Pith. sign in

REVIEW 5 major objections 5 minor 78 references

Unlocking New Paths for Science with Extreme-Mass-Ratio Inspirals: Machine Learning-Enhanced MCMC for Accurate Parameter Inversion

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read FM-MCMC, a flow-matching-enhanced parallel-tempered MCMC sampler, recovers EMRI intrinsic parameters within 1-sigma credible intervals under realistic detector noise where stand-alone MCMC fails.

desk verdict A sensible flow-matching + PTMCMC hybrid for EMRI inference, but the headline claim of unbiased recovery under fully unrestricted priors rests on a single run and an internal parameter inconsistency. read the letter →

arxiv 2508.00348 v2 pith:FQCQKI5B submitted 2025-08-01 gr-qc astro-ph.COastro-ph.IMphysics.space-ph

classification gr-qcastro-ph.COastro-ph.IMphysics.space-ph
keywords extreme-mass-ratioinspiralsgravitational-waveparameterestimationflowmatchingcontinuousnormalizingflowsparalleltemperingMCMCBayesianinferenceKerrblackholestime-delayinterferometry
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that a hybrid sampler called FM-MCMC can recover the intrinsic parameters of extreme-mass-ratio inspirals (EMRIs)—the central black hole mass, the companion mass, the black hole spin, and the orbital eccentricity—with true values inside the 1-sigma posterior credible intervals, even when the intrinsic parameters are initialized from broad priors and the data contain realistic stationary Gaussian detector noise. Stand-alone parallel-tempered MCMC, run on the same injection with the same broad intrinsic priors, gets trapped in local likelihood maxima and returns posteriors whose true values lie outside the 3-sigma region. The paper therefore positions FM-MCMC as the first method to achieve unbiased intrinsic-parameter estimation for Kerr EMRIs under noise conditions representative of planned space-based gravitational-wave observatories. A reader should care because unbiased EMRI parameter recovery is the bottleneck standing between detection and the promised science: precision tests of the Kerr/no-hair structure of black holes, formation studies, and probes of dark matter environments.

What carries the argument

The load-bearing object is the continuous normalizing flow (CNF) trained by flow matching, used as a learned proposal generator for parallel-tempered MCMC. A CNF transports samples from a simple base distribution to the posterior by integrating a neural velocity field along a time parameter; flow matching supplies a regression objective for that velocity field so the transport is learned directly from simulated signal-parameter pairs. In FM-MCMC, the trained CNF produces thousands of high-likelihood posterior samples in minutes, and these samples initialize the walkers and temperature ladder of the MCMC stage. This machinery converts a global search over a highly multimodal, degenerate prior volume into a warm-started local refinement, which is what allows the final MCMC chains to find the global mode instead of a spurious local maximum.

What would settle it

Run the same pipeline on an injected EMRI signal in which sky-location and spin-orientation angles are drawn from their full prior ranges, as the intrinsic parameters already are, and the noise realization contains a data gap or a non-stationary power spectral density; if the true intrinsic values then fall outside the 1-sigma credible intervals, the central claim does not transfer to that setting.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a normalizing-flow proposal distribution can cure the global-convergence failure of MCMC on EMRI likelihood surfaces. The method trains a continuous normalizing flow, via the flow-matching objective, to map a base Gaussian into the posterior for the four intrinsic parameters conditioned on a whitened, frequency-domain TDI signal. Those approximate posterior samples are then used as the starting distribution for parallel-tempered MCMC, whose chains produce the final posterior. On an injected Kerr EMRI signal with SNR exceeding 60 and LISA-like stationary Gaussian noise, the FM-MCMC posterior places the true intrinsic parameters inside the 1-sigma credible intervals for all four parameters, whereas MCMC initialized from the same broad priors converges to a biased mode. The paper further validates unbiasedness with a P-P plot over 1,000 test signals and reports that the full FM-MCMC run completes in about 48 hours on a single GPU, compared with days for the baseline sampler.

Load-bearing premise

The demonstration fixes or tightly initializes all extrinsic parameters and injects only stationary Gaussian noise with a known spectrum, so if a real analysis must simultaneously estimate distance, orbital phase, sky location, and spin orientation under non-stationary noise, the reported unbiased intrinsic-parameter recovery may not transfer.

Editorial extensions

If this is right

  • Bright EMRI signals with SNR above 60 could be analyzed in near real time, completing on a single GPU in about two days rather than requiring weeks of MCMC burn-in.
  • Unbiased intrinsic-parameter recovery would let a detected EMRI directly constrain the central black hole's mass and spin and the companion's mass and eccentricity, enabling precision tests of the Kerr metric.
  • Because the flow's posteriors are calibrated across the prior volume, they can serve as principled prior compression for any subsequent exact sampler, not only MCMC.
  • The hybrid design is agnostic to the waveform model, so the gains should carry over to more accurate waveform families and to other degenerate, multimodal gravitational-wave inference problems.

Reading between the lines

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

  • A natural extension the paper leaves implicit is to train the flow on the full parameter set including distance, orbital phase, sky location, and spin orientation; if that succeeds, the tight-extrinsic initialization in the current practical setting could be dropped entirely.
  • The noise assumption is the most fragile transfer point: real LISA-like data will contain gaps, glitches, and non-stationary power spectral densities, so retraining on realistic noise segments and checking the 1-sigma coverage rate would be a direct stress test.
  • The claimed 'first' status is tied to the augmented analytic kludge waveform model and the SNR-above-60 selection; with higher-fidelity self-force waveforms, the degeneracy structure may change enough to widen or shift the recovered posteriors.
  • The same flow-matching proposal could be paired with importance sampling or nested sampling instead of MCMC, potentially reducing the 48-hour wall-clock cost further.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper introduces FM-MCMC, a hybrid Bayesian inference framework that combines continuous normalizing flows (CNFs), trained via flow matching, with parallel-tempered MCMC (PTMCMC) for parameter estimation of extreme-mass-ratio inspirals (EMRIs) in LISA-like detectors. The authors generate AAK-model EMRI waveforms with Taiji's TDI-A response, inject stationary Gaussian noise, and compare standalone Eryn/PTMCMC, a pure flow-based posterior estimator (FMPE), and the combined FM-MCMC pipeline. They report that, for bright EMRIs with SNR above 60, standalone MCMC initialized from broad priors gets trapped in local maxima, whereas FM-MCMC recovers intrinsic parameters with true values inside 1-sigma credible intervals. The paper also presents a P-P plot over 1,000 test signals to calibrate the flow-based posterior and claims orders-of-magnitude speedup over Eryn.

Significance. If fully substantiated, the manuscript would address a real bottleneck in EMRI data analysis: robust global exploration of a highly multimodal, high-dimensional likelihood with practical computational cost. The hybrid initialization strategy is sensible and the idea of using a flow to provide hot starts for PTMCMC is promising. The P-P plot on 1,000 injected signals is a legitimate calibration check for the flow component, and the explicit comparison with Eryn under broad intrinsic priors is a useful stress test. However, the headline claim—'first successful estimation ... with fully unrestricted priors'—is broader than the evidence presented: the actual demonstrations fix or tightly initialize extrinsic parameters, omit distance and initial phase from the sampled space, use a single injected signal for the main FM-MCMC result, and contain an inconsistency between the injected values in Section 2 and Table 2. The strengths are real but the central claim needs substantially more evidence before it can be accepted.

major comments (5)
  1. [Section 2 and Table 1] The parameter space actually sampled by FM-MCMC omits luminosity distance and initial orbital phase, yet the abstract and Section 2 claim parameter recovery under 'fully unrestricted priors.' Table 1 lists only M, mu, a, e0, and four angular parameters, while the text fixes distance = 2.0 Gpc and initial orbital phase = 0. This is a load-bearing gap because distance and phase are strongly correlated with intrinsic parameters and are degenerate in EMRI likelihood surfaces; the claim of unrestricted-prior recovery is therefore not demonstrated for the full physical parameter space.
  2. [Section 2, initialization setting (2)] The 'practical' setting used for the main FM-MCMC and MCMC comparison initializes the extrinsic angular parameters (sky location and spin orientation) within 1e-4 of their true values, while only intrinsic parameters are drawn from broad priors. This contradicts the paper's claim of 'fully unrestricted priors' and means that the demonstrated robustness is limited to a subspace. The authors should either sample all parameters from their prior ranges in the headline experiment or explicitly qualify the claim to 'broad priors on intrinsic parameters with tightly informed extrinsic parameters.'
  3. [Table 2 vs Section 2] There is a direct numerical inconsistency in the injected parameters: Section 2 states a = 0.423 and e0 = 0.189, while Table 2 lists the injected values as a = 0.19310 and e0 = 0.48706. The recovered values in Table 2 (a = 0.1894, e0 = 0.4821) match the Table 2 injections, not the Section 2 values, and Figure 7's red lines are said to denote true values without making clear which set they use. This inconsistency must be resolved before the reported 1-sigma coverage can be evaluated at all.
  4. [Section 3 and Figure 5] The P-P plot in Figure 5 calibrates the CNF-based posterior estimator (FMPE) on 1,000 test signals, not the final PTMCMC samples produced by FM-MCMC. Since the paper's central claim is about the coverage of FM-MCMC posterior credible intervals, the P-P plot does not establish that claim. A valid calibration check for FM-MCMC would need repeated injections followed by full FM-MCMC runs, or an importance-sampling-style correction applied to the final samples; as written, the coverage evidence for the headline method rests on a single injected signal.
  5. [Section 2 and Section 4] The main FM-MCMC demonstration is based on one injected signal, with no repeated-injection runs, no convergence diagnostics (e.g., Gelman-Rubin or effective sample size), and only stationary Gaussian noise drawn from a design PSD. Real LISA/Taiji data will include nonstationary noise, gaps, and glitches, and the paper's own framing says 'realistic instrumental noise conditions.' The authors should report at least a small ensemble of injections and standard convergence checks to show that the 1-sigma coverage is not an artifact of a particularly favorable noise realization or a lucky PTMCMC run.
minor comments (5)
  1. [Section 2 text] There is a typo: 'over a a two-month observation period' should read 'over a two-month observation period.'
  2. [Section 4, Eq. (5) area] Several inline equations and symbols are corrupted (e.g., '�' placeholders for mathematical symbols), which makes parts of Section 4 hard to read. The authors should carefully proofread the typeset version, especially Eqs. (5)-(9) and (12)-(18).
  3. [Section 4.6.1] The sentence beginning 'we examine two probability distributions' is grammatically incomplete and appears to be a leftover from an earlier draft. Please rewrite it as a complete sentence.
  4. [References] Reference [61] is given as 'Du M, Liang BH, et al. Advancing Space-Based Gravitational Wave Astronomy' with the year field '2023' but no journal or arXiv identifier; please complete the citation.
  5. [Figure 6 and Figure 7 captions] The captions of Figures 6 and 7 describe the plotted curves in different terms ('blue shading' vs 'blue contours'); please make the caption terminology consistent so that readers can interpret the figures correctly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the CNF only initializes PTMCMC, and the final posterior is produced by likelihood-based MCMC against independently injected parameters.

full rationale

The paper's central claim is an empirical injection-recovery test, not a quantity defined by the method's own inputs. The CNF is trained on 20,000 waveforms with parameters drawn from the Table 1 priors (Sec. 4.5) and is used only to propose starting points; the paper states in the Figure 2c caption that 'Final posterior sampling is executed exclusively via PTMCMC’s likelihood evaluations.' Thus the recovered FM-MCMC values in Table 2 are produced by the same likelihood that defines the posterior, with independently injected parameters as targets, and standalone Eryn is used as a separate benchmark. No equation in the paper defines an output as an input, and no fitted parameter is renamed as a prediction. The self-citations [61] and [70] are background references for the Taiji orbit model and for the general efficiency of FMPE; neither is invoked as a uniqueness theorem or as the justification for the recovery result, so they are not load-bearing. Non-circular evidence gaps remain, however: Sec. 2's 'practical' setting (2) tight-initializes all extrinsic parameters (theta_S, phi_S, theta_K, phi_K) to within 1e-4 of truth, while the headline claims 'fully unrestricted priors'; luminosity distance and initial orbital phase are fixed rather than sampled; Table 2 gives injected values a = 0.19310, e0 = 0.48706, conflicting with the representative injection a = 0.423, e0 = 0.189 in Sec. 2; and the Figure 5 P-P plot calibrates the CNF proposal rather than the final PTMCMC samples. These are scope and support concerns, not circularity.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the accuracy of the AAK waveform model, stationary Gaussian Taiji noise, a prior box that contains the true source, and an initialization protocol that keeps extrinsic parameters near truth. None of these are derived in this paper; they are imported from the simulation pipeline. The chosen network and sampler hyperparameters are additional hand-set inputs. No new physical entities are introduced.

free parameters (3)
  • SNR selection threshold = SNR > 60
    Only signals above SNR 60 enter training, test, and the FM-MCMC demonstration; the unbiasedness claim is conditional on this bright-source subset.
  • Flow network hyperparameters = 21 residual blocks, 2000 epochs, Adam lr 1e-4, 2048-dim latent
    Chosen without an ablation study; performance may be sensitive to these choices.
  • PTMCMC configuration = 20 walkers, 25 temperature levels, 20,000 iterations
    Hand-set; no convergence diagnostics or sensitivity study is reported.
assumptions (5)
  • domain assumption The AAK waveform model in FastEMRIWaveforms is sufficiently accurate for EMRI parameter inference.
    Invoked in Section 4.1; if real signals differ from AAK at the required precision, all recovered parameters inherit waveform-systematic bias.
  • domain assumption Taiji noise is stationary, Gaussian, and fully described by the design PSD used in Eq. (7).
    Invoked in Section 4.4; real detector noise has transients, gaps, and nonstationarity not modeled here.
  • domain assumption The prior ranges in Table 1 contain the true source parameters.
    Training and test signals are drawn from these priors; sources outside the box are not covered by the amortized flow.
  • ad hoc to paper Extrinsic parameters can be initialized near their true values in practical analysis.
    Setting (2) tight-initializes sky location and spin orientation; this is not a fully unrestricted prior analysis and is load-bearing for the claimed intrinsic recovery.
  • standard math The Gaussian likelihood after FFT and whitening is the correct sampling distribution.
    Used for PTMCMC likelihood evaluations; standard, but depends on the noise model being exact.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unlocking New Paths for Science with Extreme-Mass-Ratio Inspirals: Machine Learning-Enhanced MCMC for Accurate Parameter Inversion." pith.science (2026). https://pith.science/paper/FQCQKI5B

@misc{pith2026250800348,
  author       = {Pith},
  title        = {Pith review of: Unlocking New Paths for Science with Extreme-Mass-Ratio Inspirals: Machine Learning-Enhanced MCMC for Accurate Parameter Inversion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FQCQKI5B}},
  note         = {Machine review of arXiv:2508.00348}
}
read the original abstract

The detection of gravitational waves from extreme-mass-ratio inspirals (EMRIs) in space-borne antennas like Taiji and LISA promises deep insights into strong-field gravity and black hole physics. However, the complex, highly degenerate, and non-convex likelihood landscapes characteristic of EMRI parameter spaces pose severe challenges for conventional Markov chain Monte Carlo (MCMC) methods. Under realistic instrumental noise and broad priors, these methods demand impractical computational costs but are prone to becoming trapped in local maxima, leading to biased and unreliable parameter estimates. To address this, we introduce Flow-Matching Markov Chain Monte Carlo (FM-MCMC), a novel Bayesian framework that integrates continuous normalizing flows (CNFs) with parallel tempering MCMC (PTMCMC). By generating high-likelihood regions via CNFs and refining them through PTMCMC, FM-MCMC enables robust exploration of the nontrivial parameter spaces, while achieving orders-of-magnitude improvement in computational efficiency and, more importantly, ensuring statistically reliable and unbiased inference. By enabling real-time, unbiased parameter inference, FM-MCMC could unlock the full scientific potential of EMRI observations, and would serve as a scalable pipeline for precision gravitational-wave astronomy.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

78 extracted references · 73 canonical work pages

  1. [1]

    Advanced LIGO

    Aasi J et al. Advanced LIGO. Class. Quant. Grav. 2015;32:074001

  2. [2]

    Observation of Gravitational Waves from a Binary Black Hole Merger

    Abbott BP et al. Observation of Gravitational Waves from a Binary Black Hole Merger. Phys. Rev. Lett. 2016;116:061102

  3. [3]

    LIGO-Virgo-KAGRA Announce the 200th Gravitational Wave Detection of O4! ����������������������������������������������

    LIGO-Virgo-KAGRA Collaboration. LIGO-Virgo-KAGRA Announce the 200th Gravitational Wave Detection of O4! ���������������������������������������������� . Accessed: July 19, 2025. 2025. 18

  4. [4]

    GWTC-2: Compact Binary Coalescences Observed by LIGO and Virgo During the First Half of the Third Observing Run

    Abbott R et al. GWTC-2: Compact Binary Coalescences Observed by LIGO and Virgo During the First Half of the Third Observing Run. Phys. Rev. X 2021;11:021053

  5. [5]

    GWTC-3: Compact Binary Coalescences Observed by LIGO and Virgo during the Second Part of the Third Observing Run

    Abbott R et al. GWTC-3: Compact Binary Coalescences Observed by LIGO and Virgo during the Second Part of the Third Observing Run. Phys. Rev. X 2023;13:041039

  6. [6]

    The LISA-Taiji Network: Precision Local- ization of Coalescing Massive Black Hole Binaries

    Ruan WH, Liu C, Guo ZK, Wu YL, and Cai RG. The LISA-Taiji Network: Precision Local- ization of Coalescing Massive Black Hole Binaries. Research 2021;2021

  7. [7]

    Verification of Laser Heterodyne Interferometric Bench for Chinese Spaceborne Gravitational Wave Detection Missions

    Xu X, Liu H, and Tan Y. Verification of Laser Heterodyne Interferometric Bench for Chinese Spaceborne Gravitational Wave Detection Missions. Research 2024;7:0302

  8. [8]

    The Taiji Program in Space for gravitational wave physics and the nature of gravity

    Hu WR and Wu YL. The Taiji Program in Space for gravitational wave physics and the nature of gravity. Natl. Sci. Rev. 2017;4:685–6

Show all 78 references
  1. [9]

    TianQin: a space-borne gravitational wave detector

    Luo J et al. TianQin: a space-borne gravitational wave detector. Class. Quant. Grav. 2016;33:035010

  2. [10]

    Laser Interferometer Space Antenna

    Amaro-Seoane P et al. Laser Interferometer Space Antenna. arXiv e-prints 2017, arXiv:1702.00786:arXiv:1702.00786

  3. [11]

    Science with the space-based interferometer LISA

    Babak S, Gair J, Sesana A, et al. Science with the space-based interferometer LISA. V: Extreme mass-ratio inspirals. Phys. Rev. D 2017;95:103012

  4. [12]

    Testing General Relativity with Low-Frequency, Space-Based Gravitational-Wave Detectors

    Gair JR, Vallisneri M, Larson SL, and Baker JG. Testing General Relativity with Low-Frequency, Space-Based Gravitational-Wave Detectors. Living Rev. Rel. 2013;16:7

  5. [13]

    Gravitational-wave cosmology with extreme mass- ratio inspirals

    Laghi D, Tamanini N, Del Pozzo W, et al. Gravitational-wave cosmology with extreme mass- ratio inspirals. Mon. Not. Roy. Astron. Soc. 2021;508:4512–31

  6. [14]

    Science with the space-based interferometer LISA

    Babak S, Gair J, Sesana A, et al. Science with the space-based interferometer LISA. V. Extreme mass-ratio inspirals. Physical Review D 2017;95

  7. [15]

    Extreme- and intermediate-mass ratio inspirals in dynamical Chern- Simons modified gravity

    Sopuerta CF and Yunes N. Extreme- and intermediate-mass ratio inspirals in dynamical Chern- Simons modified gravity. Physical Review D 2009;80

  8. [16]

    Using LISA extreme-mass-ratio inspiral sources to test off-Kerr devi- ations in the geometry of massive black holes

    Barack L and Cutler C. Using LISA extreme-mass-ratio inspiral sources to test off-Kerr devi- ations in the geometry of massive black holes. Physical Review D 2007;75

  9. [17]

    Probing fundamental physics with Extreme Mass Ratio Inspirals: a full Bayesian inference for scalar charge

    Speri L, Barsanti S, Maselli A, et al. Probing fundamental physics with Extreme Mass Ratio Inspirals: a full Bayesian inference for scalar charge. 2024. arXiv: ���������� ������� . ��� : ��������������������������������

  10. [18]

    Constraint on the Deviation of Kerr Metric via Bumpy Parameterization and Particle Swarm Optimization in Extreme Mass-Ratio Inspirals

    Zou X, Zhong X, Han WB, and Mohanty SD. Constraint on the Deviation of Kerr Metric via Bumpy Parameterization and Particle Swarm Optimization in Extreme Mass-Ratio Inspirals

  11. [19]

    Gravitational waves from extreme-mass-ratio inspirals using general parametrized metrics

    Xin S, Han WB, and Yang SC. Gravitational waves from extreme-mass-ratio inspirals using general parametrized metrics. Phys. Rev. D 8 2019;100:084055

  12. [20]

    LISA extreme-mass-ratio inspiral events as probes of the black hole mass function

    Gair JR, Tang C, and Volonteri M. LISA extreme-mass-ratio inspiral events as probes of the black hole mass function. Physical Review D 2010;81

  13. [21]

    Probing Accretion Physics with Gravitational Waves

    Speri L, Antonelli A, Sberna L, et al. Probing Accretion Physics with Gravitational Waves. Physical Review X 2023;13. 19

  14. [22]

    Disks, spikes, and clouds: distinguishing environmental effects on BBH gravitational waveforms

    Cole PS, Bertone G, Coogan A, et al. Disks, spikes, and clouds: distinguishing environmental effects on BBH gravitational waveforms. 2022. arXiv: ���� � ����� ������� . ��� : ����� � ��������������������������

  15. [23]

    Extreme dark matter tests with extreme mass ratio inspirals

    Hannuksela OA, Ng KCY, and Li TGF. Extreme dark matter tests with extreme mass ratio inspirals. Phys. Rev. D 10 2020;102:103022

  16. [24]

    Dark Matter: An Efficient Catalyst for Intermediate-mass- ratio-inspiral Events

    Yue XJ, Han WB, and Chen X. Dark Matter: An Efficient Catalyst for Intermediate-mass- ratio-inspiral Events. The Astrophysical Journal 2019;874:34

  17. [25]

    Event rate estimates for LISA extreme mass ratio capture sources

    Gair JR, Barack L, Creighton T, et al. Event rate estimates for LISA extreme mass ratio capture sources. Class. Quant. Grav. 2004;21:S1595–S1606

  18. [26]

    EMRI MC: A GPU-based code for Bayesian inference of EMRI wave- forms

    Saltas ID and Oliveri R. EMRI MC: A GPU-based code for Bayesian inference of EMRI wave- forms. arXiv preprint arXiv:2311.17174 2023

  19. [27]

    Rapid generation of fully relativis- tic extreme-mass-ratio-inspiral waveform templates for LISA data analysis

    Chua AJK, Katz ML, Warburton N, and Hughes SA. Rapid generation of fully relativis- tic extreme-mass-ratio-inspiral waveform templates for LISA data analysis. Phys. Rev. Lett. 2021;126:051102

  20. [28]

    The Mock LISA Data Challenges: From Challenge 3 to Challenge 4

    Babak S et al. The Mock LISA Data Challenges: From Challenge 3 to Challenge 4. Class. Quant. Grav. 2010;27. Ed. by Marka Z and Marka S:084009

  21. [29]

    Nonlocal parameter degeneracy in the intrinsic space of gravitational- wave signals from extreme-mass-ratio inspirals

    Chua AJK and Cutler CJ. Nonlocal parameter degeneracy in the intrinsic space of gravitational- wave signals from extreme-mass-ratio inspirals. Phys. Rev. D 2022;106:124046

  22. [30]

    Swarm Intelligence Methods for Extreme Mass Ratio Inspiral Search: First Application of Particle Swarm Optimization

    Zou XB, Mohanty SD, Luo HG, and Liu YX. Swarm Intelligence Methods for Extreme Mass Ratio Inspiral Search: First Application of Particle Swarm Optimization. Universe 2024;10

  23. [31]

    Flow Matching for Scalable Simulation-Based Inference

    Dax M, Wildberger J, Buchholz S, Green SR, Macke JH, and Scholkopf B. Flow Matching for Scalable Simulation-Based Inference. ArXiv 2023;abs/2305.17161

  24. [32]

    Detection strategies for extreme mass ratio inspirals

    Cornish NJ. Detection strategies for extreme mass ratio inspirals. Classical and Quantum Grav- ity 2011;28:094016

  25. [33]

    An algorithm for the detection of extreme mass ratio inspirals in LISA data

    Babak S, Gair JR, and Porter EK. An algorithm for the detection of extreme mass ratio inspirals in LISA data. Classical and Quantum Gravity 2009;26:135004

  26. [34]

    Flow matching for generative mod- eling

    Lipman Y, Chen RT, Ben-Hamu H, Nickel M, and Le M. Flow matching for generative mod- eling. arXiv preprint arXiv:2210.02747 2022

  27. [35]

    Eryn: a multipurpose sampler for Bayesian inference

    Karnesis N, Katz ML, Korsakova N, Gair JR, and Stergioulas N. Eryn: a multipurpose sampler for Bayesian inference. Mon. Not. Roy. Astron. Soc. 2023;526:4814–30

  28. [36]

    emcee: The MCMC Hammer

    Foreman-Mackey D, Hogg DW, Lang D, and Goodman J. emcee: The MCMC Hammer. Publ. Astron. Soc. Pac. 2013;125:306–12

  29. [37]

    Fast extreme-mass-ratio-inspiral waveforms: New tools for millihertz gravitational-wave data analysis

    Katz ML, Chua AJK, Speri L, Warburton N, and Hughes SA. Fast extreme-mass-ratio-inspiral waveforms: New tools for millihertz gravitational-wave data analysis. Phys. Rev. D 2021;104:064047

  30. [38]

    Eryn: a multipurpose sampler for Bayesian inference

    Karnesis N, Katz ML, Korsakova N, Gair JR, and Stergioulas N. Eryn: a multipurpose sampler for Bayesian inference. Monthly Notices of the Royal Astronomical Society 2023;526:4814–30. 20

  31. [39]

    ¡tt¿emcee¡/tt¿: The MCMC Hammer

    Foreman-Mackey D, Hogg DW, Lang D, and Goodman J. ¡tt¿emcee¡/tt¿: The MCMC Hammer. Publications of the Astronomical Society of the Pacific 2013;125:306–12

  32. [40]

    An efficient GPU-accelerated multi-source global fit pipeline for LISA data analysis

    Katz ML, Karnesis N, Korsakova N, Gair JR, and Stergioulas N. An efficient GPU-accelerated multi-source global fit pipeline for LISA data analysis. 2024. arXiv: ���������� ������� . ��� : ��������������������������������

  33. [41]

    Binary Black Hole Mergers in the First Advanced LIGO Observing Run

    Abbott BP, Abbott R, Abbott TD, et al. Binary Black Hole Mergers in the First Advanced LIGO Observing Run. Phys. Rev. X 4 2016;6:041015

  34. [42]

    ASTROPHYSICAL IMPLICATIONS OF THE BI- NARY BLACK HOLE MERGER GW150914

    Abbott BP, Abbott R, Abbott TD, et al. ASTROPHYSICAL IMPLICATIONS OF THE BI- NARY BLACK HOLE MERGER GW150914. The Astrophysical Journal Letters 2016;818:L22

  35. [43]

    Testing General Relativity with Low-Frequency, Space-Based Gravitational-Wave Detectors

    Gair JR, Vallisneri M, Larson SL, and Baker JG. Testing General Relativity with Low-Frequency, Space-Based Gravitational-Wave Detectors. Living Reviews in Relativity 2013;16

  36. [44]

    Gravitational Wave- forms for Compact Binaries from Second-Order Self-Force Theory

    Wardell B, Pound A, Warburton N, Miller J, Durkan L, and Le Tiec A. Gravitational Wave- forms for Compact Binaries from Second-Order Self-Force Theory. Phys. Rev. Lett. 24 2023;130:241402

  37. [45]

    Theoretical physics implications of the binary black-hole mergers GW150914 and GW151226

    Yunes N, Yagi K, and Pretorius F. Theoretical physics implications of the binary black-hole mergers GW150914 and GW151226. Phys. Rev. D 8 2016;94:084002

  38. [46]

    Tests of General Relativity with GW150914

    Abbott BP, Abbott R, Abbott TD, et al. Tests of General Relativity with GW150914. Phys. Rev. Lett. 22 2016;116:221101

  39. [47]

    Self-force and radiation reaction in general relativity

    Barack L and Pound A. Self-force and radiation reaction in general relativity. Rept. Prog. Phys. 2019;82:016904

  40. [48]

    Black hole perturbation theory and gravitational self-force

    Pound A and Wardell B. Black hole perturbation theory and gravitational self-force. arXiv e-prints 2021:2101.04592

  41. [49]

    Gravitational self-force on generic bound geodesics in Kerr spacetime

    Meent M van de. Gravitational self-force on generic bound geodesics in Kerr spacetime. Phys. Rev. D 2018;97:104033

  42. [50]

    Second-Order Self-Force Calculation of Grav- itational Binding Energy in Compact Binaries

    Pound A, Wardell B, Warburton N, and Miller J. Second-Order Self-Force Calculation of Grav- itational Binding Energy in Compact Binaries. Phys. Rev. Lett. 2020;124:021101

  43. [51]

    Gravitational-Wave Energy Flux for Compact Binaries through Second Order in the Mass Ratio

    Warburton N, Pound A, Wardell B, Miller J, and Durkan L. Gravitational-Wave Energy Flux for Compact Binaries through Second Order in the Mass Ratio. Phys. Rev. Lett. 2021;127:151102

  44. [52]

    Albertini A, Gamba R, Nagar A, and Bernuzzi S. Effective-one-body waveforms for extreme- mass-ratio binaries: Consistency with second-order gravitational self-force quasicircular results and extension to nonprecessing spins and eccentricity. Physical Review D 2024;109

  45. [53]

    Influence of mass-ratio corrections in extreme-mass-ratio inspirals for testing general relativity

    Shen P, Han WB, Zhang C, et al. Influence of mass-ratio corrections in extreme-mass-ratio inspirals for testing general relativity. Physical Review D 2023;108

  46. [54]

    LISA capture sources: Approximate waveforms, signal-to-noise ratios, and parameter estimation accuracy

    Barack L and Cutler C. LISA capture sources: Approximate waveforms, signal-to-noise ratios, and parameter estimation accuracy. Phys. Rev. D 2004;69:082005

  47. [55]

    ’Kludge’ gravitational waveforms for a test-body orbiting a Kerr black hole

    Babak S, Fang H, Gair JR, Glampedakis K, and Hughes SA. ’Kludge’ gravitational waveforms for a test-body orbiting a Kerr black hole. Phys. Rev. D 2007;75. [Erratum: Phys.Rev.D 77, 04990 (2008)]:024005. 21

  48. [56]

    Improved analytic extreme-mass-ratio inspiral model for scoping out eLISA data analysis

    Chua AJK and Gair JR. Improved analytic extreme-mass-ratio inspiral model for scoping out eLISA data analysis. Class. Quant. Grav. 2015;32:232002

  49. [57]

    Augmented kludge waveforms for detecting extreme-mass- ratio inspirals

    Chua AJK, Moore CJ, and Gair JR. Augmented kludge waveforms for detecting extreme-mass- ratio inspirals. Phys. Rev. D 2017;96:044005

  50. [58]

    Improved approximate inspirals of test-bodies into Kerr black holes

    Gair JR and Glampedakis K. Improved approximate inspirals of test-bodies into Kerr black holes. Phys. Rev. D 2006;73:064037

  51. [59]

    Assessing the data-analysis impact of LISA orbit approximations using a GPU-accelerated response model

    Katz ML, Bayle JB, Chua AJK, and Vallisneri M. Assessing the data-analysis impact of LISA orbit approximations using a GPU-accelerated response model. Phys. Rev. D 2022;106:103001

  52. [60]

    LISA Data Challenge Manual

    WG L. LISA Data Challenge Manual. ���������������������������������������������� ������������������

  53. [61]

    Advancing Space-Based Gravitational Wave Astronomy: Rapid Detection and Parameter Estimation Using Normalizing Flows

    Du M, Liang BH, Wang H, Xu P, Luo Z, and Wu Y. Advancing Space-Based Gravitational Wave Astronomy: Rapid Detection and Parameter Estimation Using Normalizing Flows. In:

  54. [62]

    Time-Delay Interferometry Simulations for the Laser Interferometer Space Antenna

    Otto M. Time-Delay Interferometry Simulations for the Laser Interferometer Space Antenna. PhD thesis. Leibniz U., Hannover, 2015. ���: �������������

  55. [63]

    Assessing the data-analysis impact of LISA orbit approximations using a GPU-accelerated response model

    Katz ML, Bayle JB, Chua AJK, and Vallisneri M. Assessing the data-analysis impact of LISA orbit approximations using a GPU-accelerated response model. Physical Review D 2022;106

  56. [64]

    Accuracy Requirements: Assessing the Impor- tance of First Post-Adiabatic Terms for Small-Mass-Ratio Binaries

    Burke O, Piovano GA, Warburton N, et al. Accuracy Requirements: Assessing the Impor- tance of First Post-Adiabatic Terms for Small-Mass-Ratio Binaries. 2024. arXiv: ���������� ������� . ��� : ��������������������������������

  57. [65]

    Overview and progress on the Laser Interferometer Space Antenna mission

    Bayle JB, Bonga B, Caprini C, et al. Overview and progress on the Laser Interferometer Space Antenna mission. Nature Astronomy 2022;6:1334–8

  58. [66]

    A roadmap of gravitational wave data analysis

    Speri L, Karnesis N, Renzini AI, and Gair JR. A roadmap of gravitational wave data analysis. Nature Astronomy 2022;6:1356–63

  59. [67]

    Numerical simulation of sky localization for LISA-TAIJI joint observation

    Wang G, Ni WT, Han WB, Yang SC, and Zhong XY. Numerical simulation of sky localization for LISA-TAIJI joint observation. Phys. Rev. D 2020;102:024089

  60. [68]

    Extreme Mass Ratio Inspirals: Perspectives for Their De- tection

    Babak S, Gair JR, and Cole RH. Extreme Mass Ratio Inspirals: Perspectives for Their De- tection. In: ��������� �� ������ �� ������������ �������. Springer International Publishing, 2015:783–812. ���: ���������������������������� . ��� : �������������������������� ��������������������

  61. [69]

    Fast �-free Inference of Simulation Models with Bayesian Con- ditional Density Estimation

    Papamakarios G and Murray I. Fast �-free Inference of Simulation Models with Bayesian Con- ditional Density Estimation. 2018. arXiv: ���������� ��������� . ��� : ������������������ ��������������

  62. [70]

    Rapid Parameter Estimation for Merging Massive Black Hole Binaries Using ODE-Based Generative Models

    Liang B, Du M, Wang H, et al. Rapid Parameter Estimation for Merging Massive Black Hole Binaries Using ODE-Based Generative Models. 2024. arXiv:���������� ������� . ��� : ������ ��������������������������

  63. [71]

    Real-Time Gravitational Wave Science with Neural Posterior Estimation

    Dax M, Green SR, Gair J, Macke JH, Buonanno A, and Sch¨ olkopf B. Real-Time Gravitational Wave Science with Neural Posterior Estimation. Physical Review Letters 2021;127. 22

  64. [72]

    Inferring Atmospheric Properties of Exoplanets with Flow Matching and Neural Importance Sampling

    Gebhard TD, Wildberger J, Dax M, Angerhausen D, Quanz SP, and Sch¨ olkopf B. Inferring Atmospheric Properties of Exoplanets with Flow Matching and Neural Importance Sampling

  65. [73]

    Variational inference with normalizing flows

    Rezende D and Mohamed S. Variational inference with normalizing flows. In: ������������� ���������� �� ������� ��������. PMLR. 2015:1530–8

  66. [74]

    Normal- izing Flows for Probabilistic Modeling and Inference

    Papamakarios G, Nalisnick E, Rezende DJ, Mohamed S, and Lakshminarayanan B. Normal- izing Flows for Probabilistic Modeling and Inference. Journal of Machine Learning Research 2021;22:1–64

  67. [75]

    ��� : ��������������������������������

    arXiv: ���������� ������������� . ��� : ��������������������������������

  68. [78]

    Neural Ordinary Differential Equa- tions

    Chen RTQ, Rubanova Y, Bettencourt J, and Duvenaud D. Neural Ordinary Differential Equa- tions. 2019. arXiv: ���������� ������� . ��� : �������������������������������� . 23

  69. [2023]

    ��� : ��������������������������������������������������

  70. [2025]

    ��� : ��������������������������������

    arXiv: ���������� ������� . ��� : ��������������������������������

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.