Pith. sign in

REVIEW 2 major objections 5 minor 43 references

The paper shows that after correcting the signal loss from foreground removal, HI-galaxy cross-power spectra recover the input dark-energy parameters w0 and wa without significant bias.

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 18:53 UTC pith:DK4RWXD4

load-bearing objection Worth engaging: the transfer-function correction logic holds and end-to-end simulations recover w0-wa, but the LSST-like claim rests on an unrealistically sharp photo-z assumption that is never tested. the 2 major comments →

arxiv 2607.17141 v2 pith:DK4RWXD4 submitted 2026-07-19 astro-ph.CO

Unveiling dark energy properties with high-sensitivity cross-correlations of neutral hydrogen intensity mapping and galaxy surveys

Alessandro Marins , Chang Feng , Filipe B. Abdalla This is my paper
classification astro-ph.CO PACS 98.80.-k95.36.+x
keywords 21 cm intensity mappingneutral hydrogendark energyw0-wa parameterizationcross-correlationforeground removalindependent component analysiscosmology
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 asks whether 21 cm neutral-hydrogen intensity mapping can do more than detect the cosmic signal: whether, after removing bright radio foregrounds and correcting the resulting bias, it can constrain how dark energy evolves. It simulates the radio sky with galactic foregrounds, cleans the maps with a blind independent-component analysis, and cross-correlates the cleaned hydrogen signal with narrow redshift slices of a galaxy survey. The central finding is that the signal loss that previously damaged such cross-correlations is avoided when the galaxy kernels are narrow, and the amplitude-corrected cross-power spectra recover the fiducial dynamical dark energy parameters (w0, wa) with no significant bias when all data sets are combined. If this result holds, combining future 21 cm surveys with optical galaxy surveys can provide an independent test of whether dark energy is a cosmological constant.

Core claim

The authors build end-to-end simulated observations: a low-redshift 21 cm intensity mapping experiment plus a galaxy survey with 40 tomographic bins (seven of which pass signal-to-noise cuts), including synchrotron and free-free foregrounds. They remove the foregrounds blind, compute transfer functions from hundreds of injected Gaussian hydrogen simulations, and correct the cross-power spectrum amplitudes. A Bayesian fit to the two-parameter dark energy equation of state (w0, wa) then yields posteriors whose best-fit values match the input cosmology. Individual data sets can show skewed one-dimensional posteriors, but when all seven sets are combined the skewness is suppressed and no signifi

What carries the argument

The central objects are the foreground- and bias-mitigated HI-galaxy cross-power spectra C^{HI,g}_ℓ(ν_i, ν_j), together with the transfer function T_ℓ(ν) that corrects the amplitude suppression produced by foreground removal. The transfer function is estimated by injecting hundreds of Gaussian hydrogen realizations through the same cleaning pipeline and cross-correlating them with galaxy maps, using an outlier-robust average to reduce realization-dependent variance. The physical mechanism that makes the scheme work is the narrow redshift kernels of the galaxy tomographic bands, which limit the overlap between the hydrogen channels and the galaxy field, so that cleaning does not erase the cor

Load-bearing premise

The unbiased parameter recovery rests on the galaxy survey having a photometric redshift error as small as 0.001, about 20 to 50 times better than the realistic scatter of the survey class considered, which keeps the galaxy kernels narrow enough that foreground removal does not erase the cross-correlation signal.

What would settle it

Redo the end-to-end forecast with a photometric redshift scatter of 0.03 and a 10-bin partition: if the recovered (w0, wa) posterior shifts away from the input values by more than the expected statistical scatter, or if the signal loss reappears in the amplitude-corrected spectra, then the paper's claim of unbiased dynamical dark energy inference is not generic.

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

If this is right

  • HI intensity mapping cross-correlated with narrow-band galaxy surveys becomes a viable independent probe of the dark energy equation of state, complementing distance-based probes.
  • Survey designers will favor spectroscopic or narrow-band galaxy samples when pairing with 21 cm experiments, because narrow redshift kernels are what suppress the signal loss.
  • The transfer-function correction procedure can be applied to data from upcoming 21 cm arrays and wide optical surveys to produce foreground-mitigated cross-spectra for parameter estimation.
  • Combining multiple redshift windows is important: the paper shows that single sets can be skewed while the combined set is unbiased.

Where Pith is reading between the lines

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

  • A plausible stress test is to rerun the pipeline with a realistic photometric redshift scatter around 0.03, as shown in one alternative configuration in the paper; the unbiased recovery is demonstrated only for the near-spectroscopic 0.001 case.
  • Because the simulations are Gaussian, nonlinear clustering and non-Gaussian covariance are not tested; a non-Gaussian mock suite would be a stronger validation before application to real data.
  • The same transfer-function approach could extend to other line-intensity cross-correlation probes, such as carbon monoxide or ionized carbon maps, where foreground cleaning also suppresses long-wavelength modes.
  • The fit is calibrated to the same fiducial cosmology used to generate the simulations, so the result checks internal consistency; the decisive test will be a blind analysis on real data with unknown input cosmology.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper presents end-to-end simulations of low-redshift HI intensity mapping cross-correlated with a narrow-band 'LSST-like' galaxy survey. The pipeline consists of FastICA foreground removal, an injection-based transfer-function correction for the signal loss induced by the cleaning, and a Bayesian fit of the dynamical dark energy parameters (w0, wa) to the foreground- and bias-mitigated HI-galaxy cross-power spectra. The authors report that the corrected spectra match the theoretical input (Fig. 4) and that the combined seven data sets / nineteen cross-spectra recover the fiducial values with no significant biases (Fig. 5). Appendix A proves that off-diagonal galaxy auto-covariance elements do not affect the transfer-function calculation. The central claim is that, for the first time, cosmological parameter inference for dynamical dark energy is tested using foreground- and bias-mitigated HI-galaxy cross-power spectra.

Significance. If the result holds, the paper is a useful proof-of-concept that blind foreground removal, when combined with a transfer-function correction, can yield unbiased dark-energy parameter constraints from HI-galaxy cross-correlations. The end-to-end validation with 2000 realizations, the Huber-estimated transfer functions, and the Appendix A proof are concrete strengths. However, the demonstration is conditional on an idealized and unrealistic photometric redshift uncertainty for an 'LSST-like' survey: σ0_z = 0.001. The paper itself shows a σz = 0.03, 10-bin configuration in Fig. 1 but does not use it in any forecast. Because the signal-loss-mitigation strategy relies on galaxy kernels being narrow relative to the HI channels, the advertised LSST applicability is not yet established.

major comments (2)
  1. [Sec. 3, Table 1; Sec. 4.2; Sec. 5; Sec. 6] The entire forecast uses σ0_z = 0.001 for the 'LSST-like' photometric survey. This is roughly 20–50 times smaller than LSST's realistic photo-z scatter (σz ≈ 0.02–0.05 for z ≲ 0.5). The σz = 0.03, Ntomo = 10 case appears only in Fig. 1 and is never propagated through the pipeline. Since the signal-loss correction in Sec. 4.2 depends on galaxy kernels that are sharply localized relative to the HI channels, the unbiased recovery shown in Fig. 5 is not demonstrated for the actual LSST-like configuration. The statement in Sec. 4 that 'spectroscopic galaxy samples with much sharper redshift coverage would be favorable' effectively concedes this dependence. Please either rerun the end-to-end pipeline with a realistic σz (e.g., the σz = 0.03 case already shown), or explicitly restrict the claims to narrow-band/spectroscopic surveys and adjust the title/abstract accordingly.
  2. [Sec. 2, selection criteria] The data selection into the seven Sets is controlled by hand-chosen SNR thresholds: ε = 0.6 for the uncorrected spectra and ε = 1 for the corrected spectra. These thresholds determine which galaxy bands and HI channels enter the forecast. No robustness test is presented for these choices, and the selection is applied to estimated spectra, so it could in principle select favorable noise or signal-loss realizations. Please demonstrate that the parameter recovery and the final contours are stable to reasonable variations in ε, or replace the hand-chosen cuts with a more principled SNR selection criterion.
minor comments (5)
  1. [Fig. 3 caption/panel] The text says the left panel corresponds to the 'tenth HI channel,' but the panel label reads 'HI channel: 9.' Please correct this inconsistency.
  2. [Sec. 2, Eq. (2)] b_HI(z) is written as a function of redshift, but the text immediately sets b_HI = 1. Clarify whether b_HI is meant to be redshift-dependent or constant.
  3. [Sec. 2, Eq. (15)] The likelihood uses only the diagonal band-power variance σb and ignores covariance between bandpowers and between overlapping cross-spectra. The p-value discussion in Sec. 5 assumes independent χ² terms. Please state explicitly whether the off-diagonal covariance is negligible for this setup, or include a simulation-based covariance.
  4. [Sec. 4.2, Eq. (12)] The transfer-function injection FICA'[X'_HI + Xobs] − FICA[Xobs] assumes approximate linearity of the foreground cleaning with respect to the injected HI signal. The end-to-end validation is reassuring, but a brief discussion or test of this linearity assumption (e.g., varying the injected amplitude) would improve the methodological rigor.
  5. [References] The reference 'Hyvärinen 1999, Survey on Independent Component Analysis' is incomplete; provide full bibliographic details. Also, some 2026 references may benefit from a note confirming their status (e.g., published or preprint).

Circularity Check

0 steps flagged

No significant circularity: the parameter-recovery demonstration is a self-contained simulation pipeline test, and the self-citations are methodological background rather than load-bearing.

full rationale

The paper's central claim is that foreground- and bias-mitigated HI-galaxy cross-power spectra can yield unbiased constraints on (w0, wa) in simulated future surveys. The derivation chain is: generate correlated Gaussian HI and galaxy maps from a theoretical model, add realistic foregrounds, remove foregrounds with FastICA, estimate a transfer function from independent injected HI mocks (Eq. 14), apply that transfer function to separate realizations, and then fit (w0, wa) with a standard Gaussian likelihood (Eq. 15). The transfer function is not a cosmological parameter fitted to the final data; it is a calibration estimated from 2000 independent realizations and applied to different maps. Therefore the posterior recovery of the fiducial cosmology is a consistency/closure test of the pipeline, not a fitted quantity renamed as a prediction. The reliance on the authors' prior work (AM2025, Marins et al. 2022) is for motivation and methodological setup (e.g., four FastICA components), not for a uniqueness theorem or to forbid alternative procedures. The unrealistic photometric redshift uncertainty (σ0_z = 0.001) and the untested broader-kernel regime are genuine realism/correctness concerns, but they do not make the derivation circular. Consequently, no step in the paper reduces to its own inputs by construction, and the circularity score is low.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The central pipeline rests on several idealized inputs, most notably the photo-z error and the linear separability assumption; these do not invalidate the simulation but limit the generality.

free parameters (4)
  • Photometric redshift error σ0_z = 0.001
    Assumed for the LSST-like survey (Table 1). Actually LSST photo-z errors are ≳0.02; this value makes galaxy bands effectively narrow, which is the key to suppressing signal loss.
  • SNR selection thresholds ε = 0.6 (no correction), 1.0 (with correction)
    Chosen by hand in Sec. 2 to select the seven data sets; different thresholds change the included cross-spectra.
  • Number of FastICA components = 4
    Selected based on the authors' prior work (Marins et al. 2022); affects how much HI signal is removed with foregrounds.
  • Multipole range = 30<ℓ<400
    Chosen to avoid foreground leakage and nonlinear scales; the ℓ range affects SNRs and parameter constraints.
axioms (5)
  • standard math Limber approximation
    Used in Eq. (6) to compute angular power spectra; valid for broad kernels but may be less accurate for the very narrow bins used here.
  • domain assumption Gaussian realizations of HI and galaxy density fields
    Sec. 4.1 generates correlated Gaussian maps; nonlinear structure is only included via halofit in the power spectrum, not in the fields.
  • ad hoc to paper Linearity of FastICA cleaning for injection method
    Eq. (12) assumes the difference of cleaned maps isolates the injected HI signal loss; this requires the foreground solution to be linearly separable from the HI component.
  • domain assumption Constant HI bias b_HI=1 and Ω_HI=4.86e-4
    Eq. (1)-(2) adopt fixed HI bias and abundance from Bull et al. 2015; in reality bias evolves and is uncertain, and the inference does not marginalize over it.
  • domain assumption DarkEnergyPPF w0-wa parametrization
    Sec. 2: the dark energy model is an effective two-parameter parametrization, not a physical model, and assumes independently conserved dark energy.

pith-pipeline@v1.3.0-alltime-deepseek · 12679 in / 15446 out tokens · 141965 ms · 2026-08-01T18:53:45.194250+00:00 · methodology

0 comments
read the original abstract

The redshifted 21 cm line from the hyperfine structure of neutral hydrogen atoms is a promising tracer for the three-dimensional evolution of our universe. Its broad spatial and temporal coverage is crucial for understanding the complex nature of dark matter and dark energy. However, it is very challenging to directly detect the 21 cm signal due to the existence of radio foreground contaminants that are orders of magnitude brighter. Therefore, mitigating the foreground contamination becomes an indispensable task for detecting the 21 cm signal, which is also expected to be correlated with the dark-matter-dominated large-scale structure. The cross-correlations between the neutral hydrogen intensity mapping and galaxy surveys in the future can not only confirm a detection of the 21 cm signal but can also be a complementary probe for 21 cm cosmology. To meet the precision requirements for cosmological studies, it is important to investigate the complex features of the estimated cross-correlations using numerical simulations. In this work, we simulate the low-redshift HI observations and narrow-band optical surveys in the future and obtain the foreground- and bias-mitigated HI-galaxy cross-power spectra, with which we perform a Bayesian analysis to infer cosmological parameters of dynamical dark energy model. The method developed in this work will be important for cosmological studies with the 21 cm intensity mapping.

Figures

Figures reproduced from arXiv: 2607.17141 by Alessandro Marins, Chang Feng, Filipe B. Abdalla.

Figure 1
Figure 1. Figure 1: The minimum and maximum redshifts of each i-bin are constructed as z i max = z i min + ∆z. The galaxy kernel for each narrow band is W (i) G (z) = bG(z)n (i) (z), (5) where bG(z) = √ 1 + z is the linear galaxy bias (Euclid Collaboration et al. 2020). We adopt the Limber approximation (Limber 1953) to calculate the cross correlations for two generic LSS tracers A and B measured at two frequencies νi and νj … view at source ↗
Figure 2
Figure 2. Figure 2 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Different types of cross-power spectra between HI IM and an LSST-like galaxy survey at two redshifts. The left panel corresponds to the second galaxy bin and the tenth HI channel, at a mean redshift ¯z = 0.21, while the right panel corresponds to the seventh galaxy bin and the 26th HI channel, at a mean redshift ¯z = 0.40. The panels display the spectra of the model (black dashed), the signal-only Gaussian… view at source ↗
Figure 4
Figure 4. Figure 4: Foreground- and bias-mitigated HI-galaxy cross-power spectra at multiple redshifts. The estimated cross-spectra are consistent with the theoretical calculations. spherical harmonic coefficients of HI and galaxy maps in each realization to get the power spectrum defined as ⟨Xˆ HI,∗ ℓm gℓ ′m′ ⟩ = CˆHI,g ℓ δℓℓ′δmm′ . (13) The transfer function is thus calculated from Tℓ(ν) = * C HI′ ,g′ ℓ (ν) C HI(2),g′ ℓ (ν)… view at source ↗
Figure 5
Figure 5. Figure 5: Posterior distributions of the cosmological parameters (w0, wa) for the dynamical dark energy models. The left panel shows the joint parameter estimation from the seven data sets with nineteen cross-power spectra, which are generated from theoretical calculations (“Theory”), Gaussian simulations (“Simulated”), and the realistic simulations with both the foreground and bias mitigated (“Estimated”). In the r… 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

43 extracted references · 4 canonical work pages · 1 internal anchor

  1. [1]

    2025, PhRvD, 112, 083515, doi: 10.1103/tr6y-kpc6

    Abdul Karim, M., Aguilar, J., Ahlen, S., et al. 2025, PhRvD, 112, 083515, doi: 10.1103/tr6y-kpc6

    doi:10.1103/tr6y-kpc6 2025

  2. [2]

    Santos, M. G. 2015a, The Astrophysical Journal, 814, 145, doi: 10.1088/0004-637x/814/2/145 10

    doi:10.1088/0004-637x/814/2/145

  3. [3]

    G., & Santos, M

    Alonso, D., Bull, P., Ferreira, P. G., & Santos, M. G. 2015b, Monthly Notices of the Royal Astronomical Society, 447, 400, doi: 10.1093/mnras/stu2474

    doi:10.1093/mnras/stu2474

  4. [4]

    2019, MNRAS, 484, 4127, doi: 10.1093/mnras/stz093

    Alonso, D., Sanchez, J., Slosar, A., & LSST Dark Energy Science Collaboration. 2019, MNRAS, 484, 4127, doi: 10.1093/mnras/stz093

    doi:10.1093/mnras/stz093 2019

  5. [5]

    J., Luciw, N

    Anderson, C. J., Luciw, N. J., Li, Y. C., et al. 2018, MNRAS, 476, 3382, doi: 10.1093/mnras/sty346

    doi:10.1093/mnras/sty346 2018

  6. [6]

    2001, PhR, 340, 291, doi: 10.1016/S0370-1573(00)00082-X

    Bartelmann, M., & Schneider, P. 2001, PhR, 340, 291, doi: 10.1016/S0370-1573(00)00082-X

    doi:10.1016/s0370-1573(00)00082-x 2001

  7. [7]

    L., Hill, R

    Bennett, C. L., Hill, R. S., Hinshaw, G., et al. 2003, ApJS, 148, 97, doi: 10.1086/377252

    doi:10.1086/377252 2003

  8. [8]

    A., et al

    Bigot-Sazy, M.-A., Dickinson, C., Battye, R. A., et al. 2015, Monthly Notices of the Royal Astronomical Society, 454, 3240, doi: 10.1093/mnras/stv2153

    doi:10.1093/mnras/stv2153 2015

  9. [9]

    G., Patel, P., & Santos, M

    Bull, P., Ferreira, P. G., Patel, P., & Santos, M. G. 2015a, ApJ, 803, 21, doi: 10.1088/0004-637X/803/1/21 —. 2015b, ApJ, 803, 21, doi: 10.1088/0004-637X/803/1/21

    doi:10.1088/0004-637x/803/1/21

  10. [10]

    P., Irfan, M

    Carucci, I. P., Irfan, M. O., & Bobin, J. 2020, Monthly Notices of the Royal Astronomical Society, 499, 304

    2020

  11. [11]

    Chang, T.-C., Pen, U.-L., Bandura, K., & Peterson, J. B. 2010, Nature, 466, 463, doi: 10.1038/nature09187

    doi:10.1038/nature09187 2010

  12. [12]

    B., Harker, G., et al

    Chapman, E., Abdalla, F. B., Harker, G., et al. 2012, MNRAS, 423, 2518, doi: 10.1111/j.1365-2966.2012.21065.x

    arXiv 2012

  13. [13]

    2025, MNRAS, 542, L1, doi: 10.1093/mnrasl/slaf058

    Chen, Z. 2025, MNRAS, 542, L1, doi: 10.1093/mnrasl/slaf058

    doi:10.1093/mnrasl/slaf058 2025

  14. [14]

    G., et al

    Cunnington, S., Li, Y., Santos, M. G., et al. 2023, MNRAS, 518, 6262, doi: 10.1093/mnras/stac3060

    doi:10.1093/mnras/stac3060 2023

  15. [15]

    L., et al

    Cunnington, S., Barberi-Squarotti, M., Bernal, J. L., et al. 2026, Astrophysics and Space Science, 371, 16, doi: 10.1007/s10509-026-04547-7

    doi:10.1007/s10509-026-04547-7 2026

  16. [16]

    2009, A&A, 493, 835, doi: 10.1051/0004-6361:200810514 Euclid Collaboration, Blanchard, A., Camera, S., et al

    Delabrouille, J., Cardoso, J.-F., Le Jeune, M., et al. 2009, A&A, 493, 835, doi: 10.1051/0004-6361:200810514 Euclid Collaboration, Blanchard, A., Camera, S., et al. 2020, A&A, 642, A191, doi: 10.1051/0004-6361/202038071

    doi:10.1051/0004-6361:200810514 2009

  17. [17]

    2008, PhRvD, 78, 087303, doi: 10.1103/PhysRevD.78.087303

    Fang, W., Hu, W., & Lewis, A. 2008, PhRvD, 78, 087303, doi: 10.1103/PhysRevD.78.087303

    doi:10.1103/physrevd.78.087303 2008

  18. [18]

    Fornazier, K. S. F., Abdalla, F. B., Remazeilles, M., et al. 2022, A&A, 664, A18, doi: 10.1051/0004-6361/202141707 G´ orski, K. M., Hivon, E., Banday, A. J., et al. 2005, ApJ, 622, 759, doi: 10.1086/427976

    doi:10.1051/0004-6361/202141707 2022

  19. [19]

    Hunter, J. D. 2007, Computing in Science and Engineering, 9, 90, doi: 10.1109/MCSE.2007.55 Hyv¨ arinen, A. 1999, Survey on Independent Component Analysis

    doi:10.1109/mcse.2007.55 2007

  20. [20]

    2011, CAMB: Code for Anisotropies in the Microwave Background, Astrophysics Source Code Library, record ascl:1102.026

    Lewis, A., & Challinor, A. 2011, CAMB: Code for Anisotropies in the Microwave Background, Astrophysics Source Code Library, record ascl:1102.026

    2011

  21. [21]

    Limber, D. N. 1953, ApJ, 117, 134, doi: 10.1086/145672 LSST Science Collaboration, Abell, P. A., Allison, J., et al. 2009, arXiv e-prints, arXiv:0912.0201, doi: 10.48550/arXiv.0912.0201

    Pith/arXiv arXiv doi:10.48550/arxiv.0912.0201 1953

  22. [22]

    2006, ApJ, 636, 21, doi: 10.1086/497068

    Ma, Z., Hu, W., & Huterer, D. 2006, ApJ, 636, 21, doi: 10.1086/497068

    doi:10.1086/497068 2006

  23. [23]

    2002, MNRAS, 334, 53, doi: 10.1046/j.1365-8711.2002.05425.x

    Maino, D., Farusi, A., Baccigalupi, C., et al. 2002, MNRAS, 334, 53, doi: 10.1046/j.1365-8711.2002.05425.x

    arXiv 2002

  24. [24]

    2008, PhRvD, 78, 023529, doi: 10.1103/PhysRevD.78.023529

    Zahn, O. 2008, PhRvD, 78, 023529, doi: 10.1103/PhysRevD.78.023529

    doi:10.1103/physrevd.78.023529 2008

  25. [25]

    Marins, A., Feng, C., & Abdalla, F. B. 2025, ApJS, 278, 32, doi: 10.3847/1538-4365/adc8a4

    work page doi:10.3847/1538-4365/adc8a4 2025

  26. [26]

    B., Fornazier, K

    Marins, A., Abdalla, F. B., Fornazier, K. S. F., et al. 2022, arXiv e-prints, arXiv:2209.11701. https://arxiv.org/abs/2209.11701

    arXiv 2022

  27. [27]

    W., Switzer, E

    Masui, K. W., Switzer, E. R., Banavar, N., et al. 2013, ApJL, 763, L20, doi: 10.1088/2041-8205/763/1/L20

    doi:10.1088/2041-8205/763/1/l20 2013

  28. [28]

    B., Novaes, C

    Motta, P., Abdalla, F. B., Novaes, C. P., et al. 2026, A&A, doi: 10.1051/0004-6361/202555763

    work page doi:10.1051/0004-6361/202555763 2026

  29. [29]

    2016, Monthly Notices of the Royal Astronomical Society, 456, 2749

    Olivari, L., Remazeilles, M., & Dickinson, C. 2016, Monthly Notices of the Royal Astronomical Society, 456, 2749

    2016

  30. [30]

    2011, Journal of Machine Learning Research, 12, 2825 Planck Collaboration, Adam, R., Ade, P

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825 Planck Collaboration, Adam, R., Ade, P. A. R., et al. 2016, A&A, 594, A10, doi: 10.1051/0004-6361/201525967

    doi:10.1051/0004-6361/201525967 2011

  31. [31]

    R., & Loeb, A

    Pritchard, J. R., & Loeb, A. 2008, PhRvD, 78, 103511, doi: 10.1103/PhysRevD.78.103511

    doi:10.1103/physrevd.78.103511 2008

  32. [32]

    2011, MNRAS, 418, 467, doi: 10.1111/j.1365-2966.2011.19497.x

    Remazeilles, M., Delabrouille, J., & Cardoso, J.-F. 2011, MNRAS, 418, 467, doi: 10.1111/j.1365-2966.2011.19497.x

    arXiv 2011

  33. [33]

    J., Bigot-Sazy, M

    Remazeilles, M., Dickinson, C., Banday, A. J., Bigot-Sazy, M. A., & Ghosh, T. 2015, MNRAS, 451, 4311, doi: 10.1093/mnras/stv1274

    doi:10.1093/mnras/stv1274 2015

  34. [34]

    2026, PhRvD, 113, 023521, doi: 10.1103/k6qq-7f7p

    Shen, D., Kokron, N., & Schaan, E. 2026, PhRvD, 113, 023521, doi: 10.1103/k6qq-7f7p

    work page doi:10.1103/k6qq-7f7p 2026

  35. [35]

    Silva, G. A. S., Novaes, C. P., Wuensche, C. A., et al. 2025, arXiv, doi: 10.48550/arxiv.2506.19068

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2506.19068 2025

  36. [36]

    P., Cunnington, S., et al

    Spinelli, M., Carucci, I. P., Cunnington, S., et al. 2021, Monthly Notices of the Royal Astronomical Society, 509, 2048, doi: 10.1093/mnras/stab3064

    doi:10.1093/mnras/stab3064 2021

  37. [37]

    R., Masui, K

    Switzer, E. R., Masui, K. W., Bandura, K., et al. 2013, MNRAS, 434, L46, doi: 10.1093/mnrasl/slt074

    doi:10.1093/mnrasl/slt074 2013

  38. [38]

    2012, ApJ, 761, 152, doi: 10.1088/0004-637X/761/2/152 The LSST Dark Energy Science Collaboration,

    Oguri, M. 2012, ApJ, 761, 152, doi: 10.1088/0004-637X/761/2/152 The LSST Dark Energy Science Collaboration,

    doi:10.1088/0004-637x/761/2/152 2012

  39. [39]

    2018, arXiv e-prints, arXiv:1809.01669, doi: 10.48550/arXiv.1809.01669

    Mandelbaum, R., Eifler, T., et al. 2018, arXiv e-prints, arXiv:1809.01669, doi: 10.48550/arXiv.1809.01669

    Pith/arXiv arXiv doi:10.48550/arxiv.1809.01669 2018

  40. [40]

    2017, MNRAS, 469, 2821, doi: 10.1093/mnras/stx949 11 van der Walt, S., Colbert, S

    Thorne, B., Dunkley, J., Alonso, D., & Næss, S. 2017, MNRAS, 469, 2821, doi: 10.1093/mnras/stx949 11 van der Walt, S., Colbert, S. C., & Varoquaux, G. 2011, Computing in Science and Engineering, 13, 22, doi: 10.1109/MCSE.2011.37

    doi:10.1093/mnras/stx949 2017

  41. [41]

    N., & Ripley, B

    Venables, W. N., & Ripley, B. D. 2013, Modern applied statistics with S, Vol. 4

    2013

  42. [42]

    W., et al

    Wolz, L., Pourtsidou, A., Masui, K. W., et al. 2022, MNRAS, 510, 3495, doi: 10.1093/mnras/stab3621

    doi:10.1093/mnras/stab3621 2022

  43. [43]

    2021, MNRAS, 504, 5231, doi: 10.1093/mnras/stab1197

    Yohana, E., Ma, Y.-Z., Li, D., Chen, X., & Dai, W.-M. 2021, MNRAS, 504, 5231, doi: 10.1093/mnras/stab1197

    doi:10.1093/mnras/stab1197 2021