Pith. sign in

REVIEW 4 major objections 5 minor 58 references

Spectroscopic Quasar Anomaly Detection (SQuAD) I: Rest-Frame UV Spectra from SDSS DR16

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 This paper claims that an automated spectral-clustering pipeline applied to rest-frame ultraviolet quasar spectra identifies 1,888 anomalous quasars in ten distinct groups, with physical causes tied to Eddington ratio and…

desk verdict A useful rare-quasar catalog that is undermined by unresolved count inconsistencies and threshold choices; worth refereeing after fixes. read the letter →

arxiv 2411.16858 v2 pith:OCBHMKNH submitted 2024-11-25 astro-ph.GA

Arihant Tiwari , M. Vivek (Indian Institute of Astrophysics) This is my paper · ORCID
classification astro-ph.GA
keywords quasarspectraanomalydetectionprincipalcomponentanalysisK-MeansclusteringbroadabsorptionlinequasarsCIVemissionSivalue-addedcatalog
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 claims that a data-driven pipeline—principal component analysis followed by hierarchical K-Means clustering on 20 PCA coefficients—can pick out genuinely rare quasar spectra from a large uniform sample of 81,814 rest-frame UV quasar spectra. Applied to the redshift range 1.88–2.47, the pipeline flags 1,888 quasars as anomalous and sorts them into ten groups: C IV Peakers, Excess Si IV emitters, Si IV Deficient objects, four broad-absorption-line subtypes, and three reddened subtypes. The paper further argues that these spectral anomalies trace real physical differences, namely lower Eddington ratios in C IV Peakers, super-solar broad-line-region metallicity in Excess Si IV emitters, and sub-solar metallicity in Si IV Deficient objects. If correct, the result provides a large value-added catalog of rare quasar outliers that can be studied statistically rather than as isolated curiosities.

What carries the argument

The load-bearing object is a 20-dimensional PCA coefficient space built from max-normalized, smoothed, resampled, and padded rest-frame spectra spanning 1250–3000 Å. Quasars are clustered with K-Means into three main clusters in that space; the cluster centroids encode mean spectral shapes, and the second PCA eigenvector carries most of the reddening information, so high PCA 2 coefficients mean redder spectra. Anomalies are declared as spectra whose Euclidean distance from their cluster centroid exceeds 5σ for clusters 1 and 2 and 4σ for cluster 3, thresholds chosen by visual inspection of the tail of the distance histograms. The flagged spectra are re-clustered, split by BAL probability, and re-clustered again to yield the ten final anomaly categories. The top 8% of spectra with the largest PCA reconstruction residuals are removed before the main clustering and handled separately.

What would settle it

Re-run the same pipeline on the same quasar sample while varying the sigma thresholds (e.g., 4σ versus 5σ for all clusters, or a fixed-percentile cut) and check whether the same ten groups emerge with stable memberships; if group membership shifts substantially, the anomaly classes are artifacts of the cut. Independently, measure Eddington ratios and broad-line-region metallicities for the flagged C IV Peakers and Excess Si IV emitters from X-ray and virial-mass data; if they match normal quasars, the paper's physical explanations fail.

Watch

Extended reading notes

Core claim

The central discovery is that unsupervised clustering in a 20-dimensional PCA coefficient space built from rest-frame 1250–3000 Å spectra separates quasar spectra into three main populations, and that objects far from their cluster centroids are not random noise but recurrent spectral types. Using a 5σ distance cut for two clusters and a 4σ cut for the third, the pipeline flags 1,994 outlier spectra in the full dataset and 1,270 in a BAL-free subset; the final catalog contains 1,888 anomalous quasars in ten groups. These are C IV Peakers, Excess Si IV emitters, Si IV Deficient objects, four broad-absorption-line subtypes (Blue BALs, Flat BALs, Reddened BALs, FeLoBALs), and three reddened non-BAL subtypes (Heavily Reddened, Moderately Reddened, and Plateau-shaped spectrum quasars). The paper attributes the C IV Peaker anomaly to lower Eddington ratios, the Excess Si IV group to super-solar broad-line-region metallicity, and the Si IV Deficient group to sub-solar metallicity, using line-ratio diagnostics and comparisons with a published quasar line catalog.

Load-bearing premise

The catalog's membership rests on the chosen sigma thresholds (5σ for two clusters, 4σ for the third) and on dropping the 8% of spectra that the model reproduces worst before clustering; if those choices shift, every group's size and composition changes.

Editorial extensions

If this is right

  • The value-added catalog gives researchers a large, homogeneous sample of rare quasar spectra—1,888 objects in ten groups—for statistical studies of extreme line emission, BAL outflows, and dust reddening, instead of relying on single-object case studies.
  • If the physical attributions are right, C IV Peakers bracket Weak-Line Quasars as the opposite end of an Eddington-ratio sequence, giving a target population for testing accretion-disk and ionizing-continuum models.
  • The Excess Si IV and Si IV Deficient groups supply two broad-line-region metallicity extremes that can be used to map chemical enrichment across quasar populations.
  • The deliberate separation of BAL and non-BAL samples demonstrates that removing a dominant outlier class before clustering exposes rarer anomaly types, a strategy that transfers to other large spectroscopic surveys.
  • The pipeline's detection of BAL quasars missed by an existing BAL_PROB flag suggests anomaly screening can reveal incompleteness in standard BAL catalogs.

Reading between the lines

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

  • Beyond the paper: the group boundaries probably depend on the chosen sigma thresholds and on the 8% reconstruction-residual cut, so a stability analysis across thresholds would tell whether the ten categories are robust or survey-specific.
  • Beyond the paper: the physical explanations for C IV Peakers and Si IV anomalies are inferred from line-ratio arguments; direct X-ray, multi-epoch, or virial-mass measurements of flagged objects would test them.
  • Beyond the paper: the same PCA-plus-K-Means recipe should transfer to optical spectra and to future surveys, where redshift windows and signal-to-noise distributions will require re-calibrating the anomaly thresholds.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript applies PCA dimensionality reduction followed by hierarchical k-means clustering to 81,814 SDSS DR16 quasar spectra in the rest-frame 1250–3000 Å window, with and without BAL quasars. Objects at 5σ (clusters 1 and 2) or 4σ (cluster 3) from their cluster centroids are labeled anomalous, and a second k-means pass groups them into what the authors condense into ten anomaly classes: C IV Peakers, Excess Si IV emitters, Si IV Deficient anomalies, four BAL subtypes, and three reddened non-BAL subtypes. The physical discussion uses line measures from Wu & Shen (2022) and literature diagnostics to attribute the anomalies to low Eddington ratios, super-solar or sub-solar BLR metallicities, and dust reddening. The paper claims 1,888 anomalous quasars and presents them as a value-added catalog.

Significance. If the catalog is reproducible, this would be a genuinely useful resource: it is one of the largest classified collections of spectroscopically rare quasars, and the grouping into physically interpretable classes is a helpful starting point for follow-up. The pipeline is clearly described in broad strokes, the two-dataset strategy is sensible, and the comparisons to the Wu & Shen (2022) measurements and external calibrations (e.g., Fu et al. 2022; Hamann et al. 2002) are appropriate. The main risk is not circularity—the anomaly detection is unsupervised—but traceability: the catalog counts and membership depend on threshold choices, manual reassignments, and an incompletely documented reconstruction-outlier path. Those issues are correctable but must be addressed before the catalog can be used.

major comments (4)
  1. [§3.1, §6, Tables 2–4] The headline count of 1,888 anomalous quasars cannot be reproduced from the paper. Section 3.1 reports 1,994 anomalies in the Full Dataset and 1,270 in the Non-BAL Only Dataset. Table 2 gives per-cluster anomaly counts that sum to 1,479 (Full) and 978 (Non-BAL), while Table 3 gives anomaly-group counts that sum to 1,542 (Full) and 916 (Non-BAL); neither matches the anomaly totals. The final group counts listed in §6 sum to 1,653 (65 + 227 + 328 + 64 + 306 + 213 + 109 + 165 + 93 + 83), and §4.2.3 states that the BAL anomaly group has 672 members although Table 4 sums to 692. The percentages quoted in §4.2.1, §4.2.2, and §4.2.4 (3.4%, 11.3%, 16.4%) imply yet another total, close to 1,994. The central catalog claim therefore lacks a defined reconciliation. Please provide a membership-flow table with counts at every step—initial anomalies, reconstruction outliers, manual removals, manual reassignments, duplicates across the two datasets, and final group assignments—and use it to derive the number 1,888 explicitly.
  2. [§3.1, Fig. 7] The anomaly definition depends on sigma thresholds that were selected by visual inspection: 5σ for clusters 1 and 2 and 4σ for cluster 3, chosen because the tails look tapering or diffuse. No stability analysis is reported. Because every group count and the final catalog membership change if these thresholds are moved, the paper needs a quantitative robustness test—for example, a scan over thresholds with reported membership overlap or rank correlation, or an objective model for the tail of the distance distribution. This is not a cosmetic issue; it determines who is in the catalog.
  3. [§2.2–2.3, Eq. (2)] The reconstruction-outlier path is load-bearing but under-specified. The top 8th percentile of the residual-error distribution from Eq. (2) is removed before clustering; then a “simple algorithm” flags spectra with large flat regions; the remaining reconstruction outliers are clustered into three classes and “assigned to their appropriate classifications.” The paper never defines the flat-region algorithm, states how many spectra were discarded versus kept, or explains how this path contributes to the final 1,888 count. Since the preprocessing in §2.1 pads spectra with the trailing flux value, which can create flat sections, the boundary between genuine reddened/BAL anomalies and padding artifacts must be quantified. Please specify the algorithm, give the counts at each substep, and show where these objects appear in the final catalog.
  4. [§5.7, §4.2.2, §4.2.4] Several membership decisions are manual and undocumented. The C IV Peakers group is said to contain roughly 350 cosmic-ray contaminants that are removed by an unspecified equivalent-width/flux cut; §4.2.2 describes quasars that were “manually identified and reassigned” from the Excess Si IV group to BAL subgroups; §4.2.4 reports 18 leaked BALs that were “manually picked out”; and §5.7 states that other machine-error anomalies were “visually identified and discarded.” Each of these decisions changes the final group counts, yet no criteria, lists, or per-step counts are provided. The paper should make these steps reproducible, including a precise definition of the C IV EW/flux cut and the number of objects removed or reassigned at each stage. As written, a user cannot tell whether the 1,888 number includes or excludes any of these manually handled objects.
minor comments (5)
  1. [Table 5] The sample catalog row for SDSS J123015.99+062056.7 lists redshift z = 1.8512, which is outside the declared sample range 1.88 ≤ z ≤ 2.47. Please verify this entry and the redshift cut used in the catalog; if this is not a typo, the selection criterion needs correction.
  2. [§5.3] The statement that a Z/Z⊙ ≈ 0.4 metallicity is “nearly 40 times lower” than that of the Excess Si IV quasars is not supported by the preceding text, since the Excess Si IV group is described only as super-solar. Please state the assumed metallicities or calibrations used to derive this factor.
  3. [§4.2.1, Table 3] The text says the C IV Peaker group contains roughly 350 contaminant spectra, but Table 3 lists only 232 members in the Full Dataset group and 167 in the Non-BAL group. Please clarify whether the 350 refers to the pooled group before cuts, and state whether those contaminants are excluded from the final catalog or only from the physical interpretation.
  4. [§4.2.3] The BAL anomaly group is described as comprising 672 objects, while Table 4 lists 692 members across the four subtypes. Please correct this internal inconsistency and verify the associated percentage of all anomalies.
  5. [§6, throughout] There are several minor presentation issues: “Plateu” should be “Plateau” in the conclusion list; the element C IV is typeset inconsistently (C iv, Civ, CIV); and the abstract says the redshift range is 1.88 < z < 2.47 while §2 uses 1.88 ≤ z ≤ 2.47. These should be harmonized in the final version.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the anomaly detection is unsupervised and physical attributions rely on external calibrations; only minor non-load-bearing self-citations and a definitional reddening label are noted, plus an unreconciled headline count that is a transparency flaw rather than circularity.

full rationale

Score 2 reflects the absence of load-bearing circularity. The pipeline is unsupervised: PCA is fit to the full and non-BAL samples, K-means clusters the PCA coefficients, and anomalies are points beyond 5 sigma (4 sigma for cluster 3) from cluster centroids; no parameter is fitted to the anomaly sample and then renamed as a prediction. The group labels (C IV Peakers, Excess Si IV emitters, Si IV Deficient, BAL sub-types, reddened sub-types) are assigned after visual inspection of composite spectra, and the physical attributions rest on external diagnostics: Fu et al. (2022) for high C IV EW and Eddington ratio, Nagao et al. (2006) and Hamann et al. (2002) for Si IV/C IV metallicity, and Wu & Shen (2022) for EW/FWHM distributions. These are independent of the PCA/clustering inputs, so the central physical claims do not reduce to the paper's own construction. The only self-citations (Vivek et al. 2012a,b; Vivek et al. 2014) appear in introductory or contextual statements about quasar variability and LoBAL outflows and are not load-bearing. A minor definitional note: Section 2.2 states that the PCA 2 coefficient is directly proportional to reddening, and later sections describe subgroups as separated by degree of reddening via PCA 2 placement; this is partly labeling with the same feature, though the authors also compare against external reddening composites and observed spectral slopes. Per the review rule on flagged limitations, a separate transparency flaw is noted: the abstract's headline '1,888 anomalous quasars' is not reconciled with the body's '1994 and 1270 anomalies' in Section 3.1 or with the Section 6 group counts summing to 1,653; the omitted reconciliation prevents independent reproduction of the headline number but is not a circular argument.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central catalog rests on several choices made by hand: sigma thresholds, k values, the reconstruction-residual cutoff, and preprocessing settings. The physical interpretations additionally assume that literature diagnostics calibrated on typical quasars transfer to these extreme objects. These assumptions do not make the detection circular, but they do mean the catalog's exact membership is not uniquely determined by the data alone.

free parameters (7)
  • sigma threshold for clusters 1 and 2 = 5 sigma
    Chosen by visual inspection of distance histograms in Section 3.1; directly controls the number of anomalies.
  • sigma threshold for cluster 3 = 4 sigma
    Chosen because cluster 3 has a longer tail; changes catalog membership.
  • reconstruction outlier percentile = top 8th percentile
    Spectra with residual error above the 92nd percentile are removed before clustering in Section 2.2; an arbitrary cutoff.
  • primary cluster count k = 3
    Selected by elbow and silhouette methods in Section 3; affects baseline clusters and outlier distances.
  • anomaly group cluster count k = 4
    Selected by elbow method in Section 3.2; determines G1-G4 grouping before manual sub-splitting.
  • Savitzky-Golay smoothing window = 5 pixels
    Preprocessing choice in Section 2.1; alters noise and line profiles before PCA.
  • BAL_PROB threshold = 0.5
    Taken from Guo and Martini 2019; defines the BAL subsample and affects the two-dataset design.
assumptions (5)
  • domain assumption Euclidean distance in 20-dimensional PCA space is a meaningful measure of spectral dissimilarity.
    Invoked in Section 3 for K-means and anomaly scoring; if false, outlier selection is arbitrary.
  • domain assumption The elbow method identifies the true number of clusters.
    Used in Section 3 to select k=3 for the full sample and k=4 for anomaly grouping.
  • domain assumption SDSS DR16 redshifts are accurate enough for rest-frame alignment.
    All spectra are shifted to rest frame using catalog redshifts in Section 2.1; redshift errors could distort line profiles and hence PCA.
  • domain assumption External line diagnostics calibrated on typical quasars transfer to these extreme objects.
    Physical attributions in Section 5 rely on Hamann et al. 2002, Nagao et al. 2006, and Fu et al. 2022 without re-deriving them for this anomalous sample.
  • ad hoc to paper Padding spectra with the trailing flux value does not create clusters dominated by artifacts.
    The procedure in Section 2.1 creates flat sections that are later flagged and removed; the paper does not quantify how much padding remains in the final sample.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Spectroscopic Quasar Anomaly Detection (SQuAD) I: Rest-Frame UV Spectra from SDSS DR16." pith.science (2026). https://pith.science/paper/OCBHMKNH

@misc{pith2026241116858,
  author       = {Pith},
  title        = {Pith review of: Spectroscopic Quasar Anomaly Detection (SQuAD) I: Rest-Frame UV Spectra from SDSS DR16},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OCBHMKNH}},
  note         = {Machine review of arXiv:2411.16858}
}
read the original abstract

We present the results of applying anomaly detection algorithms to a quasar spectroscopic sub-sample from the SDSS DR16 Quasar Catalog, covering the redshift range 1.88 < z < 2.47. Principal Component Analysis (PCA) was employed for dimensionality reduction of the quasar spectra, followed by hierarchical K-Means clustering in a 20-dimensional PCA eigenvector hyperspace. To prevent broad absorption line (BAL) quasars from being identified as the primary anomaly group, we conducted the analysis with and without them, comparing both datasets for a clearer identification of other anomalous quasar types. We identified 1,888 anomalous quasars, categorized into 10 broad groups. The anomalous groups include C IV Peakers-quasars with extremely strong and narrow C IV emission lines; Excess Si IV emitters-quasars where the Si IV line is as strong as the C IV line; and Si IV Deficient anomalies, which exhibit significantly weaker Si IV emission compared to typical quasars. The anomalous nature of these quasars is attributed to lower Eddington ratios for C IV Peakers, super-solar metallicity for Excess Si IV emitters, and sub-solar metallicity for Si IV Deficient anomalies. Additionally, we identified four groups of BAL anomalies: Blue BALs, Flat BALs, Reddened BALs, and FeLoBALs, distinguished primarily by the strength of reddening in these sources. Further, among the non-BAL quasars, we identified three types of reddened anomaly groups classified as heavily reddened, moderately reddened, and plateau-shaped spectrum quasars, each exhibiting varying degrees of reddening. The detected anomalies are presented as a value-added catalog.

Figures

Figures reproduced from arXiv: 2411.16858 by the authors.

Figure 1
Figure 1. Redshift distribution of quasars in the SDSS DR16 catalog (blue) and the selected subset of quasars (yellow) within the specified redshift range, 1.88 ≤ z ≤ 2.47. The red dashed curve represents the comoving distance, while the green dotted curve shows the lookback time as a function of redshift. The corresponding distance and time values at two key redshifts: z = 1.88 and z = 2.47, are marked. Our goal is to identi… view at source ↗
Figure 2
Figure 2. An example quasar spectrum showing the max-normalized flux before (blue) and after pre-processing (black). Note that the pre￾processing effectively reduces noise without altering the overall spectral shape. We initially attempted to utilize the AND and OR masks to obtain a clean spectrum. However, these methods proved inef￾fective as they removed large portions of the spectra, leading to detection as anomalies due t… view at source ↗
Figure 3
Figure 3. The figure shows the cumulative variance as a function of the PCA components for the PCA decomposition with 20 components, for both the Full (red) and Non-BAL Only (green) datasets. In both cases, a total explained variance of 92.1% is achieved with 20 PCA compo￾nents. we obtain a cumulative explained variance greater than 90%, which was achieved with 20 PCA components. This accounted for 92.1% of the total variance… view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: The distribution of residual error for PCA spectral reconstruction is shown for both the Full (top) and Non-BAL Only (bottom) datasets. Quasars within the red shaded region are classified as “reconstruction outliers” and are excluded from the main datasets prior to clu…
Figure 5
Figure 5. Figure 5: The flow chart shows the steps followed by our algorithm, beginning from the quasar sample selection to the final anomaly groups obtained. K-Means is well-suited for large datasets due to its simplicity and efficiency. However, the algorithm requires the user to input …
Figure 6
Figure 6. Figure 6: The figure shows the Sum of Squared Errors (SSE) and Sil￾houette Coefficients as a function of cluster numbers for the Full (left) and Non-BAL Only (right) datasets. The optimal number of clusters is determined using both the elbow (knee) method and the silhouette coef…
Figure 7
Figure 7. Figure 7: Histograms show the distribution of the euclidean distance of each point from its respective cluster centroid for the Full (above panel ) and Non-BAL Only (below panel) datasets respectively. The red shaded region marks the respective threshold limits (5σ for cluster 1…
Figure 8
Figure 8. Figure 8: The figure shows the Sum of Squared Errors (SSE) and Silhou￾ette Coefficients as a function of cluster numbers for the K-Means clus￾tering of anomalous quasars in the Full and Non-BAL Only datasets. The knee of the SSE curve occurs at 4 for both datasets, which coin￾ci…
Figure 9
Figure 9. Figure 9: Top Panel : The figure presents a 2D projection (PCA 1 versus PCA 2 coefficients) of the Full (left) and Non-BAL Only (right) datasets. Each dataset is divided into three clusters (cluster 1: yellow, cluster 2: teal, cluster 3: gray) within the 20-dimensional PCA hyper…
Figure 10
Figure 10. Figure 10: Top Panel : A 2D projection (PCA 1 versus PCA 2 coefficients) of the anomalies of the two datasets as divided into four groups each (group 1: blue, group 2: red, group 3: black, group 4: violet), in the 20 dimensional PCA hyperspace, by the second K-Means clustering a…
Figure 11
Figure 11. Figure 11: Top Panel : 2D projection of BAL anomalies (using PCA 1 versus PCA 2 coefficients) as grouped into 4 types by the secondary K-Means clustering as discussed in §3.2.1. The Reddened and FeLoBALs are placed on the upper region with higher PCA 2 coefficients pertaining to…
Figure 12
Figure 12. Figure 12: The figure shows the composite spectra for all anomaly categories identified in this project. In each plot, the gray dotted spectrum represents the Vanden Berk et al. (2001) composite. All mean spectra are aligned at the Mg ii emission peak to ensure consistent compar…
Figure 13
Figure 13. Figure 13: The figure shows the distribution of C iv equivalent width (EW) and full width at half maximum (FWHM) for all quasars in the Wu & Shen (2022) catalog (blue) compared to the Civ Peakers (orange). The Civ EW for the Civ Peakers is concentrated at the higher end of the d…
Figure 14
Figure 14. Figure 14: Top Panel : The figure shows the distribution of the ratio of Si iv to Civ emission line flux for the Excess Si iv emitters group (orange) compared to all quasars in the Wu & Shen (2022) catalog (blue). The Si iv to C iv ratio centered around 1 signifies an enhanced S…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 52 canonical work pages

  1. [1]

    M., Rosario, D., et al

    Andonie, C., Alexander, D. M., Rosario, D., et al. 2022, Monthly Notices of the Royal Astronomical Society, 517, 2577

    work page 2022

  2. [2]

    1993, Publications of the Astronomical Society of the Pacific, 105, 1032

    Artymowicz, P. 1993, Publications of the Astronomical Society of the Pacific, 105, 1032

    work page 1993

  3. [3]

    H., White, R

    Becker, R. H., White, R. L., Gregg, M. D., et al. 2000, The Astrophysical Journal, 538, 72

    work page 2000

  4. [4]

    C., Blandford, R

    Begelman, M. C., Blandford, R. D., & Rees, M. J. 1984, Rev. Mod. Phys., 56, 255

    work page 1984

  5. [5]

    Bradley, P. S. & Fayyad, U. M. 1998, in International Conference on Machine Learning

    work page 1998

  6. [6]

    M., Thomas, R., & Baron, E

    Branch, D., Leighly, K. M., Thomas, R., & Baron, E. 2002, The Astrophysical Journal, 578, L37

    work page 2002

  7. [7]

    D., Becker, R., et al

    Brotherton, M., Tran, H. D., Becker, R., et al. 2001, The Astrophysical Journal, 546, 775

    work page 2001

  8. [8]

    Carnall, A. C. 2017, SpectRes: A Fast Spectral Resampling Tool in Python

    work page 2017

Show all 58 references
  1. [9]

    M., Terndrup, D

    Choi, H., Leighly, K. M., Terndrup, D. M., et al. 2022, ApJ, 937, 74

    2022

  2. [10]

    2002, Astronomy & Astrophysics, 388, 771

    Collin, S., Boisson, C., Mouchet, M., et al. 2002, Astronomy & Astrophysics, 388, 771

    2002

  3. [11]

    M., Fan, X., Brandt, W., et al

    Diamond-Stanic, A. M., Fan, X., Brandt, W., et al. 2009, The Astrophysical Jour- nal, 699, 782 D.Napoleon, S. 2011, International Journal of Computer Applications, 13, 41

    2009

  4. [12]

    2010, The Astrophysical Journal, 717, 868

    Farrah, D., Urrutia, T., Lacy, M., et al. 2010, The Astrophysical Journal, 717, 868

    2010

  5. [13]

    J., Baldwin, J

    Ferland, G. J., Baldwin, J. A., Korista, K. T., et al. 1996, The Astrophysical Journal

    1996

  6. [14]

    2022, The Astrophysical Journal, 934, 97

    Fu, S., Brandt, W., Zou, F., et al. 2022, The Astrophysical Journal, 934, 97

    2022

  7. [15]

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

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

    2022

  8. [16]

    D., Lacy, M., White, R

    Gregg, M. D., Lacy, M., White, R. L., et al. 2002, The Astrophysical Journal, 564, 133

    2002

  9. [17]

    & Martini, P

    Guo, Z. & Martini, P. 2019, The Astrophysical Journal, 879, 72

    2019

  10. [18]

    & Maraschi, L

    Haardt, F. & Maraschi, L. 1991, ApJ, 380, L51

    1991

  11. [19]

    B., Anderson, S

    Hall, P. B., Anderson, S. F., Strauss, M. A., et al. 2002, The Astrophysical Journal Supplement Series, 141, 267

    2002

  12. [20]

    J., Warner, C., & Baldwin, J

    Hamann, F., Korista, K., Ferland, G. J., Warner, C., & Baldwin, J. 2002, The Astrophysical Journal, 564, 592

    2002

  13. [21]

    L., Ross, N., et al

    Hamann, F., Zakamska, N. L., Ross, N., et al. 2017, Monthly Notices of the Royal Astronomical Society, 464, 3431

    2017

  14. [22]

    W., Jensen, T

    Harris, D. W., Jensen, T. W., Suzuki, N., et al. 2016, AJ, 151, 155

    2016

  15. [23]

    J., Gandy, A., & Mortlock, D

    Henrion, M., Hand, D. J., Gandy, A., & Mortlock, D. J. 2013, Statistical Analysis and Data Mining: The ASA Data Science Journal, 6, 53

    2013

  16. [24]

    C., Schmidt, G

    Hines, D. C., Schmidt, G. D., Gordon, K. D., et al. 2001, The Astrophysical Journal, 563, 512

    2001

  17. [25]

    F., Strauss, M

    Hopkins, P. F., Strauss, M. A., Hall, P. B., et al. 2004, The Astronomical Journal, 128, 1112

    2004

  18. [26]

    & Blaes, O

    Koratkar, A. & Blaes, O. 1999, PASP, 111, 1

    1999

  19. [27]

    M., Cales, S., Moran, E

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

    2015

  20. [28]

    Liu, Z. et al. 2011, Procedia Environmental Sciences, 11, 256

    2011

  21. [29]

    W., Higley, A

    Lyke, B. W., Higley, A. N., McLane, J. N., et al. 2020, The Astrophysical Journal Supplement Series, 250, 8

    2020

  22. [30]

    1996, Astrophysical Journal Supplement v

    Marziani, P., Sulentic, J., Dultzin-Hacyan, D., Calvani, M., & Moles, M. 1996, Astrophysical Journal Supplement v. 104, p. 37, 104, 37

    1996

  23. [31]

    Menou, K., Berk, D. E. V ., Ivezi ´c, Ž., et al. 2001, The Astrophysical Journal, 561, 645

    2001

  24. [32]

    2012, Astronomy & Astro- physics, 541, A77

    Meusinger, H., Schalldach, P., Scholz, R.-D., et al. 2012, Astronomy & Astro- physics, 541, A77

    2012

  25. [33]

    2006, Astronomy & Astrophysics, 447, 157

    Nagao, T., Marconi, A., & Maiolino, R. 2006, Astronomy & Astrophysics, 447, 157

    2006

  26. [34]

    R., Long, D

    Newman, P. R., Long, D. C., Snedden, S. A., et al. 2004, in Ground-based In- strumentation for Astronomy, V ol. 5492, SPIE, 533–544

    2004

  27. [35]

    2022, Monthly Notices of the Royal Astro- nomical Society, 511, 5251

    Ni, Q., Brandt, W., Luo, B., et al. 2022, Monthly Notices of the Royal Astro- nomical Society, 511, 5251

    2022

  28. [36]

    2018, Monthly Notices of the Royal Astro- nomical Society, 480, 5184

    Ni, Q., Brandt, W., Luo, B., et al. 2018, Monthly Notices of the Royal Astro- nomical Society, 480, 5184

    2018

  29. [37]

    Peterson, B. M. 1997, Taxonomy of Active Galactic Nuclei (Cambridge Univer- sity Press), 21–31

    1997

  30. [38]

    M., Anderson, S

    Plotkin, R. M., Anderson, S. F., Hall, P. B., et al. 2008, The Astronomical Journal, 135, 2453

    2008

  31. [39]

    2021, The Astrophysical Journal Supplement Series, 253, 28

    Rakshit, S., Stalin, C., Kotilainen, J., & Shin, J. 2021, The Astrophysical Journal Supplement Series, 253, 28

    2021

  32. [40]

    2021, Astronomy and Computing, 34, 100437

    Reis, I., Rotman, M., Poznanski, D., Prochaska, J., & Wolf, L. 2021, Astronomy and Computing, 34, 100437

    2021

  33. [41]

    T., Hall, P

    Richards, G. T., Hall, P. B., Berk, D. E. V ., et al. 2003, The Astronomical Journal, 126, 1131

    2003

  34. [42]

    P., Hamann, F., Zakamska, N

    Ross, N. P., Hamann, F., Zakamska, N. L., et al. 2015, MNRAS, 453, 3932 Sánchez Almeida, J., Aguerri, J. A. L., Muñoz-Tuñón, C., & de Vicente, A. 2010, ApJ, 714, 487 Article number, page 17 of 20 A&A proofs: manuscript no. SQuAD1AA

    2015

  35. [43]

    2011, in 2011 31st interna- tional conference on distributed computing systems workshops, IEEE, 166– 171

    Satopaa, V ., Albrecht, J., Irwin, D., & Raghavan, B. 2011, in 2011 31st interna- tional conference on distributed computing systems workshops, IEEE, 166– 171

    2011

  36. [44]

    & Golay, M

    Savitzky, A. & Golay, M. J. E. 1964, Analytical Chemistry, 36, 1627

    1964

  37. [45]

    Shakura, N. I. & Sunyaev, R. A. 1973, A&A, 24, 337

    1973

  38. [46]

    2022, Astronomy & Astrophysics, 668, A121

    Shao, Y ., Wang, R., Weiss, A., et al. 2022, Astronomy & Astrophysics, 668, A121

    2022

  39. [47]

    Shen, Y . & Ho, L. C. 2014, Nature, 513, 210

    2014

  40. [48]

    A., Gunn, J

    Smee, S. A., Gunn, J. E., Uomoto, A., et al. 2013, The Astronomical Journal, 146, 32

    2013

  41. [49]

    M., et al

    Solarz, A., Thomas, R., Montenegro-Montes, F. M., et al. 2020, A&A, 642, A103

    2020

  42. [50]

    K., Rochman, E., & Satoto, B

    Syakur, M., Khotimah, B. K., Rochman, E., & Satoto, B. D. 2018, in IOP con- ference series: materials science and engineering, V ol. 336, IOP Publishing, 012017

    2018

  43. [51]

    R., Hall, P

    Trump, J. R., Hall, P. B., Reichard, T. A., et al. 2006, The Astrophysical Journal Supplement Series, 165, 1 Vanden Berk, D. E., Richards, G. T., Bauer, A., et al. 2001, AJ, 122, 549

    2006

  44. [52]

    2014, MNRAS, 440, 799

    Vivek, M., Srianand, R., Petitjean, P., et al. 2014, MNRAS, 440, 799

    2014

  45. [53]

    2019, Proceedings of the International Astronomical Union, 15, 285

    Wethers, C., Kotilainen, J., Schramm, M., & Schulze, A. 2019, Proceedings of the International Astronomical Union, 15, 285

    2019

  46. [54]

    & Combes, F

    Wiklind, T. & Combes, F. 1996, Nature, 379, 139

    1996

  47. [55]

    C., Vanden Berk, D

    Wilhite, B. C., Vanden Berk, D. E., Kron, R. G., et al. 2005, ApJ, 633, 638

    2005

  48. [56]

    & Shen, Y

    Wu, Q. & Shen, Y . 2022, The Astrophysical Journal Supplement Series, 263, 42

    2022

  49. [57]

    Xiong, L., Poczos, B., Connolly, A., & Schneider, J. 2018

    2018

  50. [58]

    A., Telfer, R

    Zheng, W., Kriss, G. A., Telfer, R. C., Grimes, J. P., & Davidsen, A. F. 1997, The Astrophysical Journal, 475, 469 Article number, page 18 of 20 Arihant Tiwari and M. Vivek: Spectroscopic Quasar Anomaly Detection (SQuAD) The PCA Eigenvectors for the primary K-Means clustering ...

    2001

Pith tools

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