Pith. sign in

REVIEW 3 major objections 6 minor 195 references

VAR-PZnn claims that a mixture-density network combining ZTF g-band variability features with optical, mid-infrared, and optional near-infrared colors can estimate AGN photometric redshifts with an 8.2% outlier fraction, and that mid-infrar

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 · deepseek-v4-flash

2026-08-01 20:56 UTC pith:R4QN5DWZ

load-bearing objection Well-executed, honest ML paper for AGN photo-zs; headline numbers are credible in-distribution, but the LSST scalability claim rests on an acknowledged selection-function shift. the 3 major comments →

arxiv 2607.16434 v1 pith:R4QN5DWZ submitted 2026-07-17 astro-ph.GA

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

S. Satheesh-Sheeba , P. S\'anchez-S\'aez , R. J. Assef , T. Anguita , R. Shirley , M. Salvato , P. Ar\'evalo , T T. Ananna
show 34 more authors
F. E. Bauer C. G. Bornancini W.N. Brandt D. De Cicco M. Espinoza-Ortiz J. Fagin M. Fatovi\'c A. W. Graham H. Guo L. Hernandez-Garc\'ia D. Ili\'c A. B. Kova\v{c}evi\'c P. Lira A. I. Malz M. Marculewicz D. Marsango C. Mazzucchelli T. Mkrtchyan S. Panda A. Peca V. Petrecca B. Rani C. Ricci G. T. Richards R. A. Riffel A. Rojas-Lilay\'u E. Saremi D. P. Schneider B. Sotomayor M. J. Temple A. Viitanen I. Yoon Z. Yu F. Zou
This is my paper
classification astro-ph.GA
keywords photometric redshiftsactive galactic nucleiquasarsAGN variabilitymixture density networksZTF light curvesmid-infrared photometryLSST
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.

The paper sets out to show that AGN photometric redshifts, historically unreliable because AGN power-law spectra lack the sharp features that anchor galaxy photo-zs, can be estimated accurately by a neural network that folds time-domain variability into multi-wavelength colors. On 72,728 spectroscopically confirmed AGN/QSOs, it reports a normalized median absolute deviation of 0.058 and an 8.2% catastrophic-outlier fraction, compared with 28.7% for template fitting on the same sources. Ablation tests show mid-infrared photometry is the anchor: dropping it raises outliers to roughly 35%, while adding variability on top of optical+MIR only improves the outlier fraction from 9.3% to 8.2%. If true, this offers a scalable path for the millions of AGN that LSST will find without spectra, and it quantifies how much weight to give each data type.

Core claim

The central claim is that a fully connected mixture density network—a regressor that outputs a full probability distribution over redshift—can combine 26 variability descriptors from single-band ZTF g-band light curves with Pan-STARRS optical, CatWISE mid-infrared, and optional UKIDSS near-infrared colors to reach σ_NMAD=0.058 with η=8.2% on a test set of about 14.5 thousand AGN. The paper further claims the ranking of constraints is clear: the W1−W2 mid-infrared color and the PS1 morphology flag dominate, while damped-random-walk timescale and amplitude features act as secondary refiners. Removing MIR raises the outlier fraction to 35.4% (optical+variability) or 40.5% (optical only), wherea

What carries the argument

The engine is a Mixture Density Network (MDN): a neural network whose final layer outputs the weights, means, and standard deviations of three Gaussian components, so each source gets a posterior p(z|x) instead of a single point estimate. This lets the model represent multi-modal color-redshift degeneracies. Inputs are 26 variability features (damped-random-walk timescale and amplitude, Mexican-hat power-spectrum amplitudes, structure-function slope, autocorrelation, skewness and kurtosis measures) plus optical, MIR, NIR colors and a PS1 morphology flag; variability features are quantile-mapped to Gaussians before training. Uncertainty comes from Monte-Carlo dropout with 50 stochastic passes

Load-bearing premise

The 8.2% outlier fraction is measured on 72,728 AGNs selected as variable in ZTF g-band with SDSS/DESI spectroscopy (g roughly 17–21.5), and everything rests on that training distribution representing the fainter, more obscured AGN population LSST will find without spectra—a transfer the paper itself, in Section 5.4, calls optimistic.

What would settle it

Apply the released trained model to a spectroscopically confirmed sample of faint (g≳22), type-2, or high-redshift AGNs with similar single-band light-curve coverage and compare predicted with spectroscopic redshift; if the catastrophic-outlier fraction on that out-of-distribution sample is far above 8.2% (for instance, ≳20%), the claimed transferability to LSST fails. A simpler check: retrain the model with the morphology flag removed and look for a large change in the redshift-binned outlier fraction.

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

If this is right

  • For ZTF-like time-domain surveys, AGN photo-z catalogs can be produced at η=8.2% (σ_NMAD=0.058) without spectroscopy, roughly 3.5 times fewer outliers than LRT template fitting on the same sources.
  • Mid-infrared photometry is the primary driver: removing CatWISE W1/W2 from the model raises the outlier fraction from 8.2% to about 35%, so surveys without MIR coverage should expect much larger outlier fractions.
  • Near-infrared photometry can partially substitute for MIR: PS1+UKIDSS+variability yields η=13.3%, improving to 4.6% when WISE is added, informing planning for LSST+Euclid/Roman synergies.
  • The network's uncertainty estimates are usable as a quality filter: dropping the 10% most uncertain sources lowers η from 8.2% to 5.4%.
  • Single-band variability priors should not be bolted onto SED fitting: applying VAR-PZ priors from one g-band light curve degrades LRT outliers from 28.7% to 39.4% on real data, whereas learned variability features improve the network, indicating that the way variability enters matters as much as its presence.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the reported feature ranking transfers, LSST's gains from variability may be modest for bright, type-1 AGN already well covered by optical+MIR colors; variability would earn its place mainly for sources where WISE is undetected or host-galaxy contamination is severe, an untested corollary of the UKIDSS result.
  • The high ranking of the PS1 morphology flag may partly encode the spectroscopic selection function (brighter, more point-like AGNs preferentially observed at higher redshift), so retraining without ps_score and comparing binned outlier fractions would reveal how much of the claimed accuracy is selection rather than physics; the paper itself flags this risk.
  • The failure of VAR-PZ on single-band data is not evidence against variability as a redshift tracer: LSST's multi-band light curves should break the single-band DRW degeneracy, and a direct test with simulated multi-band light curves would separate the method's potential from its single-band limitation.

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

3 major / 6 minor

Summary. This paper presents VAR-PZnn, a mixture-density network for AGN photometric redshifts that combines 26 ZTF g-band variability features with PS1 optical, CatWISE MIR, and UKIDSS NIR photometry and a PS1 morphology flag. The model is trained and tested on 72,728 spectroscopically confirmed AGNs from SDSS/DESI using a 60:20:20 split. The authors report σ_NMAD = 0.058 and η = 8.2% on the main sample, improving to 5.4% after excluding the 10% highest-uncertainty sources. Ablations show MIR photometry is the dominant constraint, with variability providing a secondary improvement (η = 9.3% without variability vs. 8.2% with it). On the UKIDSS subsample, the model reaches η = 13.3% without MIR and η = 4.6% with MIR. The paper also benchmarks against LRT template fitting (η = 28.7%) and shows that applying single-band VAR-PZ priors degrades LRT performance (η = 39.4%). The abstract and conclusions frame the method as 'a scalable approach for LSST'.

Significance. If taken at face value, the in-sample results are a solid, reproducible contribution. The paper uses a proper held-out split, standard photo-z metrics, systematic ablations, a calibration check, and a public code repository. The ranking of MIR as the dominant constraint and variability as a secondary refiner is useful for planning AGN photo-z efforts, and the cautionary result about single-band VAR-PZ priors is interesting. The main caveat is that the headline metrics are in-distribution: the test set inherits the variability-selected, spectroscopically targeted selection function. The paper itself concedes in §5.4 that the training sample is incomplete for faint/high-z AGNs and that the reported performance is an optimistic estimate for LSST. Therefore the paper's lasting value is as a framework demonstration and feature-modality comparison, not as a quantitative LSST forecast without further out-of-distribution validation.

major comments (3)
  1. [Abstract; §5.4] The abstract and summary state that VAR-PZnn 'provides a scalable approach for LSST,' but the evaluation is entirely in-distribution. The test set is a random split of a sample that is (i) pre-selected as variable by the A26 random forest, (ii) restricted to ZTF g-band detections at 17–21.5 mag, and (iii) spectroscopically targeted by SDSS/DESI. A random split preserves the selection function, so σ_NMAD=0.058 and η=8.2% measure performance on sources drawn from the same population used for training. Section 5.4 explicitly concedes that the sample is 'likely incomplete for faint or high-z AGNs' and that the reported performance 'should be regarded as an optimistic estimate.' Because the LSST claim is the stated motivation, the authors should either provide an out-of-distribution validation (e.g., a fainter or more obscured spectroscopically confirmed sample) or rewrite the abstract and co
  2. [§5.4, Fig. 6, Table 1] The feature-importance analysis ranks ps_score second (ΔRMSE=0.084), and this morphology flag is included in every ablation configuration in Table 1. The paper itself notes in §5.4 that ps_score 'may partly encode properties of the spectroscopic training sample itself,' including a Malmquist-type magnitude/morphology selection. If so, a substantial part of the reported accuracy — and the relative ranking of MIR vs. variability — could reflect the training selection rather than a universal redshift–feature relation. This risk is testable with existing data: run an ablation without ps_score, or evaluate on a magnitude/morphology-matched test subset and show that the performance and feature ranking are stable. I request this analysis, or a clear statement of why it is not possible, before the 'MIR dominant, variability refiner' conclusion is presented as robust.
  3. [§5.3, Appendix F] The interpretation of the LRT+VAR-PZ degradation (η: 28.7% → 39.4%) relies on the DRW simulation in Appendix F, but the simulated light curves are generated using the same DRW scaling relations that define VAR-PZ. The simulation therefore does not falsify the assumption that those scaling relations hold for real AGN; it only shows that idealized pure-DRW light curves with consistent uncertainties do not produce the degradation. The paper's explanation — real data contain non-DRW variability components and systematics — is plausible but not directly demonstrated. Please either add non-DRW or systematics-contaminated simulations and show that the degradation reappears, or weaken the causal claim to a hypothesis. This does not affect the ML photo-z results themselves, but it is central to the benchmark narrative in §5.3 and the abstract.
minor comments (6)
  1. [§4.3, Appendix B] There is an inconsistency in the activation function: Section 4.3 states ReLU activations, while Appendix B states LeakyReLU activations. Please clarify which was used.
  2. [§5.1, Table 1] σ_NMAD and η are reported as point estimates without uncertainties. Given the finite test size and stochastic training, a bootstrap error or multiple-seed standard deviation would make the headline numbers more interpretable, especially for the small differences between the full model and optical+MIR-only model (η = 8.2% vs. 9.3%).
  3. [§5.2] The UKIDSS subsample analysis uses a different dropout rate (0.35 vs. 0.2 for the main sample). This should be stated in the main text or table so the reader knows the architecture was not identical across all configurations.
  4. [Fig. 5] The bin at z > 3.5 has only 51 sources; this is mentioned in the text but it would be helpful to display the bin counts directly on the figure to avoid overinterpretation of the apparent improvement at the high-redshift end.
  5. [§2.4, §5.2] The UKIDSS catalog columns appear in Appendix D with nonstandard names such as 'hapermag3', 'yapermag3', 'kapermag3', 'j_1apermag3'. Please define or rename these to match the photometric bands used in the text.
  6. [§4.6] The permutation importance definition uses RMSE without (1+z) normalization. The paper justifies this for ranking, but because RMSE is dominated by high-redshift outliers, it may understate features that matter most at low z. A sentence noting this possible dependence would be appropriate.

Circularity Check

0 steps flagged

No significant circularity: the model is trained on spectroscopic labels and evaluated on a held-out split; the central claims are empirical and externally benchmarked.

full rationale

The central derivation is a supervised ML pipeline: the network is trained with spec-z labels, and the headline metrics (sigma_NMAD=0.058, eta=8.2%) are computed on a held-out test set after a random 60/20/20 split. No target quantity is used to fit the model, so the reported performance is not circular by construction. The ablation ranking (MIR dominant, variability secondary) is obtained by retraining on feature subsets and measuring held-out RMSE; it is an empirical comparison, not a definitional identity. The LRT benchmark (eta=28.7%) is an external template-fitting method (Assef et al. 2010), not a self-citation, and the comparison is performed on the same test objects. The only clearly self-referential elements are (i) the use of the authors' A26 variability-feature catalog as input features, and (ii) the Appendix F DRW simulation, which applies the authors' own VAR-PZ scaling relations to interpret why single-band VAR-PZ priors degrade LRT performance. Neither feeds the main photo-z prediction: A26 features are published independently and serve as inputs, while the simulation is a post-hoc diagnostic that does not produce the claimed photo-z results. The paper's own §5.4 explicitly concedes that the training sample is incomplete for faint/high-z AGNs and that LSST performance 'should be regarded as an optimistic estimate'; this is an honest limitation statement, not evidence of circularity. Overall, the derivation chain is self-contained against the spectroscopic labels and external benchmarks, and no fitted parameter is renamed as a prediction.

Axiom & Free-Parameter Ledger

4 free parameters · 6 axioms · 0 invented entities

No new physical entities are postulated. The only new object is the software framework VAR-PZnn, which is a method, not an invented physical entity. The free parameters are standard ML hyperparameters, chosen by hand or via validation, and the axioms are either standard statistical/math results or domain assumptions about the data, all explicitly or implicitly stated.

free parameters (4)
  • MDN mixture components K = 3
    Number of Gaussian components; K∈{1..5} tested in Appendix C, performance robust (η varies by ≤0.5 ppt); K=3 adopted as a compromise.
  • Dropout rate = 0.2 (main), 0.35 (UKIDSS+NIR config)
    Regularization strength chosen by hand; increased for the larger NIR feature set to mitigate overfitting (§4.3, §5.2).
  • Learning rate = 1e-4
    Adam optimizer hyperparameter chosen by hand (§4.3).
  • L2 weight decay = 1e-4
    Regularization hyperparameter (§4.3).
axioms (6)
  • standard math Mixture density networks and MC dropout provide valid predictive posteriors and uncertainty estimates
    Invoked in §4.3-4.4 after Bishop (1994), Gal & Ghahramani (2016), Kendall & Gal (2017).
  • domain assumption Damped random walk (DRW) model adequately describes AGN optical variability on ZTF timescales
    Used for feature extraction (GP_DRW_tau, GP_DRW_sigma) and in the Appendix F simulation validating the VAR-PZ degradation interpretation.
  • domain assumption A26 random forest variability classifier correctly separates AGN/QSO from other variable sources
    The parent sample is pre-selected on this classification (§2.1); classification errors propagate into the training set.
  • domain assumption Spectroscopic redshifts from SDSS DR16/DR19 and DESI DR1 are accurate labels
    Used as ground truth throughout (§2.3).
  • domain assumption The 60:20:20 random split yields a test set representative of the training population
    Validation and test redshift distributions track the training distribution (§3, Fig. 2); this does not cover out-of-distribution LSST sources, as conceded in §5.4.
  • domain assumption LRT templates and VAR-PZ priors are correctly applied as benchmarks
    Benchmark comparison in §5.3; LRT from Assef et al. 2010, VAR-PZ from Satheesh-Sheeba et al. 2026.

pith-pipeline@v1.3.0-alltime-deepseek · 24919 in / 12437 out tokens · 103565 ms · 2026-08-01T20:56:54.689150+00:00 · methodology

0 comments
read the original abstract

Photometric redshift estimation for active galactic nuclei (AGNs) remains a fundamental challenge for current and upcoming large-scale photometric surveys. Traditional spectral energy distribution (SED) fitting suffers from color-redshift degeneracies, particularly for AGNs whose power-law continua hide the strong spectral features required to anchor redshift estimates. While AGN variability provides additional constraining power, existing frameworks require multi-band light curves that are not always available. This work presents VAR-PZnn, a fully connected mixture density network that integrates 26 variability features extracted from ZTF g-band light curves with optical photometry from Pan-STARRS1, mid-infrared (MIR) photometry from CatWISE, and, for a subsample, NIR photometry from UKIDSS. The model is trained and tested on 72,728 spectroscopically confirmed AGNs/QSOs spanning 0.01 < z < 4.5 and g-band magnitudes from 17 to 21.5. For the main sample, we achieve \sigma_{NMAD} = 0.058 and an outlier fraction of \eta = 8.2%, which reduces to 5.4% when the 10% of sources with the highest predicted uncertainty are excluded. An ablation study demonstrates that MIR photometry provides the dominant constraint for photo-z accuracy, while variability features serve as a secondary refiner. Using UKIDSS NIR data as a proxy for future synergies between LSST and space-based missions like Euclid and Roman, we obtain \eta = 13.3% without MIR data and \eta = 4.6% when MIR is available. We benchmark against Low-Resolution Templates (LRT) SED fitting (\eta = 28.7%) and the VAR-PZ framework; applying single-band VAR-PZ priors worsens LRT performance to \eta = 39.4% due to single-band light-curve degeneracies, confirmed via simulations (\eta = 27.6% to 28.1%). This framework provides a scalable approach for the Legacy Survey of Space and Time (LSST).

Figures

Figures reproduced from arXiv: 2607.16434 by A. B. Kova\v{c}evi\'c, A. I. Malz, A. Peca, A. Rojas-Lilay\'u, A. Viitanen, A. W. Graham, B. Rani, B. Sotomayor, C. G. Bornancini, C. Mazzucchelli, C. Ricci, D. De Cicco, D. Ili\'c, D. Marsango, D. P. Schneider, E. Saremi, F. E. Bauer, F. Zou, G. T. Richards, H. Guo, I. Yoon, J. Fagin, L. Hernandez-Garc\'ia, M. Espinoza-Ortiz, M. Fatovi\'c, M. J. Temple, M. Marculewicz, M. Salvato, P. Ar\'Evalo, P. Lira, P. S\'anchez-S\'aez, R. A. Riffel, R. J. Assef, R. Shirley, S. Panda, S. Satheesh-Sheeba, T. Anguita, T. Mkrtchyan, T T. Ananna, V. Petrecca, W.N. Brandt, Z. Yu.

Figure 1
Figure 1. Figure 1: Distribution of the samples used in our analysis as a [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Spectroscopic redshift distributions of the training, vali [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Evolution of the loss function with training epoch for the [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Photo-z performance as a function of redshift bin. The solid line with circles shows the σNMAD, while the dashed line with squares indicates the η in percent. these predicted uncertainties are slightly conservative, system￾atically over-covering the true scatter (mean calibration error 0.093) [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Relative feature importance for the main sample model, [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Performance of our model across four distinct feature configurations: optical colors only (top-left), variability features [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Performance of our photo-z estimator model using AGN variability features combined with PS1+UKIDSS photometry (top row) and PS1+UKIDSS+WISE photometry (bottom row). Left panels: photometric versus spectroscopic redshifts for the test sample, with the color scale indicating the number of sources per bin. The red dashed line represents the one-to-one relation, and the shaded grey region indicates the median … view at source ↗
Figure 10
Figure 10. Figure 10: Results for the VAR-PZ framework, illustrating the perfor￾mance when variability-based constraints from ZTF light curves are combined with LRT SED modeling predictions. The scatter plot compares the resulting photo-z estimates against spectro￾scopic values for the test sample, with the one-to-one relation indicated by the red dashed line. The inset provides the corre￾sponding statistical metrics, includin… view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

195 extracted references · 9 canonical work pages · 2 internal anchors

  1. [1]

    , keywords =

    The Ensemble Photometric Variability of -0.5ex 25,000 Quasars in the Sloan Digital Sky Survey. , keywords =. doi:10.1086/380563 , archivePrefix =. astro-ph/0310336 , primaryClass =

    Pith/arXiv arXiv

  2. [2]

    , keywords =

    Quasar Variability and Gravitational Microlensing. , keywords =. doi:10.1086/310689 , adsurl =

    doi:10.1086/310689

  3. [3]

    doi:10.1086/324541 , url =

    Dario Trevese and Fausto Vagnetti , title =. doi:10.1086/324541 , url =

    doi:10.1086/324541

  4. [4]

    , keywords =

    Flexible and Scalable Methods for Quantifying Stochastic Variability in the Era of Massive Time-domain Astronomical Data Sets. , keywords =. doi:10.1088/0004-637X/788/1/33 , archivePrefix =. 1402.5978 , primaryClass =

    Pith/arXiv arXiv

  5. [5]

    Multiwavelength Monitoring of the Dwarf Seyfert 1 Galaxy NGC 4395. I. A Reverberation-based Measurement of the Black Hole Mass. , keywords =. doi:10.1086/444494 , archivePrefix =. astro-ph/0506665 , primaryClass =

    Pith/arXiv arXiv

  6. [6]

    Pereyra and Daniel E

    Nicolas A. Pereyra and Daniel E. Vanden Berk and David A. Turnshek and D. John Hillier and Brian C. Wilhite and Richard G. Kron and Donald P. Schneider and Jonathan Brinkmann , title =. doi:10.1086/500919 , url =

    doi:10.1086/500919

  7. [7]

    , keywords =

    An accretion disc model for quasar optical variability. , keywords =. doi:10.1111/j.1745-3933.2008.00480.x , archivePrefix =. 0805.0351 , primaryClass =

    arXiv 2008

  8. [8]

    , keywords =

    The Effect of a Time-varying Accretion Disk Size on Quasar Microlensing Light Curves. , keywords =. doi:10.1088/0004-637X/718/2/1079 , archivePrefix =. 1002.3126 , primaryClass =

    Pith/arXiv arXiv

  9. [9]

    , keywords =

    VAR-PZ: Constraining the photometric redshifts of quasars using variability. , keywords =. doi:10.1051/0004-6361/202557223 , archivePrefix =. 2509.13308 , primaryClass =

    Pith/arXiv arXiv

  10. [10]

    , keywords =

    Are the Variations in Quasar Optical Flux Driven by Thermal Fluctuations?. , keywords =. doi:10.1088/0004-637X/698/1/895 , archivePrefix =. 0903.5315 , primaryClass =

    Pith/arXiv arXiv

  11. [11]

    Peterson , title =

    Stefan Collier and Bradley M. Peterson , title =. , abstract =. 2001 , month =. doi:10.1086/321517 , url =

    doi:10.1086/321517 2001

  12. [12]

    , title = "

    Giveon, Uriel and Maoz, Dan and Kaspi, Shai and Netzer, Hagai and Smith, Paul S. , title = ". , volume =. 1999 , month =. doi:10.1046/j.1365-8711.1999.02556.x , url =

    arXiv 1999

  13. [13]

    , keywords =

    Selecting Quasars by Their Intrinsic Variability. , keywords =. doi:10.1088/0004-637X/714/2/1194 , archivePrefix =. 1002.2642 , primaryClass =

    Pith/arXiv arXiv

  14. [14]

    The Interplay Among Black Holes, Stars and ISM in Galactic Nuclei , year = 2004, editor =

    Quasar variability measurements with SDSS repeated imaging and POSS data. The Interplay Among Black Holes, Stars and ISM in Galactic Nuclei , year = 2004, editor =. doi:10.1017/S1743921304003126 , archivePrefix =. astro-ph/0404487 , primaryClass =

    Pith/arXiv arXiv 2004

  15. [15]

    , keywords =

    Structure Function Analysis of Long-Term Quasar Variability. , keywords =. doi:10.1086/427393 , archivePrefix =. astro-ph/0411348 , primaryClass =

    Pith/arXiv arXiv

  16. [17]

    , keywords =

    Variable Faint Optical Sources Discovered by Comparing the POSS and SDSS Catalogs. , keywords =. doi:10.1086/503672 , archivePrefix =. astro-ph/0403319 , primaryClass =

    Pith/arXiv arXiv

  17. [18]

    QUASAR OPTICAL VARIABILITY IN THE PALOMAR-QUEST SURVEY , volume=

    Bauer, Anne and Baltay, Charles and Coppi, Paolo and Ellman, Nancy and Jerke, Jonathan and Rabinowitz, David and Scalzo, Richard , year=. QUASAR OPTICAL VARIABILITY IN THE PALOMAR-QUEST SURVEY , volume=. , publisher=. doi:10.1088/0004-637x/696/2/1241 , number=

    doi:10.1088/0004-637x/696/2/1241

  18. [19]

    C. L. MacLeod and. doi:10.1088/0004-637x/721/2/1014 , url =

    doi:10.1088/0004-637x/721/2/1014

  19. [20]

    , keywords =

    Image-tube photography of a complete sample of 4C radio sources. , keywords =. doi:10.1086/111969 , adsurl =

    work page doi:10.1086/111969

  20. [21]

    , keywords =

    Variability and the nature of QSO optical-infrared continua. , keywords =. doi:10.1086/163461 , adsurl =

    work page doi:10.1086/163461

  21. [22]

    , keywords =

    An Ultraviolet Atlas of Quasar and Blazar Spectra. , keywords =. doi:10.1086/191546 , adsurl =

    work page doi:10.1086/191546

  22. [23]

    , keywords =

    Long-Term Variability of a Complete Sample of Quasars. , keywords =. doi:10.1086/115489 , adsurl =

    work page doi:10.1086/115489

  23. [24]

    , keywords =

    Optical Variability of Quasars: Statistics and Cosmological Properties. , keywords =. doi:10.1086/170365 , adsurl =

    work page doi:10.1086/170365

  24. [25]

    , year = 1994, month = may, volume =

    The variability of optically selected quasars. , year = 1994, month = may, volume =. doi:10.1093/mnras/268.2.305 , adsurl =

    doi:10.1093/mnras/268.2.305 1994

  25. [26]

    The Optical Variability of QSO's

    The optical variability of QSOs. , keywords =. doi:10.48550/arXiv.astro-ph/9506140 , archivePrefix =. astro-ph/9506140 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.astro-ph/9506140

  26. [27]

    , keywords =

    The QSO variability-luminosity-redshift relation. , keywords =. doi:10.1093/mnras/282.4.1191 , archivePrefix =. astro-ph/9608057 , primaryClass =

    Pith/arXiv arXiv

  27. [28]

    doi:10.5281/zenodo.6878414 , url =

    Weixiang Yu and Gordon Richards and Veronique Buat and William Nielsen Brandt and Manda Banerji and Qingling Ni and Raphael Shirley and Matthew Temple and Feige Wang and Jinyi Yang , title =. doi:10.5281/zenodo.6878414 , url =

    work page doi:10.5281/zenodo.6878414

  28. [30]

    Fabian , title =

    A.C. Fabian , title =. doi:10.1146/annurev-astro-081811-125521 , url =

    doi:10.1146/annurev-astro-081811-125521

  29. [31]

    American Astronomical Society Meeting Abstracts \#221 , year = 2013, series =

    AGN Science with the LSST. American Astronomical Society Meeting Abstracts \#221 , year = 2013, series =

    2013

  30. [32]

    LSST Science Book, Version 2.0 , year =

  31. [33]

    , keywords =

    Low-Resolution Spectral Templates for Active Galactic Nuclei and Galaxies from 0.03 to 30 m. , keywords =. doi:10.1088/0004-637X/713/2/970 , archivePrefix =. 0909.3849 , primaryClass =

    Pith/arXiv arXiv

  32. [34]

    Kass and Adrian E

    Robert E. Kass and Adrian E. Raftery , journal =. Bayes Factors , urldate =

  33. [35]

    , keywords =

    An Alternative Approach to Measuring Reverberation Lags in Active Galactic Nuclei. , keywords =. doi:10.1088/0004-637X/735/2/80 , archivePrefix =. 1008.0641 , primaryClass =

    Pith/arXiv arXiv

  34. [36]

    , keywords =

    Application of Stochastic Modeling to Analysis of Photometric Reverberation Mapping Data. , keywords =. doi:10.3847/0004-637X/819/2/122 , archivePrefix =. 1310.6774 , primaryClass =

    Pith/arXiv arXiv

  35. [37]

    , keywords =

    Quasar Accretion Disk Sizes from Continuum Reverberation Mapping from the Dark Energy Survey. , keywords =. doi:10.3847/1538-4357/aac9bb , archivePrefix =. 1711.11588 , primaryClass =

    Pith/arXiv arXiv

  36. [38]

    , year = 1963, month = jul, volume =

    Optical Identification of 3C 48, 3C 196, and 3C 286 with Stellar Objects. , year = 1963, month = jul, volume =. doi:10.1086/147615 , adsurl =

    doi:10.1086/147615 1963

  37. [39]

    A Survey of z>5.7 Quasars in the Sloan Digital Sky Survey. II. Discovery of Three Additional Quasars at z>6. , keywords =. doi:10.1086/368246 , archivePrefix =. astro-ph/0301135 , primaryClass =

    Pith/arXiv arXiv

  38. [40]

    Richards, G. T. and Lacy, M. and Storrie-Lombardi, L. J. and others , title =. , volume =. 2006 , doi =

    2006

  39. [41]

    J. N. Reeves and G. Wynn and P. T. O. Extreme X-ray variability in the luminous quasar. doi:10.1046/j.1365-8711.2002.06038.x , url =

    arXiv 2002

  40. [42]

    2006 , eprint=

    Optical and X-ray Variability of AGNs , author=. 2006 , eprint=

    2006

  41. [43]

    York, D. G. and Adelman, J. and Anderson, J. E. and others , title =. , volume =. 2000 , doi =

    2000

  42. [44]

    and Soares-Santos, M

    Annis, J. and Soares-Santos, M. and Strauss, M. A. and others , title =. , volume =. 2014 , doi =

    2014

  43. [45]

    Lyke, B. W. and Higley, A. N. and McLane, J. N. and others , title =. , volume =. 2020 , doi =

    2020

  44. [46]

    and Strauss, Michael A

    Shen, Yue and Richards, Gordon T. and Strauss, Michael A. and others , title =. , volume =. 2011 , doi =

    2011

  45. [47]

    Rankine, A. L. and Hewett, P. C. and Banerji, M. and Richards, G. T. , title =. , volume =. 2020 , doi =

    2020

  46. [48]

    Variable Point Sources in Sloan Digital Sky Survey Stripe 82. I. Project Description and Initial Catalog (0 hr <= <= 4 hr). , keywords =. doi:10.1088/0067-0049/186/2/233 , archivePrefix =. 0912.0976 , primaryClass =

    Pith/arXiv arXiv

  47. [49]

    Schneider, D. P. and Hall, P. B. and Richards, G. T. and others , title =. , volume =. 2007 , doi =

    2007

  48. [50]

    Abazajian, K. N. and Adelman-McCarthy, J. K. and Ag. The Seventh Data Release of the Sloan Digital Sky Survey , journal =. 2009 , doi =

    2009

  49. [51]

    and Richards, Gordon T

    Schneider, Donald P. and Richards, Gordon T. and Hall, Patrick B. and Strauss, Michael A. and Anderson, Scott F. and Boroson, Todd A. and Ross, Nicholas P. and Shen, Yue and Brandt, W. N. and Fan, Xiaohui and Inada, Naohisa and Jester, Sebastian and Knapp, G. R. and Krawczyk, Coleman M. and Thakar, Anirudda R. and Berk, Daniel E. Vanden and Voges, Wolfgan...

  50. [52]

    , keywords =

    A Modified Magnitude System that Produces Well-Behaved Magnitudes, Colors, and Errors Even for Low Signal-to-Noise Ratio Measurements. , keywords =. doi:10.1086/301004 , archivePrefix =. astro-ph/9903081 , primaryClass =

    Pith/arXiv arXiv

  51. [53]

    Lupton, R. H. and Gunn, J. E. and Szalay, A. S. , title =. , volume =. 1999 , doi =

    1999

  52. [54]

    Burke and Xin Liu and Yue Shen and Qian Yang and Charles Gammie , title =

    Colin J. Burke and Xin Liu and Yue Shen and Qian Yang and Charles Gammie , title =. Science , volume =. 2021 , doi =

    2021

  53. [55]

    Hogg and Timothy D

    Daniel Foreman-Mackey and David W. Hogg and Timothy D. Morton , title =. , volume =. 2017 , doi =

    2017

  54. [56]

    and Ichikawa, T

    Fukugita, M. and Ichikawa, T. and Gunn, J. E. and Doi, M. and Shimasaku, K. and Schneider, D. P. , title =. , volume =. 1996 , doi =

    1996

  55. [57]

    Gunn, J. E. and Carr, M. and Rockosi, C. and Sekiguchi, M. and et al. , title =. , volume =. 1998 , doi =

    1998

  56. [58]

    Smith, J. A. and Tucker, D. L. and Kent, S. and et al. , title =. , volume =. 2002 , doi =

    2002

  57. [59]

    Lupton, R. H. and Gunn, J. E. and Ivezic, Z. and Knapp, G. R. and Kent, S. , title =. Astronomical Data Analysis Software and Systems X , volume =

  58. [60]

    Gunn, J. E. and Siegmund, W. and Mannery, E. and et al. , title =. , volume =. 2006 , doi =

    2006

  59. [61]

    and Lupton, R

    Ivezic, Z. and Lupton, R. H. and Juric, M. and et al. , title =. Astronomische Nachrichten , volume =. 2004 , doi =

    2004

  60. [62]

    and Sesar, B

    Ivezic, Z. and Sesar, B. and Juric, M. and et al. , title =. , volume =. 2007 , doi =

    2007

  61. [63]

    MacLeod, C. L. and Ivezić, Ž. and Sesar, B. and et al. , title =. , volume =. 2012 , doi =

    2012

  62. [64]

    Schmidt, K. B. and Marshall, P. J. and Rix, H.-W. and Jester, S. and Hennawi, J. F. and Dobler, G. , title =. , volume =. 2010 , doi =

    2010

  63. [65]

    , title =

    Kozłowski, S. , title =. , volume =. 2016 , doi =

    2016

  64. [66]

    Bhatti, W. A. and Richmond, M. W. and Ford, H. C. and Petro, L. D. , title =. , volume =. 2010 , doi =

    2010

  65. [67]

    and Grav, T

    Mainzer, A. and Grav, T. and Masiero, J. and et al. , title =. , volume =. 2011 , doi =

    2011

  66. [68]

    and Assef, R

    Stern, D. and Assef, R. J. and Benford, D. J. and et al. , title =. , volume =. 2012 , doi =

    2012

  67. [69]

    Assef, R. J. and Stern, D. and Kochanek, C. S. and et al. , title =. , volume =. 2013 , doi =

    2013

  68. [70]

    Wright, E. L. and Eisenhardt, P. R. M. and Mainzer, A. K. and et al. , title =. , volume =. 2010 , doi =

    2010

  69. [71]

    Cutri, R. M. and Wright, E. L. and Conrow, T. and et al. , title =. 2013 , publisher =

    2013

  70. [72]

    , keywords =

    Active galactic nuclei: what's in a name?. , keywords =. doi:10.1007/s00159-017-0102-9 , archivePrefix =. 1707.07134 , primaryClass =

    Pith/arXiv arXiv

  71. [73]

    Di Matteo and V

    T. Di Matteo and V. Springel and L. Hernquist , title =. Nature , year =

  72. [74]

    D. J. Croton and V. Springel and S. D. M. White and G. De Lucia and C. S. Frenk and L. Gao and A. Jenkins and G. Kauffmann and J. F. Navarro and N. Yoshida , title =. , year =

  73. [75]

    Gebhardt and R

    K. Gebhardt and R. M. Richstone and D. K. Ajhar and M. Lauer and S. Tremaine and C. Bender and R. Bower and A. Dressler and S. M. Faber and A. V. Filippenko and R. Green and L. C. Ho and J. Kormendy and T. Rix , title =. , year =

  74. [76]

    Ferrarese and D

    L. Ferrarese and D. Merritt , title =. , year =

  75. [77]

    Kormendy and L

    J. Kormendy and L. C. Ho , title =. , year =

  76. [78]

    T. M. Heckman and P. N. Best , title =. , year =

  77. [79]

    Hoyle , title =

    B. Hoyle , title =. Astronomy & Computing , year =

  78. [80]

    Bulletin of the American Astronomical Society , year = 2019, volume =

    SDSS-V Pioneering Panoptic Spectroscopy. Bulletin of the American Astronomical Society , year = 2019, volume =

    2019

  79. [81]

    arXiv e-prints , keywords =

    The DESI Experiment Part I: Science,Targeting, and Survey Design. arXiv e-prints , keywords =. doi:10.48550/arXiv.1611.00036 , archivePrefix =. 1611.00036 , primaryClass =

    Pith/arXiv arXiv doi:10.48550/arxiv.1611.00036

  80. [82]

    Ground-based and airborne instrumentation for astronomy VI , volume=

    Prime Focus Spectrograph (PFS) for the Subaru telescope: overview, recent progress, and future perspectives , author=. Ground-based and airborne instrumentation for astronomy VI , volume=. 2016 , publisher=

    2016

Showing first 80 references.