David and Gallo, Giuseppe and Heetderks, Henry and Jelinsky, Patrick and Kent, Stephen M

    Miller, Timothy N. and Doel, Peter and Gutierrez, Gaston and Besuner, Robert and Brooks, David and Gallo, Giuseppe and Heetderks, Henry and Jelinsky, Patrick and Kent, Stephen M. and Lampton, Michael and Levi, Michael E. and Liang, Ming and Meisner, Aaron and Sholl, Michael J. and Silber, Joseph Harry and Sprayberry, David and Aguilar, Jessica Nicole and ...

    work page doi:10.3847/1538-3881/ad45fe 2024

  70. [71]

    arXiv e-prints , keywords =

    The DESI Experiment Part II: Instrument Design. arXiv e-prints , keywords =

    work page

  71. [72]

    and Ahlen, S

    Anand, Abhijeet and Guy, Julien and Bailey, Stephen and Moustakas, John and Aguilar, J. and Ahlen, S. and Bolton, A. S. and Brodzeller, A. and Brooks, D. and Claybaugh, T. and Cole, S. and de la Macorra, A. and Dey, Biprateep and Fanning, K. and Forero-Romero, J. E. and Gaztañaga, E. and Gontcho A Gontcho, S. and Gutierrez, G. and Honscheid, K. and Howlet...

    work page doi:10.3847/1538-3881/ad60c2 2024

  72. [73]

    Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pages =

    Chen, Tianqi and Guestrin, Carlos , title =. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pages =. 2016 , isbn =. doi:10.1145/2939672.2939785 , abstract =

    work page doi:10.1145/2939672.2939785 2016

  73. [74]

    Annual Review of Astronomy and Astrophysics , pages =

    Obscured Active Galactic Nuclei , doi =. Annual Review of Astronomy and Astrophysics , pages =. 2018 , author =

    work page 2018

  74. [75]

    , keywords =

    Cid Fernandes, R. and Stasińska, G. and Mateus, A. and Vale Asari, N. , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2011 , month =. doi:10.1111/j.1365-2966.2011.18244.x , url =

    work page doi:10.1111/j.1365-2966.2011.18244.x 2011

  75. [76]

    , keywords =

    Unified models for active galactic nuclei and quasars. , keywords =. doi:10.1146/annurev.aa.31.090193.002353 , adsurl =

    work page doi:10.1146/annurev.aa.31.090193.002353

  76. [77]

    2024 , month =

    Panda, Swayamtrupta and Śniegowska, Marzena , title =. 2024 , month =. doi:10.3847/1538-4365/ad344f , url =

    work page doi:10.3847/1538-4365/ad344f 2024

  77. [78]

    Unified Schemes for Radio-Loud Active Galactic Nuclei

    Unified Schemes for Radio-Loud Active Galactic Nuclei. , keywords =. doi:10.1086/133630 , archivePrefix =. astro-ph/9506063 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1086/133630

  78. [79]

    Fast and efficient identification of anomalous galaxy spectra with neural density estimation

    Fast and efficient identification of anomalous galaxy spectra with neural density estimation. , keywords =. doi:10.1093/mnras/stad2773 , archivePrefix =. 2308.00752 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stad2773

  79. [80]

    The Biases of Optical Line-Ratio Selection for Active Galactic Nuclei, and the Intrinsic Relationship between Black Hole Accretion and Galaxy Star Formation

    The Biases of Optical Line-Ratio Selection for Active Galactic Nuclei and the Intrinsic Relationship between Black Hole Accretion and Galaxy Star Formation. , keywords =. doi:10.1088/0004-637X/811/1/26 , archivePrefix =. 1501.02801 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/811/1/26

  80. [81]

    Widespread and Hidden Active Galactic Nuclei in Star-Forming Galaxies at Redshift > 0.3

    Widespread and Hidden Active Galactic Nuclei in Star-forming Galaxies at Redshift >0.3. , keywords =. doi:10.1088/0004-637X/764/2/176 , archivePrefix =. 1211.6436 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/764/2/176

Showing first 80 references.