Pith. sign in

REVIEW 4 major objections 4 minor 4 cited by

ASPIRE reuses existing posterior samples to produce unbiased new-model results without a full re-run.

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

2026-08-03 23:43 UTC pith:MFPV57FF

load-bearing objection Sound method, over-claimed guarantee: the reuse-specific validation is thinner than the abstract suggests. the 4 major comments →

arxiv 2511.04218 v2 pith:MFPV57FF submitted 2025-11-06 hep-ex astro-ph.IMgr-qc

Accelerated Sequential Posterior Inference via Reuse for Gravitational-Wave Analyses

Michael J. Williams This is my paper
classification hep-ex astro-ph.IMgr-qc
keywords Bayesian inferencesequential Monte Carlonormalizing flowsgravitational wavesposterior reusewaveform systematicsmodel comparisonlow-latency reanalysis
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.

ASPIRE is a method for Bayesian inference that takes the posterior samples and evidence estimate already computed under one model and transforms them into unbiased results under a different model, without restarting from the prior. It does this by training a normalizing flow on the old samples, extending the flow to cover any new parameters, and then using sequential Monte Carlo to anneal from the flow to the new posterior. The paper reports 4-10x reductions in likelihood evaluations and wall time while producing posteriors and evidences consistent with full re-analysis. This matters because gravitational-wave events are repeatedly reanalyzed under different waveform models and physical hypotheses, and the approach could turn those costly reanalyses into fast updates.

Core claim

ASPIRE starts from posterior samples obtained under a model M1, approximates their density with a normalizing flow q_phi, replaces samples of parameters incompatible with the new model M2 with draws from the prior, and then evolves this flow into the target posterior using a generalized SMC scheme. The SMC interpolates through the tempered distribution q_phi^(1-beta) [p(d|theta,M2) p(theta|M2)]^beta, with adaptive beta and a t-preconditioned Crank-Nicolson MCMC kernel for diversification. The paper claims this yields posterior samples and Bayesian evidence estimates that are statistically indistinguishable from a traditional full analysis, demonstrated across waveform-model switches, additio

What carries the argument

The load-bearing object is the SMC bridge defined by Eq. (2), an annealed path between the flow approximation q_phi and the target posterior under M2. Starting from a flow fitted to existing posterior samples means the sampler begins near the old answer and only pays for the difference between models. The replacement rule for incompatible parameters (drawing from the prior) extends the flow's support to the new parameter space, and the t-preconditioned Crank-Nicolson kernel performs the within-SMC MCMC diversification.

Load-bearing premise

The initial flow, after replacing incompatible parameters with prior draws, must assign non-negligible density to every region where the new model's posterior has any significant mass; otherwise the sequential Monte Carlo estimates will be biased no matter how long they run.

What would settle it

Construct a target model whose posterior has two well-separated modes, with the initial flow trained on a posterior that only covers one mode. Run ASPIRE to estimate the evidence for the target and compare it to a direct nested-sampling estimate; if the log-evidence difference exceeds the quoted uncertainty while the flow's density in the uncovered mode is negligible, the central claim fails.

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

If this is right

  • Waveform-systematics studies could be run by updating an established analysis to each alternative waveform model rather than rerunning from the prior each time.
  • Adding new physical effects such as spin precession or orbital eccentricity becomes a cheap incremental operation, enabling catalog-scale updates as models improve.
  • The evidence estimate comes out as a by-product, so model comparison between old and new models does not require a separate nested-sampling run.
  • The methodology transfers to other scientific domains where reanalysis under competing models is common, such as particle-physics reinterpretation or cosmological model comparison.

Where Pith is reading between the lines

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

  • The method's unbiasedness hinges on the flow having non-negligible density wherever the new model's likelihood is large; for models that are genuinely very different, the SMC path may become long or degenerate, so the 4-10x gain is probably not universal.
  • The replacement rule for incompatible parameters is heuristic; a more formal prior-flow over the new parameters would make the support condition easier to verify and could remove a possible failure mode.
  • Because SMC is naturally parallelizable, the approach is well suited to GPU-based, low-latency reanalysis pipelines, where full nested sampling is too slow.
  • The real-data demonstration uses an event where the two waveform models broadly agree; a deliberate stress test on an event where the models disagree strongly would sharpen the validation of the method's claims.

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

4 major / 4 minor

Summary. The paper introduces ASPIRE, a method for reusing posterior samples and Bayesian evidence estimates from one model (M1) to accelerate Bayesian inference under another model (M2). ASPIRE trains a normalizing flow on the M1 posterior samples, extends the flow to accommodate new or incompatible parameters by adding prior draws, and runs a sequential Monte Carlo (SMC) path that interpolates between the flow approximation and the M2 posterior. The authors demonstrate on simulated and real gravitational-wave data that ASPIRE reproduces baseline dynesty posteriors and evidences at roughly 4-10x lower computational cost, and they report a 100-event P-P test for statistical unbiasedness. The central claim is that ASPIRE yields unbiased posterior samples and evidence estimates without rerunning the original analysis.

Significance. If the unbiasedness claim holds, ASPIRE addresses a genuine bottleneck in gravitational-wave astronomy—repeated reanalyses of events under different waveform models or physical hypotheses—and could generalize to other Bayesian reanalysis settings. The paper ships open-source code and a data-release link, which is a strength, and the validation covers several practically relevant cases: waveform switching, spin precession, orbital eccentricity, and real data with calibration parameters. However, the strongest claim (guaranteed unbiased results in the reuse setting) rests on a support/coverage condition that is neither proved nor directly tested, and the existing P-P test does not exercise the reuse scenario. The contribution is potentially valuable, but the central statistical guarantee needs additional support.

major comments (4)
  1. [Eq. (2) and Method] The statement 'This approach guarantees unbiased results' is not supported. In the SMC path pt ∝ q_phi^{1-beta} [p(d|theta,M2)p(theta|M2)]^{beta}, the intermediate distributions have support contained in the support of q_phi for beta<1. If q_phi has very low or zero density in regions where the M2 posterior has substantial mass, the sampler cannot transport particles into those regions. The Supplemental replacement rule only adds prior draws for incompatible parameters; for shared parameters, q_phi is trained on M1 posterior samples and may not cover a shifted M2 posterior. No support condition, weight-ratio bound, or effective-sample-size diagnostic is reported. A proof of unbiasedness under explicit assumptions, or a diagnostic demonstrating adequate coverage (e.g., importance weights or ESS in the relevant region), is needed to justify the guarantee.
  2. [Fig. 5 and Results] The P-P test initializes ASPIRE from prior samples, not from a previous model's posterior, so it does not validate the reuse setting that motivates the paper. The only reuse validations are JSD comparisons on five examples (Figs. 2-4); the threshold D_JS <= 1.5 mnats is a practical heuristic, not a formal statistical test. A P-P test with ASPIRE initialized from a previous model's posterior, or a similar coverage test tailored to the reuse scenario, is necessary to substantiate the central unbiasedness claim.
  3. [Table I] For the q=4 waveform-switch case, the log-evidence difference between ASPIRE (-54803.08 +/- 0.09) and dynesty (-54802.73 +/- 0.21) is about 1.5 sigma combined, yet the text says only that 'they differ slightly.' Since evidence estimates are central to model comparison, this discrepancy needs a quantitative treatment—e.g., repeated runs, an increased number of SMC particles, or an analysis of whether the difference is consistent with sampling error. As presented, it weakens the claim that ASPIRE provides 'consistent evidence estimates' in the case where the posterior shifts most.
  4. [Abstract vs. Results] The Abstract states that ASPIRE reduces 'total likelihood evaluations and wall times by factors of up to 5.8 and 5.5', while the Results report per-sample likelihood-evaluation reductions of 7x, 6x, 10x, 7x, and 4x, and per-sample wall-time reductions of 6x, 5x, 10x, 6x, and 7x. These numbers are inconsistent or at least ambiguous. Please clarify whether the Abstract refers to total or per-sample quantities and make the reporting uniform.
minor comments (4)
  1. [Table I] The table header 'log-Bayesian evidences (basee)' contains a typo; should be 'natural log' or 'base e'.
  2. [Supplemental Material] The Supplemental says it provides 'additional figures, derivations, and experimental details,' but the derivations section is absent from the included text. If a proof of unbiasedness exists, it should be included; otherwise remove the word 'derivations.'
  3. [Fig. 5] The caption says 'Shaded regions indicate 1-, 2-, and 3-sigma confidence intervals,' but the reader must infer which shaded bands correspond to which significance; please label them explicitly.
  4. [Methods] The sentence 'This approach guarantees unbiased results, even when the target and initial posteriors occupy different regions of the parameter space' is repeated in the Discussion. This is a strong claim and should be qualified by the assumptions under which it is true, or replaced by a more precise statement.

Circularity Check

0 steps flagged

No circularity: the SMC target at beta=1 is the true M2 posterior, independent of the fitted flow proposal; self-citations are methodological and not load-bearing.

full rationale

The derivation is self-contained. Equation (2) defines the annealed target p_t ∝ q_phi^{1−beta_t} [p(d|theta,M2)p(theta|M2)]^{beta_t}; at beta_t=1 the q_phi factor vanishes, so the target is exactly the M2 posterior, and the evidence estimator follows from the SMC normalizing constant. Thus the fitted normalizing flow is a proposal distribution, not the output, and no fitted parameter is renamed as a prediction. The self-citations [22,23,36] provide prior context, implementation details, or earlier methodological validation, but the unbiasedness claim rests on the standard SMC reweighting identity and is checked against independent dynesty baseline runs and real-data comparisons. There is a genuine validation gap: the P-P test (Fig. 5) is initialized from prior samples rather than from a previous model's posterior, so it does not directly validate the reuse setting; and the q=4 evidence discrepancy in Table I is slightly larger than the quoted uncertainties. However, these are correctness/validation concerns, not circularity: no step of the claimed chain reduces to its inputs by construction. The central claim is therefore not circular.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 1 invented entities

No physical constants are fitted; the only parameters are algorithmic choices (flow training, SMC configuration, and the parameter-replacement rule), which affect efficiency and robustness but not the target posterior. The statistical guarantee relies on standard SMC consistency plus a support assumption on the flow proposal.

free parameters (3)
  • incompatible-parameter replacement rule = polarization angle, phase, and (for high mass ratio) chirp mass and mass ratio replaced with prior draws; aligned-spin s
    Hand-chosen protocol for extending M1 samples to M2; if chosen poorly the proposal may not cover the target (Supplemental, 'Analysis details').
  • normalizing-flow training hyperparameters = initial learning rate 0.001, cosine decay, batch size 500; architecture not fully specified
    Hand-selected; affect flow fidelity and hence SMC efficiency, but not statistical consistency (Supplemental, 'Implementation details').
  • SMC sampler configuration = not reported (number of particles, MCMC steps, adaptation schedule)
    Choices affect the accuracy/cost tradeoff; absence makes replication harder (Supplemental, 'Implementation details').
axioms (4)
  • standard math Tempered SMC samplers produce consistent posterior and evidence estimates when resampling and invariant MCMC kernels are used at each temperature step.
    Invoked by Eq. (2) and the claim 'this approach guarantees unbiased results'; the paper relies on standard SMC consistency without proving the conditions.
  • domain assumption The flow approximation q_phi, extended by prior draws for new or incompatible parameters, has support covering the target posterior p(theta|d,M2).
    Required for the SMC weights to be valid; the Supplemental describes the replacement rule but does not prove the support condition.
  • domain assumption The t-preconditioned Crank-Nicolson MCMC kernel leaves each tempered target invariant and mixes well enough to avoid particle degeneracy.
    Named in the Supplemental as the diversification kernel; its mixing properties in the new-parameter directions are not demonstrated.
  • domain assumption The supplied likelihood and prior for model M2 correctly describe the new model and data.
    All results depend on the waveform models and priors being implemented faithfully (Results section).
invented entities (1)
  • None no independent evidence
    purpose: No new physical entities are introduced.
    The normalizing flow is a statistical tool, not a new physical entity; no new parameters, forces, or dimensions are postulated.

pith-pipeline@v1.3.0-alltime-deepseek · 13160 in / 14103 out tokens · 138908 ms · 2026-08-03T23:43:34.227638+00:00 · methodology

0 comments
read the original abstract

We introduce Accelerated Sequential Posterior Inference via Reuse (ASPIRE), a broadly applicable framework that transforms existing posterior samples and Bayesian evidence estimates into unbiased results under alternative models without rerunning the original analysis. ASPIRE combines normalizing flows with a generalized Sequential Monte Carlo (SMC) scheme, enabling efficient updates of existing results and reducing total likelihood evaluations and wall times by factors of up to 5.8 and 5.5, respectively, with larger gains per posterior sample. This addresses a growing problem in gravitational-wave astronomy, where events must be repeatedly reanalyzed under different models or physical hypotheses. We show that ASPIRE reproduces full Bayesian results when switching waveform models or adding physical effects such as spin precession and orbital eccentricity. With this statistical robustness, ASPIRE turns repeated reanalyses into fast, reliable updates-paving the way for systematic studies of waveform systematics, scalable reanalyses across large event catalogs, and broadly applicable Bayesian reanalysis across other scientific domains.

Figures

Figures reproduced from arXiv: 2511.04218 by Michael J. Williams.

Figure 1
Figure 1. Figure 1: FIG. 1. Comparison of traditional Bayesian inference (top), [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Posterior samples obtained from analyz [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Posterior distributions for GW150914 with chirp mass [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Posterior samples obtained when adding new [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Probability–probability (P–P) plot for 100 simulated [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Ab Initio Real-Time Gravitational-Wave Parameter Estimation

    gr-qc 2026-07 accept novelty 6.0

    Slice-within-Gibbs nested sampling on modern GPUs delivers well-calibrated BNS parameter estimation in ~12 minutes uncompressed and ~89 seconds with heterodyning, from cold priors.

  2. Fortifying gravitational-wave population inference with normalizing flows

    astro-ph.HE 2026-06 conditional novelty 6.0

    Representing each gravitational-wave event's posterior with a normalizing flow lets analysts generate enough cheap posterior samples to keep the Monte-Carlo variance of population inference below threshold for catalog...

  3. Reconsidering the consistent use of precessing, higher order multipole models for gravitational wave analyses

    gr-qc 2026-01 conditional novelty 6.0

    A matched-filter SNR threshold on precession and higher-multipole content can select cheaper waveform models per event while keeping inferred black-hole mass and spin populations consistent, cutting analysis cost by t...

  4. The Sequential Monte Carlo goes NUTS: Boosting Gravitational-Wave Inference

    gr-qc 2026-01 conditional novelty 6.0

    SHARPy uses Sequential Monte Carlo with a No-U-Turn sampler in JAX to estimate gravitational-wave posteriors and evidence for binary black holes in about ten minutes.

Reference graph

Works this paper leans on

99 extracted references · 62 linked inside Pith · cited by 4 Pith papers

  1. [1]

    Aasiet al.(LIGO Scientific), Advanced LIGO, Class

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

    Pith/arXiv arXiv 2015

  2. [2]

    Acerneseet al.(VIRGO), Advanced Virgo: a second- generation interferometric gravitational wave detector, Class

    F. Acerneseet al.(VIRGO), Advanced Virgo: a second- generation interferometric gravitational wave detector, Class. Quant. Grav.32, 024001 (2015), arXiv:1408.3978 [gr-qc]

    Pith/arXiv arXiv 2015

  3. [3]

    Akutsuet al.(KAGRA), Overview of KAGRA: Detec- tor design and construction history, PTEP2021, 05A101 (2021), arXiv:2005.05574 [physics.ins-det]

    T. Akutsuet al.(KAGRA), Overview of KAGRA: Detec- tor design and construction history, PTEP2021, 05A101 (2021), arXiv:2005.05574 [physics.ins-det]

    arXiv 2021

  4. [4]

    B. P. Abbottet al.(LIGO Scientific, Virgo), GWTC- 1: A Gravitational-Wave Transient Catalog of Compact Binary Mergers Observed by LIGO and Virgo during the First and Second Observing Runs, Phys. Rev. X9, 031040 (2019), arXiv:1811.12907 [astro-ph.HE]

    Pith/arXiv arXiv 2019

  5. [5]

    Abbottet al.(LIGO Scientific, Virgo), GWTC-2: Compact Binary Coalescences Observed by LIGO and Virgo During the First Half of the Third Observing Run, Phys

    R. Abbottet al.(LIGO Scientific, Virgo), GWTC-2: Compact Binary Coalescences Observed by LIGO and Virgo During the First Half of the Third Observing Run, Phys. Rev. X11, 021053 (2021), arXiv:2010.14527 [gr- qc]

    Pith/arXiv arXiv 2021

  6. [6]

    R. Abbottet al.(LIGO Scientific, VIRGO), GWTC-2.1: Deep extended catalog of compact binary coalescences observed by LIGO and Virgo during the first half of the third observing run, Phys. Rev. D109, 022001 (2024), arXiv:2108.01045 [gr-qc]

    Pith/arXiv arXiv 2024

  7. [7]

    Abbottet al.(KAGRA, VIRGO, LIGO Scien- tific), GWTC-3: Compact Binary Coalescences Observed by LIGO and Virgo during the Second Part of the Third Observing Run, Phys

    R. Abbottet al.(KAGRA, VIRGO, LIGO Scien- tific), GWTC-3: Compact Binary Coalescences Observed by LIGO and Virgo during the Second Part of the Third Observing Run, Phys. Rev. X13, 041039 (2023), arXiv:2111.03606 [gr-qc]

    Pith/arXiv arXiv 2023

  8. [8]

    GWTC-4.0: Updating the Gravitational-Wave Tran- sient Catalog with Observations from the First Part of the Fourth LIGO-Virgo-KAGRA Observing Run (2025), arXiv:2508.18082 [gr-qc]

    Pith/arXiv arXiv 2025

  9. [9]

    L. S. Collaboration, LIGO-G2302098: LIGO-Virgo- KAGRA Cumulative Detection plot - O1-O4b,https:// dcc.ligo.org/LIGO-G2302098-v15/public(2025), ac- cessed: July 7, 2025

    2025

  10. [10]

    Abbottet al.(LIGO Scientific, Virgo, KA- GRA), Constraints on the Cosmic Expansion His- tory from GWTC–3, Astrophys

    R. Abbottet al.(LIGO Scientific, Virgo, KA- GRA), Constraints on the Cosmic Expansion His- tory from GWTC–3, Astrophys. J.949, 76 (2023), arXiv:2111.03604 [astro-ph.CO]

    Pith/arXiv arXiv 2023

  11. [11]

    GWTC-4.0: Constraints on the Cosmic Expansion Rate and Modified Gravitational-wave Propagation (2025), arXiv:2509.04348 [astro-ph.CO]

    Pith/arXiv arXiv 2025

  12. [12]

    Gamba, M

    R. Gamba, M. Breschi, S. Bernuzzi, M. Agathos, and A. Nagar, Waveform systematics in the gravitational- wave inference of tidal parameters and equation of state from binary neutron star signals, Phys. Rev. D103, 124015 (2021), arXiv:2009.08467 [gr-qc]

    Pith/arXiv arXiv 2021

  13. [13]

    A. B. Yelikar, R. O. Shaughnessy, J. Lange, and A. Z. Jan, Waveform systematics in gravitational-wave infer- ence of signals from binary neutron star merger mod- 6 els incorporating higher-order modes information, Phys. Rev. D110, 064024 (2024), arXiv:2404.16599 [gr-qc]

    Pith/arXiv arXiv 2024

  14. [14]

    Gamboaet al., Accurate waveforms for eccen- tric, aligned-spin binary black holes: The multipo- lar effective-one-body model SEOBNRv5EHM (2024), arXiv:2412.12823 [gr-qc]

    A. Gamboaet al., Accurate waveforms for eccen- tric, aligned-spin binary black holes: The multipo- lar effective-one-body model SEOBNRv5EHM (2024), arXiv:2412.12823 [gr-qc]

    Pith/arXiv arXiv 2024

  15. [15]

    Nagar, R

    A. Nagar, R. Gamba, P. Rettegno, V. Fantini, and S. Bernuzzi, Effective-one-body waveform model for non- circularized, planar, coalescing black hole binaries: The importance of radiation reaction, Phys. Rev. D110, 084001 (2024), arXiv:2404.05288 [gr-qc]

    Pith/arXiv arXiv 2024

  16. [16]

    N. V. Krishnendu and F. Ohme, Testing General Rela- tivity with Gravitational Waves: An Overview, Universe 7, 497 (2021), arXiv:2201.05418 [gr-qc]

    Pith/arXiv arXiv 2021

  17. [17]

    Abbottet al.(LIGO Scientific, VIRGO), Search for Lensing Signatures in the Gravitational-Wave Observa- tions from the First Half of LIGO–Virgo’s Third Observ- ing Run, Astrophys

    R. Abbottet al.(LIGO Scientific, VIRGO), Search for Lensing Signatures in the Gravitational-Wave Observa- tions from the First Half of LIGO–Virgo’s Third Observ- ing Run, Astrophys. J.923, 14 (2021), arXiv:2105.06384 [gr-qc]

    Pith/arXiv arXiv 2021

  18. [18]

    Skilling, Nested Sampling, AIP Conf

    J. Skilling, Nested Sampling, AIP Conf. Proc.735, 395 (2004)

    2004

  19. [19]

    Skilling, Nested sampling for general Bayesian compu- tation, Bayesian Analysis1, 833 (2006)

    J. Skilling, Nested sampling for general Bayesian compu- tation, Bayesian Analysis1, 833 (2006)

    2006

  20. [20]

    Veitchet al., Parameter estimation for compact bina- ries with ground-based gravitational-wave observations using the LALInference software library, Phys

    J. Veitchet al., Parameter estimation for compact bina- ries with ground-based gravitational-wave observations using the LALInference software library, Phys. Rev. D 91, 042003 (2015), arXiv:1409.7215 [gr-qc]

    Pith/arXiv arXiv 2015

  21. [21]

    I. M. Romero-Shawet al., Bayesian inference for compact binary coalescences with bilby: validation and applica- tion to the first LIGO–Virgo gravitational-wave transient catalogue, Mon. Not. Roy. Astron. Soc.499, 3295 (2020), arXiv:2006.00714 [astro-ph.IM]

    Pith/arXiv arXiv 2020

  22. [22]

    M. J. Williams, J. Veitch, and C. Messenger, Nested sam- pling with normalizing flows for gravitational-wave infer- ence, Phys. Rev. D103, 103006 (2021), arXiv:2102.11056 [gr-qc]

    Pith/arXiv arXiv 2021

  23. [23]

    M. J. Williams, J. Veitch, and C. Messenger, Importance nested sampling with normalising flows, Mach. Learn. Sci. Tech.4, 035011 (2023), arXiv:2302.08526 [astro- ph.IM]

    Pith/arXiv arXiv 2023

  24. [24]

    A. J. K. Chua and M. Vallisneri, Learning Bayesian posteriors with neural networks for gravitational- wave inference, Phys. Rev. Lett.124, 041102 (2020), arXiv:1909.05966 [gr-qc]

    Pith/arXiv arXiv 2020

  25. [25]

    Gabbard, C

    H. Gabbard, C. Messenger, I. S. Heng, F. Tono- lini, and R. Murray-Smith, Bayesian parameter esti- mation using conditional variational autoencoders for gravitational-wave astronomy, Nature Phys.18, 112 (2022), arXiv:1909.06296 [astro-ph.IM]

    Pith/arXiv arXiv 2022

  26. [26]

    M. Dax, S. R. Green, J. Gair, M. P¨ urrer, J. Wildberger, J. H. Macke, A. Buonanno, and B. Sch¨ olkopf, Neural Im- portance Sampling for Rapid and Reliable Gravitational- Wave Inference, Phys. Rev. Lett.130, 171403 (2023), arXiv:2210.05686 [gr-qc]

    Pith/arXiv arXiv 2023

  27. [27]

    Fairhurst, C

    S. Fairhurst, C. Hoy, R. Green, C. Mills, and S. A. Us- man, Simple parameter estimation using observable fea- tures of gravitational-wave signals, Phys. Rev. D108, 082006 (2023), arXiv:2304.03731 [gr-qc]

    Pith/arXiv arXiv 2023

  28. [28]

    D. W. Hogg and D. Foreman-Mackey, Data analysis recipes: Using Markov Chain Monte Carlo, Astrophys. J. Suppl.236, 11 (2018), arXiv:1710.06068 [astro-ph.IM]

    Pith/arXiv arXiv 2018

  29. [29]

    Lange, R

    J. Lange, R. O’Shaughnessy, and M. Rizzo, Rapid and accurate parameter inference for coalescing, precessing compact binaries (2018), arXiv:1805.10457 [gr-qc]

    Pith/arXiv arXiv 2018

  30. [30]

    N. E. Wolfe, C. Talbot, and J. Golomb, Accelerating tests of general relativity with gravitational-wave signals us- ing hybrid sampling, Phys. Rev. D107, 104056 (2023), arXiv:2208.12872 [gr-qc]

    Pith/arXiv arXiv 2023

  31. [31]

    Prathaban, H

    M. Prathaban, H. Bevins, and W. Handley, Acceler- ated nested sampling withβ-flows for gravitational waves (2024), arXiv:2411.17663 [astro-ph.IM]

    arXiv 2024

  32. [32]

    P. C. Peters and J. Mathews, Gravitational radiation from point masses in a Keplerian orbit, Phys. Rev.131, 435 (1963)

    1963

  33. [33]

    T. A. Apostolatos, C. Cutler, G. J. Sussman, and K. S. Thorne, Spin induced orbital precession and its modula- tion of the gravitational wave forms from merging bina- ries, Phys. Rev. D49, 6274 (1994)

    1994

  34. [34]

    Papamakarios, E

    G. Papamakarios, E. Nalisnick, D. J. Rezende, S. Mo- hamed, and B. Lakshminarayanan, Normalizing Flows for Probabilistic Modeling and Inference, J. Ma- chine Learning Res.22, 2617 (2021), arXiv:1912.02762 [stat.ML]

    Pith/arXiv arXiv 2021

  35. [35]

    See Supplemental Material at [URL will be inserted by publisher] for additional figures, derivations, and experi- mental details

  36. [36]

    M. J. Williams, M. Karamanis, Y. Luo, and U. Sel- jak, Validating Sequential Monte Carlo for Gravitational- Wave Inference (2025), arXiv:2506.18977 [astro-ph.IM]

    arXiv 2025

  37. [37]

    Bernardo, M

    J. Bernardo, M. Bayarri, J. Berger, A. Dawid, D. Heck- erman, A. Smith, M. West, P. Del Moral, A. Doucet, and A. Jasra, Sequential monte carlo for bayesian computa- tion, Bayesian statistics8, 1 (2011)

    2011

  38. [38]

    Del Moral, A

    P. Del Moral, A. Doucet, and A. Jasra, An adaptive se- quential monte carlo method for approximate bayesian computation, Statistics and computing22, 1009 (2012)

    2012

  39. [39]

    Fearnhead and B

    P. Fearnhead and B. M. Taylor, An Adaptive Sequential Monte Carlo Sampler, Bayesian Analysis8, 411 (2013)

    2013

  40. [40]

    M. J. Williams and KonstantinLeyde, mj-will/aspire: v0.1.0a2 (2025)

    2025

  41. [41]

    Prattenet al., Computationally efficient models for the dominant and subdominant harmonic modes of pre- cessing binary black holes, Phys

    G. Prattenet al., Computationally efficient models for the dominant and subdominant harmonic modes of pre- cessing binary black holes, Phys. Rev. D103, 104056 (2021), arXiv:2004.06503 [gr-qc]

    Pith/arXiv arXiv 2021

  42. [42]

    J. E. Thompson, E. Hamilton, L. London, S. Ghosh, P. Kolitsidou, C. Hoy, and M. Hannam, PhenomXO4a: a phenomenological gravitational-wave model for pre- cessing black-hole binaries with higher multipoles and asymmetries, Phys. Rev. D109, 063012 (2024), arXiv:2312.10025 [gr-qc]

    Pith/arXiv arXiv 2024

  43. [43]

    S. Husa, S. Khan, M. Hannam, M. P¨ urrer, F. Ohme, X. Jim´ enez Forteza, and A. Boh´ e, Frequency-domain gravitational waves from nonprecessing black-hole bina- ries. I. New numerical waveforms and anatomy of the sig- nal, Phys. Rev. D93, 044006 (2016), arXiv:1508.07250 [gr-qc]

    Pith/arXiv arXiv 2016

  44. [44]

    S. Khan, S. Husa, M. Hannam, F. Ohme, M. P¨ urrer, X. Jim´ enez Forteza, and A. Boh´ e, Frequency-domain gravitational waves from nonprecessing black-hole bi- naries. II. A phenomenological model for the ad- vanced detector era, Phys. Rev. D93, 044007 (2016), arXiv:1508.07253 [gr-qc]

    Pith/arXiv arXiv 2016

  45. [45]

    Schmidt, M

    P. Schmidt, M. Hannam, and S. Husa, Towards models of gravitational waveforms from generic binaries: A sim- ple approximate mapping between precessing and non- precessing inspiral signals, Phys. Rev. D86, 104063 (2012), arXiv:1207.3088 [gr-qc]

    Pith/arXiv arXiv 2012

  46. [46]

    Hannam, P

    M. Hannam, P. Schmidt, A. Boh´ e, L. Haegel, S. Husa, 7 F. Ohme, G. Pratten, and M. P¨ urrer, Simple Model of Complete Precessing Black-Hole-Binary Gravitational Waveforms, Phys. Rev. Lett.113, 151101 (2014), arXiv:1308.3271 [gr-qc]

    Pith/arXiv arXiv 2014

  47. [47]

    Morisaki, R

    S. Morisaki, R. Smith, L. Tsukada, S. Sachdev, S. Steven- son, C. Talbot, and A. Zimmerman, Rapid localiza- tion and inference on compact binary coalescences with the Advanced LIGO-Virgo-KAGRA gravitational-wave detector network, Phys. Rev. D108, 123040 (2023), arXiv:2307.13380 [gr-qc]

    Pith/arXiv arXiv 2023

  48. [48]

    However, incorporating eccentricity typi- cally requires reanalyzing existing events, which can be computationally expensive

    and several observed events show potential evidence for it [49–52]. However, incorporating eccentricity typi- cally requires reanalyzing existing events, which can be computationally expensive. To demonstrate the robust- ness of this approach using a simulated signal with an ec- centricity at 20 Hz of 0.25, we show how ASPIRE can ex- tend a result obtaine...

    1928

  49. [49]

    Mandel and F

    I. Mandel and F. S. Broekgaarden, Rates of compact object coalescences, Living Rev. Rel.25, 1 (2022), arXiv:2107.14239 [astro-ph.HE]

    Pith/arXiv arXiv 2022

  50. [50]

    I. M. Romero-Shaw, P. D. Lasky, E. Thrane, and J. C. Bustillo, GW190521: orbital eccentricity and signatures of dynamical formation in a binary black hole merger sig- nal, Astrophys. J. Lett.903, L5 (2020), arXiv:2009.04771 [astro-ph.HE]

    Pith/arXiv arXiv 2020

  51. [51]

    I. M. Romero-Shaw, P. D. Lasky, and E. Thrane, Four Eccentric Mergers Increase the Evidence that LIGO–Virgo–KAGRA’s Binary Black Holes Form Dynamically, Astrophys. J.940, 171 (2022), arXiv:2206.14695 [astro-ph.HE]

    Pith/arXiv arXiv 2022

  52. [52]

    Gupteet al., Evidence for eccentricity in the pop- ulation of binary black holes observed by LIGO-Virgo- KAGRA (2024), arXiv:2404.14286 [gr-qc]

    N. Gupteet al., Evidence for eccentricity in the pop- ulation of binary black holes observed by LIGO-Virgo- KAGRA (2024), arXiv:2404.14286 [gr-qc]

    Pith/arXiv arXiv 2024

  53. [53]

    Morras, G

    G. Morras, G. Pratten, and P. Schmidt, Orbital ec- centricity in a neutron star - black hole binary (2025), arXiv:2503.15393 [astro-ph.HE]

    arXiv 2025

  54. [54]

    Buonanno, B

    A. Buonanno, B. Iyer, E. Ochsner, Y. Pan, and B. S. Sathyaprakash, Comparison of post-Newtonian templates for compact binary inspiral signals in gravitational-wave detectors, Phys. Rev. D80, 084043 (2009), arXiv:0907.0700 [gr-qc]

    Pith/arXiv arXiv 2009

  55. [55]

    Moore, M

    B. Moore, M. Favata, K. G. Arun, and C. K. Mishra, Gravitational-wave phasing for low-eccentricity inspi- ralling compact binaries to 3PN order, Phys. Rev. D93, 124061 (2016), arXiv:1605.00304 [gr-qc]

    Pith/arXiv arXiv 2016

  56. [56]

    B. P. Abbottet al.(LIGO Scientific, Virgo), GW150914: First results from the search for binary black hole coa- lescence with Advanced LIGO, Phys. Rev. D93, 122003 (2016), arXiv:1602.03839 [gr-qc]

    Pith/arXiv arXiv 2016

  57. [57]

    Vietset al., Reconstructing the calibrated strain signal in the Advanced LIGO detectors, Class

    A. Vietset al., Reconstructing the calibrated strain signal in the Advanced LIGO detectors, Class. Quant. Grav.35, 095015 (2018), arXiv:1710.09973 [astro-ph.IM]

    Pith/arXiv arXiv 2018

  58. [58]

    A. G. Abacet al.(LIGO Scientific, VIRGO, KAGRA), GWTC-4.0: Methods for Identifying and Characterizing Gravitational-wave Transients (2025), arXiv:2508.18081 [gr-qc]

    Pith/arXiv arXiv 2025

  59. [59]

    Pratten, S

    G. Pratten, S. Husa, C. Garcia-Quiros, M. Colleoni, A. Ramos-Buades, H. Estelles, and R. Jaume, Setting the cornerstone for a family of models for gravitational waves from compact binaries: The dominant harmonic for nonprecessing quasicircular black holes, Phys. Rev. D 102, 064001 (2020), arXiv:2001.11412 [gr-qc]

    Pith/arXiv arXiv 2020

  60. [60]

    Pompiliet al., Laying the foundation of the effective- one-body waveform models SEOBNRv5: Improved ac- curacy and efficiency for spinning nonprecessing bi- nary black holes, Phys

    L. Pompiliet al., Laying the foundation of the effective- one-body waveform models SEOBNRv5: Improved ac- curacy and efficiency for spinning nonprecessing bi- nary black holes, Phys. Rev. D108, 124035 (2023), arXiv:2303.18039 [gr-qc]

    Pith/arXiv arXiv 2023

  61. [61]

    Buchholz, N

    A. Buchholz, N. Chopin, and P. E. Jacob, Adaptive tun- ing of hamiltonian monte carlo within sequential monte carlo, Bayesian Analysis16, 745 (2021)

    2021

  62. [62]

    Le Thu Nguyen, F

    T. Le Thu Nguyen, F. Septier, G. W. Peters, and Y. Delignon, Improving smc sampler estimate by recy- cling all past simulated particles, in2014 IEEE Work- shop on Statistical Signal Processing (SSP)(2014) pp. 117–120

    2014

  63. [63]

    Karamanis and U

    M. Karamanis and U. Seljak, Persistent sampling: En- hancing the efficiency of sequential monte carlo, Statistics and Computing35, 1 (2025)

    2025

  64. [64]

    M. M. Graham and A. J. Storkey, Continuously tempered Hamiltonian Monte Carlo, arXiv e-prints , arXiv:1704.03338 (2017), arXiv:1704.03338 [stat.CO]

    Pith/arXiv arXiv 2017

  65. [65]

    A. H. Nitz, Robust, rapid, and simple gravitational-wave parameter estimation, Phys. Rev. D112, 023032 (2025), arXiv:2410.05190 [astro-ph.IM]

    Pith/arXiv arXiv 2025

  66. [66]

    Cabourn Davieset al., Premerger observation and characterization of massive black hole binaries, Phys

    G. Cabourn Davieset al., Premerger observation and characterization of massive black hole binaries, Phys. Rev. D111, 043045 (2025), arXiv:2411.07020 [hep-ex]

    Pith/arXiv arXiv 2025

  67. [67]

    G¨ artner, N

    L. G¨ artner, N. Hartmann, L. Heinrich, M. Horstmann, T. Kuhr, M. Reboud, S. Stefkova, and D. van Dyk, Con- structing model-agnostic likelihoods, a method for the reinterpretation of particle physics results, Eur. Phys. J. C84, 693 (2024), arXiv:2402.08417 [hep-ph]

    Pith/arXiv arXiv 2024

  68. [68]

    Arjona, L

    R. Arjona, L. Espinosa-Portales, J. Garc ´ ıa-Bellido, and S. Nesseris, A GREAT model comparison against the cos- mological constant, Phys. Dark Univ.36, 101029 (2022), arXiv:2111.13083 [astro-ph.CO]

    Pith/arXiv arXiv 2022

  69. [69]

    Z. G. Lane, A. Seifert, R. Ridden-Harper, and D. L. Wiltshire, Cosmological foundations revisited with Pan- theon+, Mon. Not. Roy. Astron. Soc.536, 1752 (2025), arXiv:2311.01438 [astro-ph.CO]

    Pith/arXiv arXiv 2025

  70. [70]

    M. J. Williams, Accelerated Sequential Posterior Infer- ence via Reuse for Gravitational-Wave Analyses - Data Release (2025)

    2025

  71. [71]

    M. J. Williams, Accelerated Sequential Posterior Infer- ence via Reuse for Gravitational-Wave Analyses - Code Release (2025)

    2025

  72. [72]

    M. J. Williams, aspire-inference, Python Package Index (PyPI) (2025)

    2025

  73. [73]

    M. J. Williams, aspire-bilby, Python Package Index (PyPI) (2025)

    2025

  74. [74]

    M. J. Williams, aspire-gw, Python Package Index (PyPI) (2025)

    2025

  75. [75]

    Abbottet al.(LIGO Scientific, Virgo), Open data from the first and second observing runs of Advanced LIGO and Advanced Virgo, SoftwareX13, 100658 (2021), arXiv:1912.11716 [gr-qc]

    R. Abbottet al.(LIGO Scientific, Virgo), Open data from the first and second observing runs of Advanced LIGO and Advanced Virgo, SoftwareX13, 100658 (2021), arXiv:1912.11716 [gr-qc]

    Pith/arXiv arXiv 2021

  76. [76]

    for Python Data API Standards, array-api-compat, https://pypi.org/project/array-api-compat/ (2025)

    C. for Python Data API Standards, array-api-compat, https://pypi.org/project/array-api-compat/ (2025)

    2025

  77. [77]

    Ashtonet al., BILBY: A user-friendly Bayesian infer- ence library for gravitational-wave astronomy, Astrophys

    G. Ashtonet al., BILBY: A user-friendly Bayesian infer- ence library for gravitational-wave astronomy, Astrophys. J. Suppl.241, 27 (2019), arXiv:1811.02042 [astro-ph.IM]

    Pith/arXiv arXiv 2019

  78. [78]

    Talbot, G

    C. Talbot, G. Ashton, M. H¨ ubner, M. Pitkin, plasky, asb5468, M. J. Williams, R. Smith, A. Vijaykumar, SMorisaki, J. Veitch, N. Sarin, D. Macleod, D. Williams, MarcArene, JasperMartins, C. P. L. Berry, V. Ray- mond, Ceciliogq, I. Markin, D. Keitel, AlexandreGoet- tel, L. Pompili, M. Wright, oliviawilk, noahewolfe, jacob- golomb, S. Wu, R. Udall, and M. P...

    2025

  79. [79]

    Foreman-Mackey, corner.py: Scatterplot matrices in python, The Journal of Open Source Software1, 24 (2016)

    D. Foreman-Mackey, corner.py: Scatterplot matrices in python, The Journal of Open Source Software1, 24 (2016)

    2016

  80. [80]

    J. S. Speagle, dynesty: a dynamic nested sampling package for estimating Bayesian posteriors and evi- 8 dences, Mon. Not. Roy. Astron. Soc.493, 3132 (2020), arXiv:1904.02180 [astro-ph.IM]

    Pith/arXiv arXiv 2020

Showing first 80 references.