Pith. sign in

REVIEW 3 major objections 5 minor 78 references

Estimating Bolometric Luminosities of Type 1 Quasars with Self-Organizing Maps

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A self-organizing map trained on 11 photometric bands estimates bolometric luminosities of type 1 quasars with about 4.9% scatter, roughly four times better than single-band bolometric corrections.

desk verdict Useful methods paper with code and catalog, and the headline 4.9% scatter should be read as internal precision, not demonstrated absolute accuracy. read the letter →

arxiv 2506.04329 v2 pith:2MKIWNVV submitted 2025-06-04 astro-ph.CO astro-ph.GAastro-ph.HE

Jie Chen , Linhua Jiang , Shengxiu Sun , Zijian Zhang , Mouyuan Sun This is my paper
classification astro-ph.COastro-ph.GAastro-ph.HE
keywords bolometricluminositytype1quasarsself-organizingmapsspectralenergydistributionscorrectionsmulti-bandphotometrySDSSDR16Qmachinelearning
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 self-organizing map trained on 11 rest-frame monochromatic luminosities, from 3.2 $\mu$m to 2 keV, can estimate the bolometric luminosity of type 1 quasars with $\sigma_f \approx 4.9\%$, roughly four times better than single-band bolometric corrections. The authors build a fiducial sample of 79,163 SDSS quasars at $0.5\leq z\leq 2$, correct their multi-band photometry for Galactic extinction, intergalactic-medium absorption, host-galaxy light, and emission lines, and integrate the resulting spectral energy distributions (SEDs) from 4 $\mu$m to 10 keV to define $L_{\rm bol}$. The trained map assigns each new quasar to a cell of similar SEDs and uses the cell's median luminosity as the estimate. If this holds, large spectroscopic quasar samples can be assigned bolometric luminosities quickly from public survey photometry, with uncertainties roughly a factor of four smaller than traditional bolometric corrections.

What carries the argument

The central object is the self-organizing map (SOM), an unsupervised neural network that projects the 11-dimensional space of observed rest-frame monochromatic luminosities onto a 45$\times$45 grid of cells while preserving topological order. After training on 80% of the fiducial sample, each cell contains quasars with similar SEDs, and the median and dispersion of the true $\log L_{\rm bol}$ within that cell are assigned to any new quasar whose best-matching cell is found. Missing bands, notably W3 and W4, are filled with k-nearest-neighbor imputation before projection, which lets the same trained map handle quasars with limited photometry. The map's topology-based averaging is what converts multi-band SED similarity into a luminosity estimate.

What would settle it

Take a subsample of quasars with actual X-ray detections and individually decomposed host galaxies, compute $L_{\rm bol}$ by integrating the observed SEDs from the far-IR to hard X-rays, and compare with SOM estimates from the released code; if the scatter is substantially larger than $\sigma_f\approx 5\%$ or shows a luminosity-dependent offset, the small scatter is an artifact of training and testing on the same corrected photometry.

Watch

Extended reading notes

Core claim

The paper's central claim is that a 45$\times$45 self-organizing map trained on 11 rest-frame monochromatic luminosities reproduces the bolometric luminosities of type 1 quasars with fractional residual scatter $\sigma_f \approx 4.9\%$, compared with 18$-$29% for single-band bolometric corrections and 13.6% for three-band multiple linear regression. The map is trained on SEDs constructed from multi-band photometry after corrections for Galactic extinction, IGM absorption, host-galaxy contamination, and emission lines, and each cell's median $\log L_{\rm bol}$ becomes the estimator for any quasar mapped to that cell. The authors apply the trained model to 633,373 SDSS DR16Q quasars, release the code and the resulting catalog, and recalculate bolometric corrections at 1450, 3000, 5100, and 10,000 \AA{} rest-frame wavelengths together with updated multi-linear regression relations.

Load-bearing premise

The load-bearing premise is that the 'true' luminosities used to train and test the map are accurate, since they come from integrating the same corrected photometry that feeds the map; any bias in the host-galaxy, intergalactic-medium, or predicted X-ray corrections would move targets and estimates together, and the paper itself notes that only interpolation error is included in its uncertainties, so the 4.9% scatter does not bound the absolute error.

Editorial extensions

If this is right

  • For the 633,373 DR16Q quasars with released estimates, bolometric luminosities now come with documented per-object systematic uncertainties instead of a single assumed bolometric correction.
  • Recalibrated bolometric corrections at 1450, 3000, and 5100 \AA{} (2.93, 3.72, and 6.43) make the luminosity dependence of the correction explicit, so future work can avoid the bias of a constant bolometric correction.
  • Quasars with incomplete photometry remain usable: KNN-imputed SEDs recover $L_{\rm bol}$ with $\sigma_f$ of roughly 13$-$16% when four to eight bands are available.
  • Because the method handles a wide range of luminosity and redshift and needs no low-resolution SED fitting, it can be applied directly to the next generation of wide multi-band quasar samples.

Reading between the lines

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

  • One extension the authors leave implicit is that the same trained map can serve as a general SED classifier, predicting any rest-frame band luminosity or color from partial photometry, not just $L_{\rm bol}$.
  • Because 85% of the training targets use X-ray luminosities predicted from the UV relation, the 4.9% figure should be read as precision relative to those corrected targets; an absolute check needs quasars with measured X-ray fluxes and individually decomposed host galaxies.
  • The paper's own limited-data tests imply that UV coverage, not IR coverage, controls accuracy: dropping to four NUV-to-optical bands raises $\sigma_f$ to about 14%, so future surveys that add near-UV photometry will benefit most.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This paper proposes a self-organizing map (SOM) method for estimating bolometric luminosities of unobscured type 1 quasars from multi-band photometry. The authors build a fiducial sample of 79,163 SDSS DR16 quasars at 0.5≤z≤2 with GALEX, SDSS, UKIDSS/UHS/VHS, WISE, and X-ray data, correct for Galactic extinction, IGM absorption, host galaxy contamination, and emission lines, integrate the rest-frame SEDs over 4 µm–10 keV to define Lbol, and train a 45×45 SOM on 11 monochromatic luminosities from 3.2 µm to 2 keV. They report a fractional residual dispersion σf≈4.9% against their own Lbol values, substantially smaller than single- or multi-band linear regression and bolometric corrections, and they apply the method to 633,373 DR16Q quasars, releasing code and a catalog.

Significance. If the claimed accuracy were absolute, the method would be a valuable community resource: it is fast, uses SED diversity rather than a single band, includes a held-out test set, checks selection effects, tests limited-data recovery, and ships public code and a large catalog. The comparison with bolometric corrections and linear regressions on the same labels is internally instructive and demonstrates that the SOM recovers the authors' own Lbol definition far better than single-band methods. However, the central claim that the method 'reduces systematical uncertainties' is not yet supported, because the labels are constructed from the same photometry that provides the SOM inputs; the small σf measures self-consistency, not absolute accuracy. The only independent comparison, against Wu & Shen (2022), shows a median offset an order of magnitude larger than the quoted scatter.

major comments (3)
  1. [§4.3, §4.5, Figs. 10 and 13] The central quantitative claim, σf≈4.9% in §4.3/Figure 10, is a recovery test of the authors' own Lbol definition. The 'true' Lbol is obtained by integrating SEDs built from the same photometric bands that form the 11 SOM inputs, after the same corrections: IGM absorption (§3.1), host galaxy subtraction (§3.3), emission-line corrections (§3.4), interpolation/imputation (§3.5), and the Steffen et al. X-ray relation for 85% of sources (§2.4). Biases in any of these corrections shift labels and features coherently, so the small residual scatter does not bound the absolute error. This is not only a philosophical point: §4.5/Figure 13 reports a 14.8% median offset and 43.6% 1σ scatter against the independent Wu & Shen (2022) estimates, an order of magnitude larger than σf. The authors themselves state in §3.5 that 'the uncertainty of bolometric luminosity is underestimated in this work' and in §5 that systematic uncertainties will be characterized in future work. I therefore request an external validation, e.g., comparison with SED-fitting-based Lbol values (Lusso et al. 2012; Duras et al. 2020) for a matched sample, or a simulation with injected SEDs of known Lbol, together with a propagation of the dominant correction uncertainties (host fraction, IGM transmission, Steffen relation scatter) into the predicted Lbol. Without this, the abstract's claim that the method 'reduces systematical uncertainties' is premature.
  2. [§4.5, §4.6, Table 5] The application to DR16Q in §4.5 and the code's scaling option in §4.6 extrapolate the SOM beyond its training domain. The fiducial SOM is trained on 0.5≤z≤2 quasars, but the released catalog covers z<5 and includes a scaling parameter for faint (log Lbol < 45.13) and bright (>46.74) ends with no validation shown. The IGM transmission correction is extended to z=5 and to u, g, r bands, which were not part of the §3.1 calculation, and the per-source uncertainties in the catalog do not include extrapolation error. I ask for a validation subset (e.g., z>2 or very luminous quasars with X-ray detections or SED-fitting Lbol) or, alternatively, an explicit statement that the claimed σf≈4.9% and the released uncertainties apply only within the training domain.
  3. [§3.5, §4.1, §4.4] Section 3.5 states that about 20% of the fiducial SEDs do not reach 4 µm and uses KNNImputer to fill in the missing W3/W4 photometry before computing Lbol; these same imputed values enter both the SOM input features and the target labels. The sentence that conclusions remain 'quantitatively consistent' when excluding these samples is not accompanied by the actual numbers. More importantly, it is unclear whether the KNN imputation is performed before the 80/20 train/test split described in §4.1. If the imputation models are fit on the full sample, test objects can contribute to their own imputation, which would artificially lower the test-set σf. Please clarify the order of operations and, if necessary, rerun the test-set validation with imputation fit only on the training set, reporting the affected values.
minor comments (5)
  1. [§3.1] The text refers to 'Figure 3.1' for the IGM transmission example, but figures are numbered sequentially elsewhere (Figure 1 is the redshift/L2500 plot in §2). Please renumber or reference the actual figure.
  2. [Table 5] Table 5 lists the default integration range as 'from 4 µm to 2 keV' and describes 'wave range' as being in Hz, while §3.5 and §5 state the default range is 4 µm–10 keV. Please make the default range and units consistent between the text, table, and code.
  3. [Appendix A, Table 6] The column numbering in Table 6 repeats '4' twice (X RAY MATCHED and Chandra FLUX) and the subsequent column numbers are shifted; the column headings should be renumbered.
  4. [§2.1] The flag name 'ERRBITS=0' appears to be a typo; if it refers to UKIDSS/UHS quality flags, please use the actual flag names from those catalogs.
  5. [§4.3] The σf values quoted in the text for the BC method at 1450, 3000, and 5100 Å are given as 20.3%, 18.0%, and 29.0%, while Table 2 reports 0.203, 0.180, and 0.290 in fractional units; please unify the presentation of these numbers.

Circularity Check

2 steps flagged · score 6.0 of 10

The 4.9% scatter is internally anchored: 'true' Lbol and the SOM inputs derive from the same photometry and the same statistical corrections, so the claimed accuracy does not bound absolute systematic error.

  1. self definitional [Sections 3.5 and 4.1 (Eq. 2; SOM input definition)]
    "Bolometric luminosity is the integrated area under the SED curve. Its mathematical definition is as follows, Lbol = Z ∞ 0 Lνdν = Z ∞ −∞ ln(10)νLνd log(ν). ... We use an integration range of 4 µm-10 keV (default range) in this paper. ... In this work, the input parameter space has 11 dimensions representing the rest-frame quasar SEDs (i.e., 11 monochromatic luminosities)."

    The 11 input features are monochromatic luminosities drawn from the same rest-frame SED whose integral defines Lbol. The label is not an independent measurement: Lbol is a deterministic functional of the same photometry and the same corrections (IGM transmission, host-galaxy fraction, KNN-imputed W3/W4, and the Steffen et al. L_UV-L_X relation for 85% of quasars). A SOM can therefore recover Lbol with small residuals by learning the quadrature-like mapping from the 11 fluxes to their own integral. Any systematic error in the shared corrections shifts inputs and labels together and is invisible in the residual. The train/test split only verifies interpolation of the authors' own Lbol definition, not agreement with an absolute measurement.

  2. other [Section 2.4 (Steffen et al. relation) combined with Section 3.5 (integration range) and Section 4.1 (2 keV input)]
    "For quasars without X-ray detections (85%), we estimate their X-ray flux using the correlation defined by the LUV − LX relation based on the luminosities at 2500 Å and 2 keV. Steffen et al. (2006) used 333 quasars ... log(L2keV) = (0.721±0.011) log(L2500)+(4.531±0.688)."

    For 85% of the sample, the 2 keV luminosity is not observed but is predicted from L2500. That same predicted 2 keV flux is both one of the 11 SOM inputs and part of the integrated Lbol (the default range is 4 µm-10 keV). Any error in the Steffen relation is therefore common-mode: it enters the features and the label identically and cannot appear in the quoted σf. The reported scatter tests the internal consistency of the adopted corrections, not the absolute accuracy of the X-ray-inclusive Lbol.

full rationale

The paper does not rely on a load-bearing self-citation chain, and the SOM itself is unsupervised with a held-out test set; there is no uniqueness theorem or ansatz smuggled in via citation. The core circularity is structural: the 'true' Lbol used for training and testing is computed by integrating the very SEDs that provide the 11 input features, after the same statistical corrections are applied to both sides. Hence the headline accuracy (σf ≈ 4.9%) measures how well a smooth function of 11 SED points reproduces the integral of those same SED points, not agreement with an independent bolometric measurement. The external comparison with Wu & Shen (2022) shows a median offset of 14.8% and a 1σ scatter of 43.6%, an order of magnitude larger than the quoted σf, confirming that the claimed precision is internal. The MLR relations are also fit and evaluated on the same sample. The released code and catalog remain useful as a fast, reproducible estimator of the authors' own Lbol definition, but the central claim of reduced systematic uncertainty is not yet supported by an absolute accuracy assessment.

Assumptions & free parameters 6 free parameters · 8 assumptions · 0 invented entities

The central Lbol values rest on several adopted calibrations (host fraction, IGM transmission, L_UV-L_X, line equivalent widths) plus hand-chosen algorithmic settings (SOM size, KNN neighbors, integration limits). None of these is fitted in this paper to an external standard, so they are carried into the claim as assumptions.

free parameters (6)
  • Host galaxy fraction at 5100 A (Jalan et al. 2023 relation) = adopted relation, not refitted; reduces Lbol by about 0.05 dex on average
    Used in Section 3.3 to remove host galaxy contamination; the relation has large scatter and the spiral template choice shifts the zero point of Lbol.
  • IGM transmission model parameters = b=30 km/s; absorber distribution from Faucher-Giguere (2020); power-law index -0.44
    Adopted in Section 3.1 to correct FUV and NUV photometry; these choices directly set the UV flux and hence Lbol.
  • KNNImputer number of neighbors = 100
    Chosen in Sections 3.5 and 4.4 to impute missing W3 and W4 and to recover limited SEDs; affects reconstructed SEDs for about 20% of full-coverage quasars and all limited-SED applications.
  • SOM map size and training settings = 45x45 map; sampling rate 90%; quantization error 0.11
    Chosen by hand in Section 4.1; controls resolution and scatter of the Lbol estimator.
  • Bolometric integration range = 4 um to 10 keV (default)
    Chosen in Section 3.5; bolometric correction values depend strongly on the long-wavelength boundary, and changing the range changes Lbol by construction.
  • Scaling parameter for faint and bright ends = option in code; thresholds log(Lbol)<45.13 and >46.74
    Added in Section 4.6 to handle low-sample ends; iterative scaling can bias recovered luminosities at the extremes.
assumptions (8)
  • domain assumption Adopted flat cosmology H0=70 km/s/Mpc, Omega_M=0.3, Omega_Lambda=0.7.
    Stated in Section 1 and used for luminosity distances; standard but a choice.
  • domain assumption Integration of the SED from 4 um to 10 keV yields the bolometric luminosity; reprocessed IR and coronal X-rays are excluded by default.
    Section 3.5; the definition of bolometric luminosity is contingent on how reprocessed radiation is treated, and the boundary choice affects all Lbol values.
  • domain assumption Isotropic emission is assumed; optional viewing-angle correction f=0.75 from Nemmen and Brotherton (2010).
    Section 3.6; anisotropic emission would change the true luminosity, and the correction is optional in code.
  • domain assumption Steffen et al. (2006) L_UV-L_X relation applies to the 85% of quasars without X-ray detections.
    Section 2.4, Eq. (1); used to estimate 2 keV flux and thus the X-ray contribution to Lbol for most sources.
  • domain assumption IGM transmission model with Poisson-distributed HI clouds (Faucher-Giguere 2020) correctly describes average LyC and LyL absorption.
    Section 3.1; adopted for FUV and NUV correction; wrong transmission directly shifts the UV SED and Lbol.
  • domain assumption Host galaxy fraction at 5100 A from Jalan et al. (2023) and the spiral template of Assef et al. (2010) describe host contamination for this sample.
    Section 3.3; applied as a deterministic correction despite the large scatter in the underlying relation.
  • domain assumption Mg II equivalent width relations (based on Wu and Shen 2022 continuum fits) can infer Ly alpha, C IV, H beta, and H alpha line fluxes for the K-correction.
    Section 3.4; the line corrections change Lbol by about 0.02 dex and depend on these empirical relations.
  • ad hoc to paper GALEX FUV nonlinearity is negligible for sources fainter than the 16 mag calibration threshold.
    Section 2.3; the paper explicitly assumes no correction is needed because the sample is fainter than the calibrated regime.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Estimating Bolometric Luminosities of Type 1 Quasars with Self-Organizing Maps." pith.science (2026). https://pith.science/paper/2MKIWNVV

@misc{pith2026250604329,
  author       = {Pith},
  title        = {Pith review of: Estimating Bolometric Luminosities of Type 1 Quasars with Self-Organizing Maps},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2MKIWNVV}},
  note         = {Machine review of arXiv:2506.04329}
}
abstract

We present a new method to calculate bolometric luminosities for unobscured, type 1 quasars with multi-band photometric data. Bolometric luminosity is a fundamental property to understand quasars and it is commonly estimated from monochromatic luminosities using bolometric corrections that often neglect quasar SED diversity. We take advantage of the fact that most quasars now have multi-band observations from UV to mid-IR, and construct SEDs for a well-defined sample of SDSS quasars at $0.5\leq z\leq 2$. Based on this fiducial sample, we explore quasar SEDs, their diversity, and their relations with bolometric luminosities. We then use unsupervised neural network self-organizing maps (SOM) to describe the SED diversity and compute the bolometric luminosities with a fully-trained SOM model. This method reduces systematical uncertainties compared to the traditional method. In addition, we update the multi-linear regression relations between bolometric luminosity and monochromatic luminosities at restframe 1450\r{A}, 3000\r{A}, and 5100\r{A}. Our method is applicable to large quasar samples with a wide range of luminosity and redshift. We have applied it to the SDSS DR16 quasars. We have also made our code publicly available.

Figures

Figures reproduced from arXiv: 2506.04329 by the authors.

Figure 1
Figure 1. Redshift and L2500 distributions of the quasars in our sample. The quasars are represented by the blue con￾tours. The contours indicate 0.005, 0.25, 0.5, 0.75, and 0.95 of the normalized distributions. The histograms of redshift and luminosity are normalized by the total number of the quasars. total flux to describe them in each band. For SDSS, we use PSFMAG. For other data, we adopt aperture mag￾nitudes with apertu… view at source ↗
Figure 2
Figure 2. Multi-band SEDs of the quasars in our sample. The upper panel shows the number of SEDs averaged at each frequency. In the lower panel, the solid curve represents the median SED of all quasars in the sample and the dashed curves represent its 1σ ranges. The data points are photometric data of randomly selected 800 quasars. 2.2. Mid-IR Data We match 128,498 SDSS quasars (with a match length of 1′′) to the AllWISE Data… view at source ↗
Figure 3
Figure 3. shows the diversity of the quasar SEDs in the UV to IR range. In this figure, we display 4000 randomly selected quasars from our sample and their SEDs are normalized at 5100 ˚A. These 4000 UV-IR SEDs follow the same L5100 distribution of the whole sample. The large scatter suggests a highly diverse nature of quasar SEDs that cannot be well explained by a single quasar template. Note that the y-axis is on a logarithm… view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Relation between bolometric luminosity and monochromatic luminosity at 1450 ˚A, 3000 ˚A, 5100 ˚A, and 10,000 ˚A. The black lines indicate the BC fitting and the red lines show the linear regression. In practice, the observed wavelengths are limited. The choice of the i…
Figure 5
Figure 5. Figure 5: Bolometric correction as a function of bolometric luminosity. The dashed horizontal lines represent the median values of the bolometric corrections. wavelengths, i.e., 1450 ˚A, 3000 ˚A, 5100 ˚A, and 10,000 ˚A. We optimize the multiple linear regression (MLR) by minimiz…
Figure 6
Figure 6. Figure 6: Distribution of 11 normalized monochromatic luminosities on the 45×45 trained SOM. matic luminosities in the range of 3.2 µm to 760 ˚A, along with the 2 keV luminosity. Our photometric data are available for most quasars in this wavelength range. To evaluate the precis…
Figure 8
Figure 8. Figure 8: Six median quasar SEDs (solid curves) with 1σ dispersions (shaded area) from six cells, i.e., the black open circles in [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 7
Figure 7. Figure 7: Upper panel: SOM with cells color-coded by the median values of log(Lbol). The gray contours represent the levels of the log(Lbol) values. The open black squares represent six cells: Cell 1 = [33, 38], Cell 2 = [12, 38], Cell 3 = [12, 23], Cell 4 = [33, 18], Cell 5 = […
Figure 10
Figure 10. Figure 10: Fractional residuals from different methods, in￾cluding BCs at 1450 ˚A, 3000 ˚A, and 5100 ˚A, linear regres￾sion with monochromatic luminosities at 1450 ˚A, 3000 ˚A, and 5100 ˚A, multiple linear regression with monochromatic lumi￾nosities at 1450 ˚A, 3000 ˚A, and 5100…
Figure 9
Figure 9. Figure 9: Comparison between the true log(Lbol) and SOM￾derived log(Lbol) values in the training and test sets. The red line is the one-to-one line. We calculate the median and dispersion of fractional residuals between the SOM-derived bolometric luminos￾ity and the true value. …
Figure 11
Figure 11. Figure 11: An example for KNNImputer to recover limited SEDs. The upper panel shows input and recovered SEDs. The input data only contains 4 band luminosities from NUV to optical (red crosses). The gray dots denote the recov￾ered SED by KNNImputer, and the red dashed curve rep￾r…
Figure 13
Figure 13. Figure 13: Comparison between log(Lbol) from this work and log(Lbol, Wu) from Wu & Shen (2022) for recalculated DR16Q sample. The contour level is labeled with a closed percentile. The gray dashed lines, from top to bottom, rep￾resent a line one dex higher, the one-to-one line, …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

78 extracted references · 7 canonical work pages

  1. [1]

    2016, MNRAS, 456, 1618, doi: 10.1093/mnras/stv2703

    Acquaviva, V. 2016, MNRAS, 456, 1618, doi: 10.1093/mnras/stv2703

    work page doi:10.1093/mnras/stv2703 2016

  2. [2]

    J., Kochanek, C

    Assef, R. J., Kochanek, C. S., Brodwin, M., et al. 2010, ApJ, 713, 970, doi: 10.1088/0004-637X/713/2/970 17

    doi:10.1088/0004-637x/713/2/970 2010

  3. [3]

    2023, ApJ, 957, 19, doi: 10.3847/1538-4357/acf21a

    Auge, C., Sanders, D., Treister, E., et al. 2023, ApJ, 957, 19, doi: 10.3847/1538-4357/acf21a

    doi:10.3847/1538-4357/acf21a 2023

  4. [4]

    M., & Brunner, R

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

    doi:10.1142/s0218271810017160 2010

  5. [5]

    2019, arXiv e-prints, arXiv:1904.07248, doi: 10.48550/arXiv.1904.07248

    Baron, D. 2019, arXiv e-prints, arXiv:1904.07248, doi: 10.48550/arXiv.1904.07248

    arXiv doi:10.48550/arxiv.1904.07248 2019

  6. [6]

    Barthel, P. D. 1989, ApJ, 336, 606, doi: 10.1086/167038

    doi:10.1086/167038 1989

  7. [7]

    2014, Ap&SS, 354, 103, doi: 10.1007/s10509-014-1935-6

    Bianchi, L. 2014, Ap&SS, 354, 103, doi: 10.1007/s10509-014-1935-6

    doi:10.1007/s10509-014-1935-6 2014

  8. [8]

    2017, ApJS, 230, 24, doi: 10.3847/1538-4365/aa7053

    Bianchi, L., Shiao, B., & Thilker, D. 2017, ApJS, 230, 24, doi: 10.3847/1538-4365/aa7053

    doi:10.3847/1538-4365/aa7053 2017

Show all 78 references
  1. [9]

    R., et al

    Brightman, M., Balokovi´ c, M., Ballantyne, D. R., et al. 2017, ApJ, 844, 10, doi: 10.3847/1538-4357/aa75c9

    2017 doi

  2. [10]

    2023, Bulletin of the AAS, 55

    Bruursema, J., Vrba, F., Munn, J., et al. 2023, Bulletin of the AAS, 55

    2023

  3. [11]

    2023, Nature Astronomy, 7, 1506, doi: 10.1038/s41550-023-02088-5 Calistro Rivera, G., Lusso, E., Hennawi, J

    Cai, Z.-Y., & Wang, J.-X. 2023, Nature Astronomy, 7, 1506, doi: 10.1038/s41550-023-02088-5 Calistro Rivera, G., Lusso, E., Hennawi, J. F., & Hogg, D. W. 2016, ApJ, 833, 98, doi: 10.3847/1538-4357/833/1/98

    2023 doi

  4. [12]

    Camarota, L., & Holberg, J. B. 2014, MNRAS, 438, 3111, doi: 10.1093/mnras/stt2422

    2014 doi

  5. [13]

    Chatfield, C., & Collins, A. J. 1980, Principal component analysis (Boston, MA: Springer US), 57–81, doi: 10.1007/978-1-4899-3184-9 4

    1980 doi

  6. [14]

    M., Wright, E

    Cutri, R. M., Wright, E. L., Conrow, T., et al. 2021, VizieR Online Data Catalog: AllWISE Data Release (Cutri+ 2013), VizieR On-line Data Catalog: II/328. Originally published in: IPAC/Caltech (2013)

    2013

  7. [15]

    L., et al

    Davidzon, I., Laigle, C., Capak, P. L., et al. 2019, arXiv e-prints, arXiv:1905.13233. https://arxiv.org/abs/1905.13233

    2019 arXiv

  8. [16]

    2020, A&A, 636, A73, doi: 10.1051/0004-6361/201936817

    Duras, F., Bongiorno, A., Ricci, F., et al. 2020, A&A, 636, A73, doi: 10.1051/0004-6361/201936817

    2020 doi

  9. [17]

    A., et al

    Dye, S., Lawrence, A., Read, M. A., et al. 2018, MNRAS, 473, 5113, doi: 10.1093/mnras/stx2622

    2018 doi

  10. [18]

    J., McDowell, J

    Elvis, M., Wilkes, B. J., McDowell, J. C., et al. 1994, ApJS, 95, 1, doi: 10.1086/192093

    1994 doi

  11. [19]

    N., Evans, J

    Evans, I. N., Evans, J. D., Mart ´ ınez-Galarza, J. R., et al. 2024, ApJS, 274, 22, doi: 10.3847/1538-4365/ad6319 Faucher-Gigu` ere, C.-A. 2020, MNRAS, 493, 1614, doi: 10.1093/mnras/staa302

    2024 doi

  12. [20]

    Fitzpatrick, E. L. 1999, PASP, 111, 63, doi: 10.1086/316293

    1999 doi

  13. [21]

    E., et al

    Fukugita, M., Ichikawa, T., Gunn, J. E., et al. 1996, AJ, 111, 1748, doi: 10.1086/117915

    1996 doi

  14. [22]

    M., Turner, T

    George, I. M., Turner, T. J., Yaqoob, T., et al. 2000, ApJ, 531, 52, doi: 10.1086/308461

    2000 doi

  15. [23]

    2019, ApJ, 877, 117, doi: 10.3847/1538-4357/ab1be5 HI4PI Collaboration, Ben Bekhti, N., Fl¨ oer, L., et al

    Hemmati, S., Capak, P., Masters, D., et al. 2019, ApJ, 877, 117, doi: 10.3847/1538-4357/ab1be5 HI4PI Collaboration, Ben Bekhti, N., Fl¨ oer, L., et al. 2016, A&A, 594, A116, doi: 10.1051/0004-6361/201629178

    2019 doi

  16. [24]

    W., Bovy, J., & Lang, D

    Hogg, D. W., Bovy, J., & Lang, D. 2010, arXiv e-prints, arXiv:1008.4686, doi: 10.48550/arXiv.1008.4686

    2010 doi arXiv

  17. [25]

    F., Richards, G

    Hopkins, P. F., Richards, G. T., & Hernquist, L. 2007, ApJ, 654, 731, doi: 10.1086/509629

    2007 doi

  18. [26]

    H., & Agol, E

    Hubeny, I., Blaes, O., Krolik, J. H., & Agol, E. 2001, ApJ, 559, 680, doi: 10.1086/322344

    2001 doi

  19. [27]

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

    2007 doi

  20. [28]

    Jalan, P., Rakshit, S., Woo, J.-J., Kotilainen, J., & Stalin, C. S. 2023, MNRAS, 521, L11, doi: 10.1093/mnrasl/slad014

    2023 doi

  21. [30]

    2023, ApJ, 954, 156, doi: 10.3847/1538-4357/aceb5e

    Kim, D., Im, M., Kim, M., et al. 2023, ApJ, 954, 156, doi: 10.3847/1538-4357/aceb5e

    2023 doi

  22. [31]

    1982, Biological Cybernetics, 43, 59, doi: 10.1007/BF00337288

    Kohonen, T. 1982, Biological Cybernetics, 43, 59, doi: 10.1007/BF00337288

    1982 doi

  23. [32]

    M., Richards, G

    Krawczyk, C. M., Richards, G. T., Mehta, S. S., et al. 2013, ApJS, 206, 4, doi: 10.1088/0067-0049/206/1/4 La Torre, V., & Pacucci, F. 2024, arXiv e-prints, arXiv:2410.11951, doi: 10.48550/arXiv.2410.11951 La Torre, V., Sajina, A., Goulding, A. D., et al. 2024, AJ, 167, 261, do...

    2013 doi arXiv

  24. [35]

    W., Higley, A

    Lyke, B. W., Higley, A. N., McLane, J. N., et al. 2020, ApJS, 250, 8, doi: 10.3847/1538-4365/aba623

    2020 doi

  25. [36]

    1971, ApJL, 164, L73, doi: 10.1086/180695

    Lynds, R. 1971, ApJL, 164, L73, doi: 10.1086/180695

    1971 doi

  26. [37]

    Lyu, J., & Rieke, G. H. 2017, ApJ, 841, 76, doi: 10.3847/1538-4357/aa7051

    2017 doi

  27. [38]

    2004, MNRAS, 351, 169, doi: 10.1111/j.1365-2966.2004.07765.x

    Marconi, A., Risaliti, G., Gilli, R., et al. 2004, MNRAS, 351, 169, doi: 10.1111/j.1365-2966.2004.07765.x

    2004

  28. [39]

    C., Fanson, J., Schiminovich, D., et al

    Martin, D. C., Fanson, J., Schiminovich, D., et al. 2005, ApJL, 619, L1, doi: 10.1086/426387

    2005 doi

  29. [40]

    G., Banerji, M., Gonzalez, E., et al

    McMahon, R. G., Banerji, M., Gonzalez, E., et al. 2013, The Messenger, 154, 35

    2013

  30. [41]

    2024, A&A, 682, A34, doi: 10.1051/0004-6361/202347165

    Merloni, A., Lamer, G., Liu, T., et al. 2024, A&A, 682, A34, doi: 10.1051/0004-6361/202347165

    2024 doi

  31. [42]

    2014, SOMPY: A Python Library for Self Organizing Map (SOM)

    Moosavi, V., Packmann, S., & Vall´ es, I. 2014, SOMPY: A Python Library for Self Organizing Map (SOM)

    2014

  32. [43]

    A., et al

    Morrissey, P., Conrow, T., Barlow, T. A., et al. 2007, ApJS, 173, 682, doi: 10.1086/520512

    2007 doi

  33. [44]

    S., & Brotherton, M

    Nemmen, R. S., & Brotherton, M. S. 2010, MNRAS, 408, 1598, doi: 10.1111/j.1365-2966.2010.17224.x 18 Chen et al

    2010

  34. [45]

    D., & Thorne, K

    Novikov, I. D., & Thorne, K. S. 1973, in Black Holes (Les Astres Occlus), 343–450

    1973

  35. [46]

    2011, Journal of Machine Learning Research, 12, 2825, doi: 10.48550/arXiv.1201.0490

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825, doi: 10.48550/arXiv.1201.0490

    2011 doi arXiv

  36. [47]

    Perez, F., & Granger, B. E. 2007, Computing in Science and Engineering, 9, 21, doi: 10.1109/MCSE.2007.53

    2007 doi

  37. [48]

    X., Worseck, G., & O’Meara, J

    Prochaska, J. X., Worseck, G., & O’Meara, J. M. 2009, ApJL, 705, L113, doi: 10.1088/0004-637X/705/2/L113

    2009 doi

  38. [49]

    2015, A&A, 577, A121, doi: 10.1051/0004-6361/201425246

    Ranalli, P., Georgantopoulos, I., Corral, A., et al. 2015, A&A, 577, A121, doi: 10.1051/0004-6361/201425246

    2015 doi

  39. [50]

    2024, arXiv e-prints, arXiv:2406.17598, doi: 10.48550/arXiv.2406.17598

    Ren, W., Guo, H., Shen, Y., et al. 2024, arXiv e-prints, arXiv:2406.17598, doi: 10.48550/arXiv.2406.17598

    2024 doi arXiv

  40. [51]

    T., Lacy, M., Storrie-Lombardi, L

    Richards, G. T., Lacy, M., Storrie-Lombardi, L. J., et al. 2006, ApJS, 166, 470, doi: 10.1086/506525

    2006 doi

  41. [52]

    J., Trakhtenbrot, B., Lutz, D., et al

    Rosario, D. J., Trakhtenbrot, B., Lutz, D., et al. 2013, A&A, 560, A72, doi: 10.1051/0004-6361/201322196

    2013 doi

  42. [53]

    C., Brotherton, M

    Runnoe, J. C., Brotherton, M. S., & Shang, Z. 2012a, MNRAS, 422, 478, doi: 10.1111/j.1365-2966.2012.20620.x —. 2012b, MNRAS, 426, 2677, doi: 10.1111/j.1365-2966.2012.21644.x

    2012

  43. [54]

    F., & Finkbeiner, D

    Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103, doi: 10.1088/0004-637X/737/2/103

    2011 doi

  44. [55]

    F., Meisner, A

    Schlafly, E. F., Meisner, A. M., & Green, G. M. 2019, ApJS, 240, 30, doi: 10.3847/1538-4365/aafbea

    2019 doi

  45. [56]

    J., Finkbeiner, D

    Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525, doi: 10.1086/305772

    1998 doi

  46. [57]

    Schmidt, M., & Green, R. F. 1983, ApJ, 269, 352, doi: 10.1086/161048

    1983 doi

  47. [58]

    I., & Sunyaev, R

    Shakura, N. I., & Sunyaev, R. A. 1976, MNRAS, 175, 613, doi: 10.1093/mnras/175.3.613

    1976 doi

  48. [59]

    S., Wills, B

    Shang, Z., Brotherton, M. S., Wills, B. J., et al. 2011, ApJS, 196, 2, doi: 10.1088/0067-0049/196/1/2

    2011 doi

  49. [60]

    2020, ApJ, 899, 112, doi: 10.3847/1538-4357/aba8a1

    Treister, E. 2020, ApJ, 899, 112, doi: 10.3847/1538-4357/aba8a1

    2020 doi

  50. [61]

    T., Strauss, M

    Shen, Y., Richards, G. T., Strauss, M. A., et al. 2011, ApJS, 194, 45, doi: 10.1088/0067-0049/194/2/45

    2011 doi

  51. [62]

    A., Siemiginowska, A., & ˙Zycki, P

    Sobolewska, M. A., Siemiginowska, A., & ˙Zycki, P. T. 2004a, ApJ, 608, 80, doi: 10.1086/392529 —. 2004b, ApJ, 617, 102, doi: 10.1086/425262

    doi

  52. [63]

    2016, MNRAS, 458, 2288, doi: 10.1093/mnras/stw444

    Stalevski, M., Ricci, C., Ueda, Y., et al. 2016, MNRAS, 458, 2288, doi: 10.1093/mnras/stw444

    2016 doi

  53. [64]

    T., Strateva, I., Brandt, W

    Steffen, A. T., Strateva, I., Brandt, W. N., et al. 2006, AJ, 131, 2826, doi: 10.1086/503627

    2006 doi

  54. [65]

    2025, arXiv e-prints, arXiv:2501.10793, doi: 10.48550/arXiv.2501.10793

    Su, T., Guo, Q., Qiao, E., et al. 2025, arXiv e-prints, arXiv:2501.10793, doi: 10.48550/arXiv.2501.10793

    2025 doi

  55. [66]

    B., Vanden Berk, D

    Trammell, G. B., Vanden Berk, D. E., Schneider, D. P., et al. 2007, AJ, 133, 1780, doi: 10.1086/511817

    2007 doi

  56. [67]

    D., Lamer, G., et al

    Traulsen, I., Schwope, A. D., Lamer, G., et al. 2020, A&A, 641, A137, doi: 10.1051/0004-6361/202037706

    2020 doi

  57. [68]

    2001, Bioinformatics, 17, 520, doi: 10.1093/bioinformatics/17.6.520

    Troyanskaya, O., Cantor, M., Sherlock, G., et al. 2001, Bioinformatics, 17, 520, doi: 10.1093/bioinformatics/17.6.520

    2001 doi

  58. [69]

    2003, ApJ, 598, 886, doi: 10.1086/378940 van der Walt, S., Colbert, S

    Ueda, Y., Akiyama, M., Ohta, K., & Miyaji, T. 2003, ApJ, 598, 886, doi: 10.1086/378940 van der Walt, S., Colbert, S. C., & Varoquaux, G. 2011, Computing in Science and Engineering, 13, 22, doi: 10.1109/MCSE.2011.37 Vanden Berk, D. E., Richards, G. T., Bauer, A., et al. 2001, A...

    2003 doi

  59. [70]

    J., Ivezi, Z., & Gray, A

    VanderPlas, J., Connolly, A. J., Ivezi, Z., & Gray, A. 2012, in 2012 Conference on Intelligent Data Understanding, 47–54, doi: 10.1109/CIDU.2012.6382200

    2012

  60. [71]

    V., & Fabian, A

    Vasudevan, R. V., & Fabian, A. C. 2007, MNRAS, 381, 1235, doi: 10.1111/j.1365-2966.2007.12328.x

    2007

  61. [72]

    2003, ApJ, 582, 559, doi: 10.1086/344675

    Volonteri, M., Haardt, F., & Madau, P. 2003, ApJ, 582, 559, doi: 10.1086/344675

    2003 doi

  62. [73]

    E., Kilic, M., Bergeron, P., & Leiphart, N

    Wall, R. E., Kilic, M., Bergeron, P., & Leiphart, N. D. 2023, MNRAS, 523, 4067, doi: 10.1093/mnras/stad1699

    2023 doi

  63. [74]

    E., Kilic, M., Bergeron, P., et al

    Wall, R. E., Kilic, M., Bergeron, P., et al. 2019, MNRAS, 489, 5046, doi: 10.1093/mnras/stz2506

    2019 doi

  64. [75]

    A., Coriat, M., Traulsen, I., et al

    Webb, N. A., Coriat, M., Traulsen, I., et al. 2020, A&A, 641, A136, doi: 10.1051/0004-6361/201937353

    2020 doi

  65. [76]

    L., Eisenhardt, P

    Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010, AJ, 140, 1868, doi: 10.1088/0004-6256/140/6/1868

    2010 doi

  66. [77]

    2022, ApJS, 263, 42, doi: 10.3847/1538-4365/ac9ead

    Wu, Q., & Shen, Y. 2022, ApJS, 263, 42, doi: 10.3847/1538-4365/ac9ead

    2022 doi

  67. [78]

    G., Adelman, J., Anderson, Jr., J

    York, D. G., Adelman, J., Anderson, Jr., J. E., et al. 2000, AJ, 120, 1579, doi: 10.1086/301513

    2000 doi

  68. [79]

    2014, ARA&A, 52, 529, doi: 10.1146/annurev-astro-082812-141003

    Yuan, F., & Narayan, R. 2014, ARA&A, 52, 529, doi: 10.1146/annurev-astro-082812-141003

    2014 doi

  69. [80]

    B., Liu, X

    Yuan, H. B., Liu, X. W., & Xiang, M. S. 2013, MNRAS, 430, 2188, doi: 10.1093/mnras/stt039

    2013 doi

  70. [81]

    C., & Shangguan, J

    Zhuang, M.-Y., Ho, L. C., & Shangguan, J. 2021, ApJ, 906, 38, doi: 10.3847/1538-4357/abc94d

    2021 doi

Pith tools

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