Pith. sign in

REVIEW 1 major objections 7 minor 1 cited by

Tests for model misspecification in simulation-based inference: from local distortions to global model checks

T0 review · 1 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Model misspecification in simulation-based inference can be tested by distorting the simulator, and in a well-defined limit the resulting neural statistics reduce to matched-filter SNRs and chi-square statistics.

desk verdict A sound and useful SBI misspecification-testing framework, but the advertised classical connections are a narrow limiting case and the code reference is incomplete. read the letter →

arxiv 2412.15100 v2 pith:N2DAKNW7 submitted 2024-12-19 astro-ph.IM astro-ph.COcs.LGgr-qc

Noemi Anau Montel , James Alvey , Christoph Weniger This is my paper · ORCID
classification astro-ph.IMastro-ph.COcs.LGgr-qc
keywords simulation-basedinferencemodelmisspecificationhypothesistestinganomalydetectionvalidationmatchedfilteringchi-squaregoodness-of-fitgravitationalwaves
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

Simulation-based inference usually answers the question 'which parameters fit the data?'; this paper asks 'is the simulator itself wrong?'. It proposes to test a base simulator by building a large ensemble of alternative hypotheses in which the simulator is augmented with stochastic distortions, then training neural networks to estimate the log-likelihood-ratio test statistic for each distortion. Under a Gaussian base likelihood, the large-sample limit, and distortions that leave the base model's maximum-likelihood estimate essentially unchanged, each localized statistic equals the squared matched-filter signal-to-noise ratio plus a constant, and the sum over unit-vector distortions equals Pearson's chi-square plus a constant. This yields interpretable, simulation-based anomaly detection, model validation, and residual analysis, with a Monte Carlo global p-value that corrects for the many correlated tests. The framework is demonstrated on a white-noise toy model and on the gravitational-wave event GW150914, where it finds no significant misspecification; the practical interest is that this is a step toward end-to-end simulation-based pipelines that validate the simulator itself.

What carries the argument

The machinery is the ensemble of localized likelihood-ratio test statistics $t_i(x)=-2\ln[p_{\mathrm{sim}}(x)/p_{\mathrm{dist}}(x|i)]$, each comparing the base simulator to a simulator augmented with a stochastic additive distortion in direction $n^{(i)}$. These are estimated by neural networks from simulated data pairs, either through a binary cross-entropy classifier loss or through a Gaussian regression loss on the matched-filter amplitude $\epsilon$ and its variance; the regression variant directly outputs the quantities entering the SNR. The analytic bridge is Eq. (B18), $t_i(x)\simeq \mathrm{SNR}_i^2(x)+C$, with $\mathrm{SNR}_i(x)=\Delta x^\top\Sigma^{-1}n^{(i)}/\sqrt{(n^{(i)})^\top\Sigma^{-1}n^{(i)}}$, and its unit-vector limit Eq. (B22) giving $t_{\mathrm{sum}}=\chi^2+\mathrm{const}$. Aggregation of arbitrary subsets of localized tests, plus a Monte Carlo global p-value over the minimum p-value across all tests, turns the ensemble into a trials-corrected model check.

What would settle it

In the Section III white-noise example, inject a known correlated distortion, then compute the neural test statistics with the base-model parameters held fixed at their true values and compare them to the analytic $\mathrm{SNR}^2+C$ prediction; if the previously observed mismatch (Figure 7) disappears, Eq. B7 is exactly the failure point, while any residual mismatch would implicate another assumption.

Watch

Extended reading notes

Core claim

The paper's central claim is that model misspecification in simulation-based inference can be treated as a high-volume hypothesis-testing problem: one defines many distorted simulators, trains a multi-output network to approximate the ratio of each distorted model to the base model, and uses the outputs as localized test statistics, with arbitrary aggregations serving as global checks. The analytic core is the derivation that, for additive stochastic distortions in a Gaussian likelihood with large samples and with the base model's maximum-likelihood parameters nearly unaffected by the distortion, each marginalized localized statistic is $t_i(x)\simeq \mathrm{SNR}_i^2(x)+C$ (Eq. B18), and for distortions along standard basis vectors the aggregated statistic becomes $t_{\mathrm{sum}}(x)=\chi^2(x)+\mathrm{const}$ (Eq. B22). These identities connect neural test statistics to matched filtering and goodness-of-fit testing, so simulation-based results inherit classical interpretability. The paper also introduces a self-calibrating training strategy in which distortion amplitudes are tied to the learned noise variance through a chosen maximum SNR, and demonstrates the full pipeline on real gravitational-wave data.

Load-bearing premise

The analytic claims stand on the assumption that adding a distortion leaves the base model's fitted parameters essentially unchanged (Eq. B7); when a distortion shifts the maximum-likelihood parameters, the matched-filter and chi-square identifications are no longer valid.

Editorial extensions

If this is right

  • Localized neural test statistics for Gaussian base models can be read as matched-filter SNR maps, giving a spatial and correlation-scale localization of where the model fails.
  • The aggregated test over unit-vector distortions provides a chi-square-like goodness-of-fit summary with a Monte Carlo calibrated p-value, so model validation needs no analytic likelihood.
  • The self-calibrating distortion-amplitude rule sets training distortions to be detectable but not absurd, which makes the method practical for high-dimensional data.
  • Because all alternative hypotheses are scored by one trained network and significance is Monte Carlo based, the pipeline is efficient and automatically accounts for the look-elsewhere effect.
  • The GW150914 demonstration shows the framework can serve as a post-inference quality check on real gravitational-wave data.

Reading between the lines

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

  • Editorial inference: if the matched-filter and chi-square identifications survive beyond exactly Gaussian settings, the framework offers a principled prescription for designing SBI diagnostics generally: choose distortion directions, train the SNR or BCE network, and interpret outputs as classical statistics; the limits of that interpretation can be probed by injecting distortions that strongly sh
  • The authors note that using posterior predictive samples for the gravitational-wave null can overestimate distortions degenerate with parameter changes; one testable extension is to re-run the check drawing parameters from the full prior, which should make the global p-values more conservative and better calibrated.
  • The mismatch between neural and profiled-analytic statistics when the MLE absorbs a distortion (Figure 7) suggests that the size of that mismatch could itself be used as a diagnostic for parameter-distortion degeneracy, rather than only as a limitation.
  • The flexibility of arbitrary aggregated tests, such as a double-excess statistic, is a direct consequence of the framework; one could benchmark these against classical bump-hunt searches to see whether simulation-based aggregation recovers or improves their power.
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

1 major / 7 minor

Summary. The paper introduces a simulation-based inference framework for model misspecification testing based on many hypothesis tests against stochastically distorted simulators. The null hypothesis is the base simulator's marginal distribution, and each alternative is defined by an additive distortion in a specified direction. Test statistics are learned with neural networks using either a binary cross-entropy (BCE) loss or a Gaussian SNR loss, and individual and global p-values are estimated by Monte Carlo sampling under the null. Under Gaussian-likelihood, large-sample, and non-shifting-MLE assumptions, the localized statistics are shown to reduce to matched-filter SNRs and the aggregate to a chi-square statistic. The method is demonstrated on a white-noise toy model with several distortion correlation scales and on GW150914, where no significant misspecification is reported.

Significance. The framework is a useful step toward unified anomaly detection and model validation in SBI, and the Monte Carlo p-value calibration is a real strength: it guarantees valid tests under H0 for any learned statistic, even when the analytic connections break down. The derivations in Appendix B are careful and explicitly conditioned, and the toy-model comparisons in Figures 6 and 7 provide both validation and an honest picture of the limits. The public code (once the repository URL is completed) should aid reproducibility. The main weakness is that the interpretability of the learned statistics as classical matched-filter and chi-square quantities rests on Eq. B7, a condition whose domain of validity is not quantified; the authors should scope their claims more carefully and provide a diagnostic or quantitative criterion for when that condition holds.

major comments (1)
  1. [II.4, Eq. (9), Fig. 7, Appendix B] The paper's headline interpretability result, t_i(x) ~ SNR_i^2(x) + C (Eq. 9, derived as Eq. B18) and t_sum(x) = chi^2 + const (Eqs. 11/B22), is derived under Eq. B7, which requires the base-model MLE to be essentially unchanged by the distortion. This is a geometric restriction rather than a large-sample effect: for a linear model mu(Theta)=M Theta it reads M^T Sigma^{-1} n(i) = 0. The paper's own Fig. 7 and Appendix D show that when the MLE absorbs part of the distortion, the analytic profiled statistic and the neural marginalized statistic differ substantially, so the matched-filter and chi-square identifications are not valid in the regime that the global model-checking leg targets. I recommend that the authors (i) state explicitly in Section II.4 that Eq. (9) holds only in the orthogonal-distortion limit and refer to Fig. 7; (ii) provide a quantitative diagnostic for when Eq. B7 is approximately satisfied, for example by comparing the model's ability to absorb the distortion (e.g., ||M^T Sigma^{-1} n(i)||) with the distortion's SNR; and (iii) adjust the 'Principled' conclusion in Section VI so that the classical-statistics interpretation is presented as a limiting case. This issue does not invalidate the testing framework, since the Monte Carlo p-values remain calibrated for any network statistic, but it is load-bearing for the paper's interpretability claim.
minor comments (7)
  1. [Eq. (12) and Eq. (B23)] The relation for b appears to be misprinted. From Eq. B16, sigma_epsilon* = 1/sqrt(n^T Sigma^{-1} n), so b = SNRmax sigma_epsilon* should place the square root in the denominator. Please check Eq. (12) and Eq. B23 for consistency.
  2. [Eq. (5)] The summation in the aggregated test statistic runs from i = 0 to Nalt, but t_0 is not defined; it should presumably start at i = 1.
  3. [Appendix C, Eq. C2] The proportional relation t_hat_i(x) proportional to epsilon_{i,phi}/sigma_{i,phi} is inconsistent with Eq. B18, which involves (epsilon/sigma)^2 plus a constant. The notation is ambiguous and should be corrected to avoid implying that the test statistic is the SNR itself rather than its square.
  4. [Section II.3] The global p-value procedure is described in words; a precise algorithm or pseudocode would help, and the number of Monte Carlo samples Nmc used in the examples should be stated, since the reported global p-values (e.g., 7.57 x 10^{-3} in Fig. 2) imply a finite resolution.
  5. [Code availability] The code repository is given as "/gtbNoemiAM/mist"; the full URL (for example, https://github.com/gtbNoemiAM/mist) should be provided.
  6. [Section IV.2] The sentence noting that posterior-predictive samples 'will slightly overestimate distortions that are degenerate with the effect of model parameter changes' is ambiguous; please clarify whether this makes the test conservative or anti-conservative for such distortions.
  7. [Abstract] The phrase 'We also make explicit analytic connections to classical techniques' could be qualified with 'in limiting cases' earlier in the sentence, to align the abstract with the conditional treatment in the body, though the abstract already contains the caveat 'where they are valid' at the end.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the analytic matched-filter and chi-square connections are derived from explicit assumptions, and the neural test statistics are validated against independent profiled expectations and Monte Carlo null sampling.

full rationale

The paper's central analytic results (ti(x) ≃ SNR_i^2(x) + C, Eq. 9/B18, and tsum = chi^2 + const, Eq. 11/B22) are derived in Appendix B as an approximation to the marginal likelihood ratio under stated assumptions: Gaussian likelihood, large-sample limit, and Eq. B7, which requires the base-model MLE to be insensitive to the distortion. These are genuine analytic derivations from the likelihood-ratio definition, not fitted parameters renamed as predictions. The failure of Eq. B7 for distortions that the MLE partially absorbs is explicitly acknowledged in Appendix D and Figure 7, where the paper states that the mismatch is expected and that there is no reason the profiled analytic and marginalized neural statistics should agree when the MLE is significantly shifted. This is a stated validity limitation, not a circular step. The significance estimates in Section II.3 are Monte Carlo p-values computed under the null hypothesis from simulations, so no real-data fitting enters the calibration; the global p-value procedure is a standard trials-correction using null samples. The SNR training strategy (Appendix C) does define the network output as an estimate of the matched-filter amplitude and variance, so its agreement with the SNR interpretation is partly by construction, but the paper does not present this as an independent prediction; it is presented as an equivalent training objective motivated by the independently derived Eq. B18, and the BCE classifier strategy (Appendix A) provides an independent estimate of the same likelihood-ratio statistic that is compared against the analytic expectation in Figures 6 and 7. The authors' self-citations (e.g., Refs. 6, 12, 31) are contextual applications or practical suggestions about prior truncation; none of them carries the load-bearing uniqueness or derivation claims of this paper. Overall, the derivation chain is self-contained and the central claims reduce neither to fitted inputs nor to a self-citation chain.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

No invented physical entities. The central method rests on standard SBI approximations; the analytic classical limits require Gaussian likelihood, large samples, and Eq. B7. Free parameters are mostly the user-chosen distortion library and amplitude scale.

free parameters (4)
  • Distortion amplitude bound b = Adaptively set: b = SNRmax * sigma; toy converges near b ~ 4.2 after training
    Defines the scale of alternative hypotheses in Eq. (8); affects power and the constant C in Eq. (B18), though p-values are Monte Carlo calibrated.
  • SNRmax = 5 (Section III.1)
    User-chosen maximum SNR used to set b in Eq. (12); the 'self-calibrating' scheme still depends on this choice.
  • Distortion kernel sizes / correlation scales = Toy A=5, B=21, C=61 bins; GW correlated distortion = 11 processed time steps
    The library of alternative hypotheses is hand-chosen; tests can only find distortions that resemble the chosen augmentations.
  • Analysis window and processing choices for GW = 4 s window, 20-1024 Hz bandpass, 0.1 Hz notches, downsampling factor 8
    Chosen following jimgw and the paper's processing; these affect the exact null distribution but not the framework's logic.
assumptions (5)
  • domain assumption Large-sample Gaussian approximation of the base likelihood in model parameters (Eq. B3).
    Used in Appendix B to connect marginal and profile likelihoods; requires enough data for the Laplace approximation to hold.
  • domain assumption The MLE of the base model parameters is not significantly shifted by the distortion, Theta*x is approximately Theta*(x - epsilon n(i)) (Eq. B7).
    This is load-bearing for the matched-filter and chi-square identifications; Figure 7 shows the analytic and neural results diverge when the MLE absorbs the distortion.
  • domain assumption The base likelihood is Gaussian in data space (Eq. B13).
    Needed only for the analytic limiting results, not for the general BCE-based framework.
  • standard math Calibrated binary classifiers estimate likelihood ratios (Appendix A, from Ref. [17]).
    The BCE training strategy relies on this existing result; it is not derived in the paper.
  • standard math Monte Carlo p-values from null samples are uniform and correctly calibrated (Eq. 6).
    Standard property assumed for the individual and global significance estimates.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tests for model misspecification in simulation-based inference: from local distortions to global model checks." pith.science (2026). https://pith.science/paper/N2DAKNW7

@misc{pith2026241215100,
  author       = {Pith},
  title        = {Pith review of: Tests for model misspecification in simulation-based inference: from local distortions to global model checks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N2DAKNW7}},
  note         = {Machine review of arXiv:2412.15100}
}
read the original abstract

Model misspecification analysis strategies, such as anomaly detection, model validation, and model comparison are a key component of scientific model development. Over the last few years, there has been a rapid rise in the use of simulation-based inference (SBI) techniques for Bayesian parameter estimation, applied to increasingly complex forward models. To move towards fully simulation-based analysis pipelines, however, there is an urgent need for a comprehensive simulation-based framework for model misspecification analysis. In this work, we provide a solid and flexible foundation for a wide range of model discrepancy analysis tasks, using distortion-driven model misspecification tests. From a theoretical perspective, we introduce the statistical framework built around performing many hypothesis tests for distortions of the simulation model. We also make explicit analytic connections to classical techniques: anomaly detection, model validation, and goodness-of-fit residual analysis. Furthermore, we introduce an efficient self-calibrating training algorithm that is useful for practitioners. We demonstrate the performance of the framework in multiple scenarios, making the connection to classical results where they are valid. Finally, we show how to conduct such a distortion-driven model misspecification test for real gravitational wave data, specifically on the event GW150914.

Figures

Figures reproduced from arXiv: 2412.15100 by the authors.

Figure 1
Figure 1. FIG. 1. Summary illustration of the presented framework for tests of model misspecification in SBI (see Section [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Comprehensive summary of the framework results for the instructive example presented in Section [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Same as Figure [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Illustration of the adaptive training of distortion amplitudes in our framework. The figure shows how the generated [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Similar to Figures [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. A comparison of the training strategies and analytical expectation for correlated distortions. Each column is for data [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Mitigating Model Misspecification in Simulation-Based Inference for Galaxy Clustering

    astro-ph.CO 2025-07 conditional novelty 5.0 of 10

    A two-step method (coefficient pruning plus learned robust transformation) fixes model misspecification in the SimBIG wavelet-scattering analysis of BOSS galaxy clustering and produces tight Lambda-CDM constraints.

Reference graph

Works this paper leans on

58 extracted references · 19 canonical work pages · cited by 1 Pith paper

  1. [1]

    p-values for anomaly detection

    work page

  2. [2]

    p-value for model validation Global p-value of all tests

    work page

  3. [3]

    bump-hunt

    Residual analysis 2) Residual variance analysisInsight FIG. 1. Summary illustration of the presented framework for tests of model misspecification in SBI (see Section II for details). Left panel : An ensemble of localized test statistics is learned by neural networks (see Appendix A for details); they are typically more sensitive towards isolated distorti...

    work page

  4. [4]

    Skilling, Bayesian Analysis 1, 833 (2006)

    J. Skilling, Bayesian Analysis 1, 833 (2006)

    2006

  5. [5]

    Neyman and E

    J. Neyman and E. S. Pearson, Philosophical Transactions of the Royal Society of London. Series A, Containing Pa- pers of a Mathematical or Physical Character 231, 289 (1933)

    work page 1933

  6. [6]

    T. A. Collaboration, Physics Letters B 716, 1 (2012)

    work page 2012

  7. [7]

    Cranmer, J

    K. Cranmer, J. Brehmer, and G. Louppe, Proceedings of the National Academy of Sciences 117, 30055 (2020), arXiv:1911.01429 [cs, stat]

    arXiv 2020

  8. [8]

    Mishra-Sharma and K

    S. Mishra-Sharma and K. Cranmer, Physical Review D 105, 063017 (2022), arXiv:2110.06931 [astro-ph.HE]

    arXiv 2022

Show all 58 references
  1. [9]

    Anau Montel, A

    N. Anau Montel, A. Coogan, C. Correa, K. Karchev, and C. Weniger, Monthly Notices of the Royal Astronomi- cal Society 518, 2746 (2022), arXiv:2205.09126 [astro- ph.CO]

    2022 arXiv

  2. [10]

    Tucci and F

    B. Tucci and F. Schmidt, Journal of Cosmology and Astroparticle Physics 05, 063 (2024), arXiv:2310.03741 [astro-ph.CO]

    2024 arXiv

  3. [11]

    Fast likelihood-free inference in the LSS Stage IV era,

    G. F. Abell´ an, G. Ca˜ nas Herrera, M. Martinelli, O. Savchenko, D. Sciotti, and C. Weniger, “Fast likelihood-free inference in the LSS Stage IV era,” (2024), arXiv:2403.14750 [astro-ph.CO]

    2024 arXiv

  4. [12]

    Alsing, T

    J. Alsing, T. Charnock, S. Feeney, and B. Wandelt, Monthly Notices of the Royal Astronomical Society 488, 4440 (2019), arXiv:1903.00007 [astro-ph.CO]

    2019 arXiv

  5. [13]

    Sensitivity Analysis of Simulation-Based Inference for Galaxy Clustering,

    C. Modi, S. Pandey, M. Ho, C. Hahn, B. R.- S. Blancard, and B. Wandelt, “Sensitivity Analysis of Simulation-Based Inference for Galaxy Clustering,” (2023), arXiv:2309.15071 [astro-ph.CO]

    2023 arXiv

  6. [14]

    Karchev and R

    K. Karchev and R. Trotta, (2024), arXiv:2409.03837 [astro-ph.CO]

    2024 arXiv

  7. [15]

    Bhardwaj, J

    U. Bhardwaj, J. Alvey, B. K. Miller, S. Nissanke, and C. Weniger, Physical Review D 108, 042004 (2023), arXiv:2304.02035 [gr-qc]

    2023 arXiv

  8. [16]

    M. Dax, S. R. Green, J. Gair, J. H. Macke, A. Buonanno, and B. Sch¨ olkopf, Phys. Rev. Lett.127, 241103 (2021), arXiv:2106.12594 [gr-qc]

    2021 arXiv

  9. [17]

    Jeffrey and B

    N. Jeffrey and B. D. Wandelt, Machine Learning: Sci- ence and Technology 5, 015008 (2024), arXiv:2305.11241 [astro-ph, stat]

    2024 arXiv

  10. [18]

    Gessey-Jones and W

    T. Gessey-Jones and W. J. Handley, (2023), arXiv:2309.06942 [astro-ph.IM]

    2023 arXiv

  11. [19]

    A. S. Mancini, M. M. Docherty, M. A. Price, and J. D. McEwen, RAS Techniques and Instruments 2, 710 (2023), arXiv:2207.04037 [astro-ph, physics:physics]

    2023 arXiv

  12. [20]

    Approximat- ing likelihood ratios with calibrated discriminative clas- sifiers,

    K. Cranmer, J. Pavez, and G. Louppe, “Approximat- ing likelihood ratios with calibrated discriminative clas- sifiers,” (2016), arXiv:1506.02169 [stat.AP]

    2016 arXiv

  13. [21]

    Confidence sets and hypothesis testing in a likelihood-free inference setting,

    N. Dalmasso, R. Izbicki, and A. B. Lee, “Confidence sets and hypothesis testing in a likelihood-free inference setting,” (2020), arXiv:2002.10399 [stat.ME]

    2020 arXiv

  14. [22]

    Learning Optimal Test Statistics in the Presence of Nuisance Parameters,

    L. Heinrich, “Learning Optimal Test Statistics in the Presence of Nuisance Parameters,” (2022), arXiv:2203.13079 [physics, stat]

    2022 arXiv

  15. [23]

    KiDS-SBI: Simulation-Based Inference Analysis of KiDS-1000 Cosmic Shear,

    M. von Wietersheim-Kramsta, K. Lin, N. Tessore, B. Joachimi, A. Loureiro, R. Reischke, and A. H. Wright, “KiDS-SBI: Simulation-Based Inference Analysis of KiDS-1000 Cosmic Shear,” (2024)

    2024

  16. [24]

    White, Econometrica 50, 1 (1982), publisher: [Wiley, Econometric Society]

    H. White, Econometrica 50, 1 (1982), publisher: [Wiley, Econometric Society]

    1982

  17. [25]

    Cannon, D

    P. Cannon, D. Ward, and S. M. Schmon, Investi- gating the Impact of Model Misspecification in Neural 12 Simulation-based Inference, Tech. Rep. arXiv:2209.01845 (arXiv, 2022) arXiv:2209.01845 [cs, stat] type: article

    2022 arXiv

  18. [26]

    Detecting Model Misspecification in Amortized Bayesian Inference with Neural Networks: An Extended Investigation,

    M. Schmitt, P.-C. B¨ urkner, U. K¨ othe, and S. T. Radev, “Detecting Model Misspecification in Amortized Bayesian Inference with Neural Networks: An Extended Investigation,” (2024), arXiv:2406.03154

    2024 arXiv

  19. [27]

    Huang, A

    D. Huang, A. Bharti, A. Souza, L. Acerbi, and S. Kaski, Advances in Neural Information Processing Systems 36, 7289 (2023)

    2023

  20. [28]

    Addressing Misspecification in Simulation-based Inference through Data-driven Calibration,

    A. Wehenkel, J. L. Gamella, O. Sener, J. Behrmann, G. Sapiro, M. Cuturi, and J.-H. Jacobsen, “Addressing Misspecification in Simulation-based Inference through Data-driven Calibration,” (2024)

    2024

  21. [29]

    D. Ward, P. Cannon, M. Beaumont, M. Fasiolo, and S. Schmon, Advances in Neural Information Processing Systems 35, 33845 (2022)

    2022

  22. [30]

    Dellaporta, J

    C. Dellaporta, J. Knoblauch, T. Damoulas, and F.- X. Briol, in Proceedings of The 25th International Con- ference on Artificial Intelligence and Statistics (PMLR,

  23. [31]

    Scalable inference with Autoregressive Neural Ratio Estimation,

    N. Anau Montel, J. Alvey, and C. Weniger, “Scalable inference with Autoregressive Neural Ratio Estimation,” (2023), arXiv:2308.08597 [astro-ph.IM]

    2023 arXiv

  24. [32]

    Practical statistics for the lhc,

    K. Cranmer, “Practical statistics for the lhc,” (2015), arXiv:1503.07622 [physics.data-an]

    2015 arXiv

  25. [33]

    Truncated Marginal Neural Ratio Estima- tion,

    B. K. Miller, A. Cole, P. Forr´ e, G. Louppe, and C. Weniger, “Truncated Marginal Neural Ratio Estima- tion,” (2021)

    2021

  26. [34]

    Se- quential neural likelihood: Fast likelihood-free inference with autoregressive flows,

    G. Papamakarios, D. C. Sterratt, and I. Murray, “Se- quential neural likelihood: Fast likelihood-free inference with autoregressive flows,” (2019), arXiv:1805.07226 [stat.ML]

    2019 arXiv

  27. [35]

    B. P. Abbott et al. (LIGO Scientific, Virgo), Phys. Rev. Lett. 116, 241102 (2016), arXiv:1602.03840 [gr-qc]

    2016 arXiv

  28. [36]

    S. S. Wilks, The Annals of Mathematical Statistics 9, 60 (1938)

    1938

  29. [37]

    Gross and O

    E. Gross and O. Vitells, The European Physical Journal C 70, 525–530 (2010)

    2010

  30. [38]

    B. P. Abbott et al. (LIGO Scientific, Virgo), Phys. Rev. Lett. 116, 061102 (2016), arXiv:1602.03837 [gr-qc]

    2016 arXiv

  31. [39]

    K. K. H. Lam, K. W. K. Wong, and T. D. P. Edwards, Phys. Rev. D 109, 124009 (2024), arXiv:2306.17245 [gr- qc]

    2024 arXiv

  32. [40]

    B. P. Abbott et al. (LIGO Scientific, Virgo), Phys. Rev. D 93, 122003 (2016), arXiv:1602.03839 [gr-qc]

    2016 arXiv

  33. [41]

    B. P. Abbott et al. (LIGO Scientific, Virgo), Class. Quant. Grav. 34, 104002 (2017), arXiv:1611.07531 [gr- qc]

    2017 arXiv

  34. [42]

    Gamba, M

    R. Gamba, M. Breschi, S. Bernuzzi, M. Agathos, and A. Nagar, Phys. Rev. D 103, 124015 (2021), arXiv:2009.08467 [gr-qc]

    2021 arXiv

  35. [43]

    Abbott et al

    R. Abbott et al. (LIGO Scientific, Virgo), SoftwareX 13, 100658 (2021), arXiv:1912.11716 [gr-qc]

    2021 arXiv

  36. [44]

    B. P. Abbott et al. (LIGO Scientific, Virgo), Phys. Rev. Lett. 116, 131103 (2016), arXiv:1602.03838 [gr-qc]

    2016 arXiv

  37. [45]

    Legin, M

    R. Legin, M. Isi, K. W. K. Wong, Y. Hezaveh, and L. Perreault-Levasseur, (2024), arXiv:2410.19956 [astro- ph.IM]

    2024 arXiv

  38. [46]

    S. B. Coughlin et al. , Phys. Rev. D 99, 082002 (2019), arXiv:1903.04058 [astro-ph.IM]

    2019 arXiv

  39. [47]

    S. Khan, S. Husa, M. Hannam, F. Ohme, M. P¨ urrer, X. Jim´ enez Forteza, and A. Boh´ e, Phys. Rev. D 93, 044007 (2016), arXiv:1508.07253 [gr-qc]

    2016 arXiv

  40. [48]

    K. W. K. Wong, M. Isi, and T. D. P. Edwards, Astro- phys. J. 958, 129 (2023), arXiv:2302.05333 [astro-ph.IM]

    2023 arXiv

  41. [49]

    T. D. P. Edwards, K. W. K. Wong, K. K. H. Lam, A. Coogan, D. Foreman-Mackey, M. Isi, and A. Zimmer- man, Phys. Rev. D 110, 064028 (2024), arXiv:2302.05329 [astro-ph.IM]

    2024 arXiv

  42. [50]

    S. Husa, S. Khan, M. Hannam, M. P¨ urrer, F. Ohme, X. Jim´ enez Forteza, and A. Boh´ e, Phys. Rev. D 93, 044006 (2016), arXiv:1508.07250 [gr-qc]

    2016 arXiv

  43. [51]

    Lee, J.-h

    J.-Y. Lee, J.-h. Kim, M. Jung, B. K. Oh, Y. Jo, S. Park, J. Lee, Y.-S. Ting, and H. S. Hwang, The Astrophysical Journal 975, 38 (2024)

    2024

  44. [52]

    Cosmology with one galaxy? – the astrid model and robustness,

    N. Echeverri, F. Villaescusa-Navarro, C. Chawak, Y. Ni, C. Hahn, E. Hernandez-Martinez, R. Teyssier, D. Angles- Alcazar, K. Dolag, and T. Castro, “Cosmology with one galaxy? – the astrid model and robustness,” (2023), arXiv:2304.06084 [astro-ph.CO]

    2023 arXiv

  45. [53]

    N. S. M. de Santi, H. Shao, F. Villaescusa-Navarro, L. R. Abramo, R. Teyssier, P. Villanueva-Domingo, Y. Ni, D. Angl´ es-Alc´ azar, S. Genel, E. Hern´ andez-Mart ´ ınez, U. P. Steinwandel, C. C. Lovell, K. Dolag, T. Castro, and M. Vogelsberger, The Astrophysical Journal 952, 69 (2023)

    2023

  46. [54]

    Domain-adaptive neural posterior esti- mation for strong gravitational lens analysis,

    P. Swierc, M. Tamargo-Arizmendi, A. ´Ciprijanovi´ c, and B. D. Nord, “Domain-adaptive neural posterior esti- mation for strong gravitational lens analysis,” (2024), arXiv:2410.16347 [astro-ph.IM]

    2024 arXiv

  47. [56]

    Cosmological analysis with calibrated neu- ral quantile estimation and approximate simulators,

    H. Jia, “Cosmological analysis with calibrated neu- ral quantile estimation and approximate simulators,” (2024), arXiv:2411.14748 [astro-ph.CO]

    2024 arXiv

  48. [57]

    Robustness of neural ratio and posterior estimators to distributional shifts for population-level dark mat- ter analysis in strong gravitational lensing,

    A. Filipp, Y. Hezaveh, and L. Perreault-Levasseur, “Robustness of neural ratio and posterior estimators to distributional shifts for population-level dark mat- ter analysis in strong gravitational lensing,” (2024), arXiv:2411.05905 [astro-ph.CO]

    2024

  49. [58]

    Cross-entropy loss functions: Theoretical analysis and applications,

    A. Mao, M. Mohri, and Y. Zhong, “Cross-entropy loss functions: Theoretical analysis and applications,” (2023), arXiv:2304.07288 [cs.LG]. 13 APPENDICES Appendix A describes a general training strategy for our model misspecification testing framework. An alternative training str...

    2023 arXiv

  50. [2022]

    943–970, iSSN: 2640-3498

    pp. 943–970, iSSN: 2640-3498

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.