Can machine learning improve the detectability and disentanglement of the gravitational-wave background?

Hugo Einsle; Jishnu Suresh; Mairi Sakellariadou; Marie Anne Bizouard; Tania Regimbau

REVIEW 4 major objections 4 minor 25 references

This paper claims that a custom multi-scale multi-headed autoencoder followed by MCMC sampling can detect a compact-binary-coalescence gravitational-wave background as faint as 4.3e-9 at 25 Hz in 108 days of fourth-observing-run-like mock d

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

2026-08-04 00:49 UTC pith:VE3SLFVU

load-bearing objection A useful, honest simulation study that extends the authors' ML pipeline to O4a-like noise, with promising but not yet cleanly supported detection thresholds because the autoencoder train/test split is never documented. the 4 major comments →

arxiv 2608.00281 v1 pith:VE3SLFVU submitted 2026-07-31 gr-qc astro-ph.HEastro-ph.IM

Can machine learning improve the detectability and disentanglement of the gravitational-wave background?

Hugo Einsle , Marie Anne Bizouard , Tania Regimbau , Mairi Sakellariadou , Jishnu Suresh This is my paper

classification gr-qc astro-ph.HEastro-ph.IM PACS 04.30.-w04.80.Nn07.05.Mh

keywords gravitational-wave backgroundautoencoderBayesian inferenceMarkov chain Monte Carlocompact binary coalescencecosmological backgroundspectral indexstochastic background

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 authors are trying to show that a machine-learning pipeline, built from a custom autoencoder and Bayesian inference, can detect the gravitational-wave background from compact binary mergers and separate it from a fainter cosmological background. Using 108 days of simulated noise that mimics the first period of the fourth observing run, they find decisive evidence for a compact-binary background at amplitude 4.3e-9 at 25 Hz in realistic noise, and they can recover a flat-spectrum cosmological component down to 9.7e-10. In blind comparisons, the pipeline recovers amplitudes and spectral indices more accurately than the standard cross-correlation approach, and it succeeds in separating the two background components where the standard approach fails. If these results carry over to real data, they would make gravitational-wave background detection and component separation achievable sooner and with less observing time.

Core claim

The paper establishes that the MSMHAutoencoder—a deep convolutional autoencoder with multiple scales and heads—can separate a stochastic gravitational-wave background from detector noise in 108-day mock datasets, and that subsequent MCMC sampling recovers the injected amplitude and spectral index within 1-sigma credible intervals. Setting a detection threshold at log10 noise Bayes factor greater than 3, they find a minimal detectable compact-binary background amplitude of 4.3^{+0.5}_{-0.4}×10^{-9} at 25 Hz in realistic noise, improving to 2.6e-9 when spectral lines are removed. They further inject a flat-spectrum cosmological background on top of a CBC background at the detection threshold a

What carries the argument

The MSMHAutoencoder is a deep convolutional neural network with Inception-like blocks designed to learn the stationary detector noise and subtract it, leaving the gravitational-wave background; it includes specialized sub-networks intended to suppress narrow spectral lines. After denoising, a Markov chain Monte Carlo stage samples the posterior of amplitude and spectral index for competing hypotheses, and the evidence ratio (Bayes factor) quantifies both detection of a signal and the presence of a cosmological component beyond the CBC background via B_cosmo = Z_CBC+cosmo / Z_CBC.

Load-bearing premise

The detection thresholds assume the autoencoder was tested only on signal injections and noise realizations it never saw during training, and that the Gaussian mock data faithfully represent real detector noise; the paper does not document a train/test split and explicitly says real-data validation is still required.

What would settle it

Retrain the autoencoder on half of the mock datasets and evaluate on the held-out half with fresh noise realizations and injections at the claimed threshold amplitudes; if the noise Bayes factor no longer exceeds log10(B)>3 at 4.3e-9, the reported detection thresholds are inflated by train/test leakage.

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

If this is right

With 108 days of fourth-observing-run-like data, a compact-binary background as faint as ~4.5e-9 at 25 Hz can be detected with decisive evidence in realistic noise; removing spectral lines lowers the threshold to ~2.6e-9.
A flat cosmological background can be isolated down to ~9.7e-10 even when a CBC background near the detection threshold is present.
The proposed pipeline recovers the spectral index more accurately than the standard cross-correlation approach and estimates the total background amplitude with uncertainties around 1e-10.
In two-component scenarios, the pipeline separates astrophysical and cosmological contributions, something the standard cross-correlation analysis in this test did not achieve.
Spectral lines, particularly the 60 Hz line, are the dominant sensitivity-limiting feature, costing 30-60% of performance.

Where Pith is reading between the lines

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

If the simulated sensitivity transfers to real data, the method could cut the observing time needed for a background detection by roughly a factor of five relative to cross-correlation, making earlier science with current detectors possible.
The paper does not report a formal train/test split; unless the evaluation injections and noise realizations were held out during training, the claimed Bayes factors are optimistic. This is the first thing to verify.
The observed tendency to overestimate the cosmological component while underestimating the CBC component suggests the Bayesian separation stage, rather than the denoiser, drives the apportionment; alternative priors or spectral parametrizations may be needed when the two components overlap.
The architecture's sensitivity to spectral lines implies that real-data gains might come from better line suppression in the network rather than notching, which also removes signal; this extension is flagged by the authors as follow-up work.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

A useful, honest simulation study that extends the authors' ML pipeline to O4a-like noise, with promising but not yet cleanly supported detection thresholds because the autoencoder train/test split is never documented.

read the letter

The punchline: this is a legitimate extension of the authors' MSMHAutoencoder program, and the new numerical results are worth taking seriously. But there is a real methodological gap — no train/test split is documented for the autoencoder — so the headline thresholds (4.3e-9 for a CBC background, 9.7e-10 for a cosmological component) should be treated as conditional on a clean held-out evaluation that the paper does not describe.

What is actually new: the O4a-like PSD study, the line-sensitivity analysis identifying the 60 Hz line as the dominant culprit, and the blind comparison against pygwb. The comparison is genuinely informative: DeepGWB recovers amplitudes and spectral indices more accurately than pygwb on the tested injections, and it is the only one of the two that separates combined CBC+cosmological components. The paper also says plainly that real-data validation is mandatory, which is the right kind of caveat.

Where it is soft: first, the Methods state only that "independent training runs" are performed for each mock dataset; there is no mention of held-out noise realizations or held-out signal injections. If the network trained on the same 108-day realizations it was tested on, the Bayes factors and thresholds reflect memorization, not generalization. The Fig. 3 caption's mention of curriculum learning focused on the lowest-amplitude signals makes me worry the injections themselves were part of training. This is the load-bearing weakness and it is fixable with a clean train/test split. Second, the likelihood used on the autoencoder's denoised output is never specified, which makes it hard to reproduce the MCMC stage. Third, the blind test is conducted by the same team on their own simulations; that is not a fatal flaw, but it is a weaker form of validation than an independent check. The comparison to a theoretical cross-correlation SNR=3 threshold is also approximate, as the paper acknowledges.

These are addressable conditions, not internal contradictions. The paper does not oversell itself; it explicitly flags the need for real-data validation and even quantifies the sensitivity loss from spectral lines. The architecture is the authors' prior work, so novelty is moderate, but the O4a-specific results are new and useful to the community.

My read: a serious referee should see this. The open questions — train/test split, likelihood specification, and ideally a code/data release — are exactly what peer review should catch. I would not desk-reject it, but I would send it back with a request for a held-out evaluation before the detection thresholds are advertised.

Referee Report

4 major / 4 minor

Summary. The paper presents a hybrid machine-learning pipeline (MSMHAutoencoder + MCMC) for detecting and separating gravitational-wave backgrounds in simulated LIGO noise. Using two 108-day O4a-like mock datasets (one with spectral lines, one without), the authors report decisive (log10 B > 3) detection of a CBC background down to ~4.3e-9 (with lines) and ~2.6e-9 (line-free), and separation of a flat cosmological component down to ~9.7e-10 (with lines) and ~6.4e-10 (line-free). They compare to pygwb in a blind-test appendix, claiming more accurate amplitude and spectral-index recovery. The manuscript includes posterior plots, Bayes factors, and tables of recovered parameters.

Significance. If the reported sensitivities are unbiased, this would be a substantial improvement over standard cross-correlation methods for stochastic backgrounds, and would be the first demonstration of component separation in O4a-like conditions. Strengths: the use of standard O4a PSDs, comparisons to a public pipeline, credible intervals, and explicit recognition that real-data validation is required. The main weakness is that the machine-learning evaluation is not documented tightly enough to rule out training/evaluation leakage or an underspecified statistical model. The paper's value is methodological; it sets up a clear next step rather than claiming a direct detection.

major comments (4)

[Methods, 'We perform independent training runs...'] No train/validation/test split is described. The detection thresholds are computed on the same dataset type used for training, and Figure 3's caption mentions curriculum learning focusing on the lowest amplitude signals, which suggests injected signals may have been used in training. If the test injections or noise realizations were not held out, the reported Bayes factors could be inflated by memorization. Please specify exactly how the data used for Figures 2-4 and Tables I-II were separated from training data (e.g., disjoint noise realizations and signal injections), and how the autoencoder output is calibrated on held-out data.
[Methods, MCMC stage] The likelihood used on the autoencoder output is never stated. The paper says 'we characterize the recovered signal components via MCMC' but does not define the noise model, the likelihood, or how the autoencoder output is converted to a spectrum with uncertainties. Without this, the Bayes factors and credible intervals are not reproducible. Please provide the likelihood, any assumed correlations between bins or time segments, and the data reduction from autoencoder output to likelihood input.
[Results, cross-correlation comparison] The claim of a factor 2 improvement in amplitude and factor ~5 in observing time is based on an approximate relation log B ~ SNR^2/2, but no derivation or reference is given for the cross-correlation sensitivity at O4a, and the mapping between log10 B = 3 and SNR = 3 is not argued. The authors already have a direct pygwb comparison in Table I; they should either use that to draw the sensitivity comparison or state all assumptions and compute both thresholds in a unified way.
[Appendix, blind-test comparison] The term 'blind' is not defined; it is not stated who generated the injections and whether either pipeline was run without knowledge of the injected parameters. Also, the autoencoder training for the blind test is not described with respect to whether the blind datasets were seen during training. This matters for the claim that DeepGWB outperforms pygwb. Please clarify the blinding protocol and the training/evaluation split used in the appendix.

minor comments (4)

[Throughout] Typographical errors: 'o test this capability' should be 'To test'; 'as we shown' should be 'as we show'; appendix contains 'stochasticWe', 'with byDeepGWB', 'pygwbwas unable', and a repeated sentence about pygwb failing.
[Reference [18]] The reference is incomplete: 'R. Tania and S. Jishnu, 10.48550/arXiv.2506.12237' lacks a title and the authors' full names. Please correct.
[Figures 2 and 3] The color scale indicating log10(B) is described in the caption but no color bar is shown in the figure. Add a color bar or an explicit legend so the reader can map marker color to evidence values.
[Discussion, factor ~5] If sensitivity amplitude scales as T^{-1/2} for fixed SNR, a factor 2 improvement in amplitude corresponds to a factor 4 in observing time. The paper states 'factor ~5'; please clarify whether this includes other effects or is an approximation.

Circularity Check

1 steps flagged

The reported detection thresholds are measured on the same mock datasets used to train the autoencoder, so the central sensitivity claim is not established as an independent prediction.

specific steps

fitted input called prediction [Methods (mock dataset generation and autoencoder training); Results (detection threshold); Fig. 3 caption]
"We perform independent training runs for each mock dataset using the MSMHAutoencoder architecture, following the methodology described in [17]. ... Both datasets include the same injected GWB signals. ... The non-uniformly varying scale of the ellipses between log10(Ωα) = −7.6 and −8.6 is due to the curriculum learning focusing on the lowest amplitude signals."

The autoencoder is trained on the same mock datasets (which contain the injected GWB signals) that are then used to evaluate detection thresholds and Bayes factors; the paper does not document any held-out test split. The curriculum-learning note confirms training targeted the low-amplitude injections whose recovery is later quoted as the sensitivity limit. Thus the recovered amplitudes and log10(B)>3 thresholds are estimates of the network's ability to reconstruct its own training data, not predictions on unseen data. The claimed detectability figures are therefore not independent of the fitted model.

full rationale

The central claim is an empirical performance measurement, so it does not reduce to an input by an equation; however, the training/evaluation design as described is circular in the ML sense. Methods states training runs are performed on each mock dataset, and the mock datasets are explicitly said to contain the injected GWB signals that are later recovered in Results. No train/test split is mentioned anywhere in the manuscript, and the Fig. 3 caption shows curriculum learning focused on the lowest-amplitude signals—exactly the regime of the quoted detection threshold. This makes the reported sensitivity an evaluation on the training distribution rather than a generalization result. The self-citation of the MSMHAutoencoder from [17] is not itself circular, since [17] is a separate published method; the limitation statement that real-data validation is 'mandatory' is a scope caveat, not a circular step. If a held-out test set exists and is described in [17], the circularity would be resolved, but based on the text before us the prediction is fitted input. Score 6 reflects one central 'prediction' (the detection threshold) reducing to the training input by construction.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

The central claim rests on the assumed two-component spectral model, the representativeness of O4a-like Gaussian mock data, and the unstated pipeline details (autoencoder training and MCMC likelihood). No new physical entities are introduced; the free parameters are the machine-learning hyperparameters inherited from [17], the chosen priors, and the analysis frequency band.

free parameters (3)

Autoencoder training hyperparameters and curriculum regime = not reported (inherited from [17])
Detection thresholds and Bayes factors depend on the network architecture, training schedule, and curriculum focusing on low amplitudes; none are specified in this paper.
Prior bounds on model parameters = log10 Ωα ∈ U(−13,−7), α ∈ U(−5,5), log10 Ω0 ∈ U(−13,−7)
Chosen by hand; Bayes factors depend on prior volume, though the ranges are wide and intended to be uninformative.
Analysis frequency range = 20-100 Hz
The quoted thresholds are restricted to this band; extending it could improve sensitivity, so the choice affects the central numbers.

axioms (4)

domain assumption The GWB is a superposition of a CBC component with Ω ∝ f^{2/3} below ~100 Hz and a flat cosmological component with Ω = const.
The injected signals and the likelihood model both assume these spectral shapes; the separation claim is specific to this model. Methods paragraph.
domain assumption The O4a-like noise PSDs (with and without spectral lines) and Gaussian realizations used in the mock datasets are representative of LIGO Hanford/Livingston O4a data.
The quoted detection thresholds are conditional on this representativeness; the authors note real data is needed for validation. Methods and Discussion.
domain assumption The mock datasets contain only Gaussian noise plus the injected GWB; no environmental correlated noise between detectors is included.
Real stochastic searches must contend with correlated noise (e.g., Schumann resonances); excluding it makes the simulated thresholds optimistic. Abstract and Methods.
domain assumption The autoencoder output is suitable for the MCMC likelihood used for evidence computation.
The paper does not specify the likelihood on the denoised output; if the autoencoder reshapes noise non-Gaussianly, the Bayes factors could be biased. Methods/Results.

pith-pipeline@v1.3.0-alltime-deepseek · 9289 in / 19035 out tokens · 178687 ms · 2026-08-04T00:49:38.433569+00:00 · methodology

0 comments

read the original abstract

Gravitational waves from compact binary coalescences and from early Universe processes are expected to form a gravitational-wave background. We employ a custom deep learning multi-scale multi-headed autoencoder architecture to isolate gravitational-wave background from detector noise, followed by a Markov chain Monte Carlo inference stage to separate the astrophysical and cosmological components. Analyzing $108$-day mock datasets representative of the first period of the fourth LIGO-Virgo-KAGRA observing run, we show that we can detect with high confidence --- $\log_{10}$ noise Bayes factor larger than 3 --- a compact binary coalescence gravitational-wave background with an amplitude of $4.3^{+0.5}_{-0.4}\times10^{-9}$ at $f_{\rm ref}=25\,\mathrm{Hz}$, which is a factor $\sim5$ higher than the amplitude expected from compact binary sources. We also show that we can isolate a cosmological -- assumed flat spectrum -- gravitational-wave background as weak as $ 9.7^{+2.5}_{-2.4} \times 10^{-10}$ from the expected compact binary coalescence gravitational-wave background within simulated Gaussian noise mimicking the LIGO detectors sensitivity achieved in the fourth observing run. In blind-test comparisons with the standard \texttt{pygwb} pipeline, we show that our method achieves more accurate amplitude and spectral-index recovery and enables the separation of astrophysical and cosmological background components.

Figures

Figures reproduced from arXiv: 2608.00281 by Hugo Einsle, Jishnu Suresh, Mairi Sakellariadou, Marie Anne Bizouard, Tania Regimbau.

**Figure 3.** Figure 3: shows similar results with spectral lines being removed from the O4a data. The detection sensitivity improves to Ωα = 2.6 +0.1 −0.1 × 10−9 for an injected amplitude of Ωα = 2.6 × 10−9 , recovered with log10(B) = 3.7. The ellipses are also slightly smaller than the ones for the O4a dataset containing spectral lines. By keeping or suppressing each spectral line in the data, we conclude that the 60 Hz line i… view at source ↗

**Figure 2.** Figure 2: we draw the credible-region contours of the recovered log10(Ωα) and α parameters of each injected astrophysical GWB. The faintest background amplitude is Ωα = 4.3 +0.5 −0.4 ×10−9 recovered for an injected CBC background of Ωα = 4.5 × 10−9 with log10(B) = 3.3. The α spectral index is well estimated within 1σ accuracy [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 4.** Figure 4: FIG. 4. Posterior distributions of the cosmological GWB am [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗

discussion (0)

Reference graph

Works this paper leans on

25 extracted references · 18 linked inside Pith

[1]

Christensen, Rept

N. Christensen, Rept. Prog. Phys.82, 016903 (2019), arXiv:1811.08797 [gr-qc]

Pith/arXiv arXiv 2019
[2]

Caprini and D

C. Caprini and D. G. Figueroa, Class. Quant. Grav.35, 163001 (2018), arXiv:1801.04268 [astro-ph.CO]

Pith/arXiv arXiv 2018
[3]

Regimbau, Res

T. Regimbau, Res. Astron. Astrophys.11, 369 (2011), arXiv:1101.2762 [astro-ph.CO]

Pith/arXiv arXiv 2011
[4]

B. P. Abbott et al. (KAGRA, LIGO Scientific, Virgo), Living Rev. Rel.19, 1 (2016), arXiv:1304.0670 [gr-qc]

Pith/arXiv arXiv 2016
[5]

A. G. Abac et al. (LIGO Scientific, VIRGO, KAGRA), (2025), arXiv:2510.26848 [gr-qc]

arXiv 2025
[6]

(2026), arXiv:2605.27225 [gr-qc]

Pith/arXiv arXiv 2026
[8]

Allen and J

B. Allen and J. D. Romano, Phys. Rev. D59, 102001 (1999), arXiv:gr-qc/9710117 [gr-qc]

Pith/arXiv arXiv 1999
[9]

P. M. Meyers, K. Martinovic, N. Christensen, and M. Sakellariadou, Phys. Rev. D102, 102005 (2020), arXiv:2008.00789 [gr-qc]

Pith/arXiv arXiv 2020
[10]

Martinovic, P

K. Martinovic, P. M. Meyers, M. Sakellariadou, and N. Christensen, Phys. Rev. D103, 043023 (2021), arXiv:2011.05697 [gr-qc]

Pith/arXiv arXiv 2021
[11]

Aasi et al

J. Aasi et al. (LIGO Scientific), Class. Quant. Grav.32, 074001 (2015), arXiv:1411.4547 [gr-qc]

Pith/arXiv arXiv 2015
[12]

Regimbau, M

T. Regimbau, M. Evans, N. Christensen, E. Katsavouni- dis, B. Sathyaprakash, and S. Vitale, Phys. Rev. Lett. 118, 151105 (2017), arXiv:1611.08943 [astro-ph.CO]

Pith/arXiv arXiv 2017
[13]

Sharma and J

A. Sharma and J. Harms, Phys. Rev. D102, 063009 (2020), arXiv:2006.16116 [gr-qc]

Pith/arXiv arXiv 2020
[14]

B. Zhou, L. Reali, E. Berti, M. ¸ calı¸ skan, C. Creque- Sarbinowski, M. Kamionkowski, and B. S. Sathyaprakash, Phys. Rev. D108, 064040 (2023), arXiv:2209.01310 [gr-qc]

Pith/arXiv arXiv 2023
[15]

Zhong, B

H. Zhong, B. Zhou, L. Reali, E. Berti, and V. Mandic, Phys. Rev. D110, 064047 (2024), arXiv:2406.10757 [gr- qc]

Pith/arXiv arXiv 2024
[16]

Biscoveanu, C

S. Biscoveanu, C. Talbot, E. Thrane, and R. Smith, Phys. Rev. Lett.125, 241101 (2020), arXiv:2009.04418 [astro-ph.HE]

Pith/arXiv arXiv 2020
[17]

Einsle, M.-A

H. Einsle, M.-A. Bizouard, T. Regimbau, and M. Sakellariadou, Phys. Rev. D112, 063056 (2025), arXiv:2506.14764 [gr-qc]

Pith/arXiv arXiv 2025
[18]

for a full description). This produces a character- istic energy density spectrum Ω astro(f) = Ω α(f /fref )α whereα= 2/3 for frequencies up to∼100 Hz, reflect- ing inspiral-phase emission and the full contribution of the source population. Above this frequency, the spec- trum flattens due to the reduced contribution from high- mass and high-redshift syst...

2030
[19]

Tania and S

R. Tania and S. Jishnu, 10.48550/arXiv.2506.12237, arXiv:2506.12237 [gr-qc]

doi:10.48550/arxiv.2506.12237
[20]

Auclair et al., JCAP04, 034 (2020), arXiv:1909.00819 [astro-ph.CO]

P. Auclair et al., JCAP04, 034 (2020), arXiv:1909.00819 [astro-ph.CO]

Pith/arXiv arXiv 2020
[21]

P. A. R. Ade et al. (BICEP, Keck), Phys. Rev. Lett.127, 151301 (2021), arXiv:2110.00483 [astro-ph.CO]

arXiv 2021
[22]

A. G. Abac et al. (LIGO Scientific, Virgo, KAGRA), (2025), arXiv:2508.20721 [gr-qc]

Pith/arXiv arXiv 2025
[23]

Soni et al

S. Soni et al. (LIGO Scientific), Class. Quant. Grav.42, 085016 (2025), arXiv:2409.02831 [astro-ph.IM]

arXiv 2025
[24]

Allen, J

B. Allen, J. D. E. Creighton, E. E. Flanagan, and J. D. Romano, Phys. Rev. D67, 122002 (2003)

2003
[25]

Thrane and J

E. Thrane and J. D. Romano, Phys. Rev. D88, 124032 (2013), arXiv:1310.5300 [astro-ph.IM]

Pith/arXiv arXiv 2013
[26]

A. I. Renzini, A. Romero-Rodr ´ ıguez, C. Talbot, M. Lalle- man, S. Kandhasamy, K. Turbang, S. Biscoveanu, K. Martinovic, P. Meyers, L. Tsukada, K. Janssens, D. Davis, A. Matas, P. Charlton, G.-C. Liu, I. Dvorkin, S. Banagiri, S. Bose, T. Callister, F. De Lillo, L. D’Onofrio, F. Garufi, G. Harry, J. Lawrence, V. Mandic, A. Macquet, I. Michaloliakos, S. Mi...

2023

[1] [1]

Christensen, Rept

N. Christensen, Rept. Prog. Phys.82, 016903 (2019), arXiv:1811.08797 [gr-qc]

Pith/arXiv arXiv 2019

[2] [2]

Caprini and D

C. Caprini and D. G. Figueroa, Class. Quant. Grav.35, 163001 (2018), arXiv:1801.04268 [astro-ph.CO]

Pith/arXiv arXiv 2018

[3] [3]

Regimbau, Res

T. Regimbau, Res. Astron. Astrophys.11, 369 (2011), arXiv:1101.2762 [astro-ph.CO]

Pith/arXiv arXiv 2011

[4] [4]

B. P. Abbott et al. (KAGRA, LIGO Scientific, Virgo), Living Rev. Rel.19, 1 (2016), arXiv:1304.0670 [gr-qc]

Pith/arXiv arXiv 2016

[5] [5]

A. G. Abac et al. (LIGO Scientific, VIRGO, KAGRA), (2025), arXiv:2510.26848 [gr-qc]

arXiv 2025

[6] [6]

(2026), arXiv:2605.27225 [gr-qc]

Pith/arXiv arXiv 2026

[7] [8]

Allen and J

B. Allen and J. D. Romano, Phys. Rev. D59, 102001 (1999), arXiv:gr-qc/9710117 [gr-qc]

Pith/arXiv arXiv 1999

[8] [9]

P. M. Meyers, K. Martinovic, N. Christensen, and M. Sakellariadou, Phys. Rev. D102, 102005 (2020), arXiv:2008.00789 [gr-qc]

Pith/arXiv arXiv 2020

[9] [10]

Martinovic, P

K. Martinovic, P. M. Meyers, M. Sakellariadou, and N. Christensen, Phys. Rev. D103, 043023 (2021), arXiv:2011.05697 [gr-qc]

Pith/arXiv arXiv 2021

[10] [11]

Aasi et al

J. Aasi et al. (LIGO Scientific), Class. Quant. Grav.32, 074001 (2015), arXiv:1411.4547 [gr-qc]

Pith/arXiv arXiv 2015

[11] [12]

Regimbau, M

T. Regimbau, M. Evans, N. Christensen, E. Katsavouni- dis, B. Sathyaprakash, and S. Vitale, Phys. Rev. Lett. 118, 151105 (2017), arXiv:1611.08943 [astro-ph.CO]

Pith/arXiv arXiv 2017

[12] [13]

Sharma and J

A. Sharma and J. Harms, Phys. Rev. D102, 063009 (2020), arXiv:2006.16116 [gr-qc]

Pith/arXiv arXiv 2020

[13] [14]

B. Zhou, L. Reali, E. Berti, M. ¸ calı¸ skan, C. Creque- Sarbinowski, M. Kamionkowski, and B. S. Sathyaprakash, Phys. Rev. D108, 064040 (2023), arXiv:2209.01310 [gr-qc]

Pith/arXiv arXiv 2023

[14] [15]

Zhong, B

H. Zhong, B. Zhou, L. Reali, E. Berti, and V. Mandic, Phys. Rev. D110, 064047 (2024), arXiv:2406.10757 [gr- qc]

Pith/arXiv arXiv 2024

[15] [16]

Biscoveanu, C

S. Biscoveanu, C. Talbot, E. Thrane, and R. Smith, Phys. Rev. Lett.125, 241101 (2020), arXiv:2009.04418 [astro-ph.HE]

Pith/arXiv arXiv 2020

[16] [17]

Einsle, M.-A

H. Einsle, M.-A. Bizouard, T. Regimbau, and M. Sakellariadou, Phys. Rev. D112, 063056 (2025), arXiv:2506.14764 [gr-qc]

Pith/arXiv arXiv 2025

[17] [18]

for a full description). This produces a character- istic energy density spectrum Ω astro(f) = Ω α(f /fref )α whereα= 2/3 for frequencies up to∼100 Hz, reflect- ing inspiral-phase emission and the full contribution of the source population. Above this frequency, the spec- trum flattens due to the reduced contribution from high- mass and high-redshift syst...

2030

[18] [19]

Tania and S

R. Tania and S. Jishnu, 10.48550/arXiv.2506.12237, arXiv:2506.12237 [gr-qc]

doi:10.48550/arxiv.2506.12237

[19] [20]

Auclair et al., JCAP04, 034 (2020), arXiv:1909.00819 [astro-ph.CO]

P. Auclair et al., JCAP04, 034 (2020), arXiv:1909.00819 [astro-ph.CO]

Pith/arXiv arXiv 2020

[20] [21]

P. A. R. Ade et al. (BICEP, Keck), Phys. Rev. Lett.127, 151301 (2021), arXiv:2110.00483 [astro-ph.CO]

arXiv 2021

[21] [22]

A. G. Abac et al. (LIGO Scientific, Virgo, KAGRA), (2025), arXiv:2508.20721 [gr-qc]

Pith/arXiv arXiv 2025

[22] [23]

Soni et al

S. Soni et al. (LIGO Scientific), Class. Quant. Grav.42, 085016 (2025), arXiv:2409.02831 [astro-ph.IM]

arXiv 2025

[23] [24]

Allen, J

B. Allen, J. D. E. Creighton, E. E. Flanagan, and J. D. Romano, Phys. Rev. D67, 122002 (2003)

2003

[24] [25]

Thrane and J

E. Thrane and J. D. Romano, Phys. Rev. D88, 124032 (2013), arXiv:1310.5300 [astro-ph.IM]

Pith/arXiv arXiv 2013

[25] [26]

A. I. Renzini, A. Romero-Rodr ´ ıguez, C. Talbot, M. Lalle- man, S. Kandhasamy, K. Turbang, S. Biscoveanu, K. Martinovic, P. Meyers, L. Tsukada, K. Janssens, D. Davis, A. Matas, P. Charlton, G.-C. Liu, I. Dvorkin, S. Banagiri, S. Bose, T. Callister, F. De Lillo, L. D’Onofrio, F. Garufi, G. Harry, J. Lawrence, V. Mandic, A. Macquet, I. Michaloliakos, S. Mi...

2023