Pith. sign in

REVIEW 4 major objections 5 minor 49 references

Feldman-Cousins' ML Cousin: Sterile Neutrino Global Fits using Simulation-Based Inference

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A machine-learning pipeline can run sterile neutrino global fits about 200 times faster, the authors report.

desk verdict Honest and useful SBI proof-of-concept for sterile neutrino global fits, but the 'full-fledged Bayesian analysis' claim outruns what is validated; the speedup is real, the credibility regions are not yet. read the letter →

arxiv 2501.08988 v1 pith:KFB3BLNK submitted 2025-01-15 hep-ex

Joshua Villarreal , John M. Hardin , Janet M. Conrad This is my paper · ORCID
classification hep-ex
keywords simulation-basedinferencesterileneutrinosglobalfitsFeldman-CousinsmethodnormalizingflowsBayesianneuralnetworkscredibilityregionsneutrinooscillation
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

For small-signal particle physics analyses, Wilks' theorem can fail, and the standard fix, the Feldman-Cousins procedure, becomes computationally intractable when many experiments are combined. This paper claims that a simulation-based inference pipeline, dubbed FCMLC, can approximate a full Bayesian analysis about 200 times faster, reducing a sterile neutrino global fit from roughly 510,000 to 2,500 computing hours. The pipeline replaces the high-dimensional data-to-parameter problem with a two-step density estimate: a stochastic neural network predicts the oscillation parameters, and a normalizing flow learns the conditional distribution of true parameters given those predictions. The authors demonstrate the method on a joint analysis of electron neutrino and antineutrino disappearance experiments under a 3+1 sterile neutrino model, reproducing the qualitative features of published single-experiment limits and of a Wilks-based global fit. If the claim holds, it makes full global fits feasible for small-signal searches where Feldman-Cousins is too expensive.

What carries the argument

The load-bearing object is the posterior identity $p(y|x) \approx \int p(y|\hat{y})\, p(\hat{y}|x)\, d\hat{y}$, which follows from the law of total probability together with the approximation that the neural network's predicted parameters $\hat{y}$ screen off the data $x$ from the true parameters $y$. The machinery has three parts: a dropout-at-inference neural network that turns one experimental data vector into many randomized parameter predictions, a kernel density estimate that converts those predictions into $p(\hat{y}|x)$, and a normalizing flow, an invertible neural transformation of a base distribution into a flexible conditional density, that supplies $p(y|\hat{y})$. Numerical integration over a 50 by 50 grid in $\hat{y}$ and thresholding of the resulting $p(y|x)$ builds the credibility regions used for fits and exclusions.

What would settle it

Compare two mock data sets that the trained network maps to the same predicted parameters but that were generated from different true model points; if the FCMLC posteriors for the two data sets differ materially, the conditional-independence approximation $p(y|\hat{y},x) \approx p(y|\hat{y})$ fails. A more direct version would bin test samples by the predicted parameters and check whether the distribution of true parameters is independent of the original data within each bin; a systematic dependence would falsify the method's core approximation.

Watch

Extended reading notes

Core claim

The paper's central claim is that the posterior density over the true oscillation parameters $y$ given experimental data $x$ can be written, under the conditional-independence approximation $p(y|\hat{y},x) \approx p(y|\hat{y})$, as $p(y|x) \approx \int p(y|\hat{y})\, p(\hat{y}|x)\, d\hat{y}$. Here $\hat{y}$ are the parameters predicted by a dropout-at-inference neural network, $p(\hat{y}|x)$ is estimated by a kernel density estimate over repeated network calls, and $p(y|\hat{y})$ is learned by a normalizing flow. Credibility regions are obtained by numerically solving for the level set of this approximate posterior that contains the desired probability mass. The authors show that for STEREO-only, BEST-only, and combined electron-flavor disappearance fits, FCMLC produces contours that qualitatively match the published STEREO exclusion, the BEST allowed region, and a Wilks-based global fit, including the gallium-anomaly region, while reducing the total runtime by a factor of about 200 relative to a Feldman-Cousins treatment.

Load-bearing premise

The method assumes the network's predicted parameters capture all the information in the data that matters for the true parameters, so that once the prediction is known, the original data tells you nothing more about the physics.

Editorial extensions

If this is right

  • Sterile neutrino global fits on electron neutrino and antineutrino disappearance data become practical, taking about 2,500 hours instead of about 510,000 hours for the same realization generation.
  • Single-experiment fits, and the extension from one experiment to a global combination, add little runtime once the neural network and normalizing flow are trained.
  • FCMLC reproduces the qualitative shapes of the published STEREO and BEST results and of the Wilks-based global fit, including the parameter region associated with the gallium anomaly.
  • Because every component is modular, the same recipe can be applied to other small-signal global fits where Wilks' theorem fails and Feldman-Cousins is too costly.
  • The fast posterior can serve as a first pass to localize parameter space before running a higher-fidelity Feldman-Cousins analysis.

Reading between the lines

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

  • The conditional-independence assumption is effectively a sufficiency claim about the learned network output. A direct test would be to compare FCMLC posteriors for two data vectors that map to the same predicted parameters; any substantial difference would indicate that $x$ carries information beyond $\hat{y}$, and the coverage mismatch seen in the paper's Fig. 4 may already be a symptom of this.
  • Because the training grid is uniform in log-space and the authors note that exclusion curves depend strongly on that choice, the credibility regions should be read as prior-dependent summaries rather than frequentist confidence intervals until a physically motivated prior is introduced.
  • The current demonstration is restricted to two oscillation parameters; extending to all 3+1 channels would test whether the normalizing flow's density estimate and the conditional-independence approximation survive in higher dimension, and could be done with modest changes to the pipeline.
  • One could calibrate FCMLC coverages empirically, adjusting credibility thresholds until observed coverage matches nominal coverage, to obtain frequentist-style limits at a fraction of the cost of a full Feldman-Cousins calculation.
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

4 major / 5 minor

Summary. The paper introduces FCMLC, a simulation-based inference procedure intended to approximate Bayesian posterior distributions for global fits in particle physics. A dropout neural network maps each experimental realization x to a compact prediction y_hat, a KDE converts repeated network calls into p(y_hat|x), a normalizing flow learns p(y|y_hat), and Eq. (2) integrates over y_hat to estimate p(y|x). The method is demonstrated on electron neutrino/antineutrino disappearance data from STEREO, BEST, and a combined global fit, and the authors claim a roughly 200-fold runtime reduction relative to a Feldman-Cousins analysis. The paper is explicitly framed as a proof-of-concept, and it honestly lists several limitations, including imperfect coverage and the training grid acting as an implicit prior.

Significance. If the method's validity were established, it would offer a practical way to perform approximate global fits in regimes where Feldman-Cousins is computationally prohibitive. The qualitative agreement with STEREO's exclusion curve, BEST's allowed region, and the Wilks-based global fit is encouraging and suggests that the architecture can capture relevant structure. The paper is also transparent about its limitations, which is a strength. However, the central claim of approximating a 'full-fledged Bayesian analysis' is not yet supported by the presented validation, primarily because the sufficiency assumption behind Eq. (2) is untested and the coverage checks show miscalibration.

major comments (4)
  1. [II B, Eq. (2)] The simplification p(y|y_hat,x) ≈ p(y|y_hat) asserts that the neural-network output y_hat is a sufficient statistic of the data x for the parameters y. This is the load-bearing step of the method, yet no diagnostic is provided to test whether the network output captures all relevant information. The coverage checks in Sec. III C and Fig. 4 do not validate this sufficiency: they sample y from the training prior and can show good marginal coverage even when the conditional posterior is biased, because overcoverage in some regions compensates undercoverage in others. The visible deviation from the 45-degree line in Fig. 4 is consistent with this concern. The abstract's characterization of the result as 'a full-fledged Bayesian analysis' is therefore not supported by the evidence presented.
  2. [IV B] The authors state that exclusion curves drawn from the estimated posterior depend strongly on the choice of training grid, which acts as an implicit prior. This undermines the interpretation of the drawn regions as Bayesian credibility regions under a physically motivated prior. Moreover, the coverage-check procedure in Sec. II C 2 samples parameters from the same log-uniform grid, so it demonstrates only that the method is self-consistent with its training distribution, not that it produces correct posterior inference for arbitrary physical parameters. The paper should either recalibrate the procedure to yield valid frequentist or Bayesian coverage over the parameter space of interest, or rescope the claims to describe the output as an approximate, prior-dependent surrogate rather than a Bayesian posterior.
  3. [III C, Fig. 4] The coverage miscalibration is acknowledged but only conjecturally attributed to instability of low-credibility contours. Because the paper's value proposition is speed combined with trustworthy inference regions, the miscalibration needs to be quantified and, if possible, mitigated—for example by posterior recalibration or by restricting claims to the high-credibility region where the authors argue the agreement is adequate. Without such analysis, the reader cannot assess whether the 200x speedup is achieved at an acceptable cost in statistical validity.
  4. [IV A, Table II] The runtime comparison in Table II is plausible, but the headline factor of ~200 compares a frequentist Feldman-Cousins confidence-interval construction with a procedure that produces approximate posterior regions under an implicit training-grid prior. These are not equivalent inference outputs. The paper should state more explicitly what the speedup is buying and how the FCMLC regions should be interpreted relative to frequentist confidence intervals, especially because the method does not provide a direct frequentist coverage guarantee. This is essential for readers who may use the runtime figure to decide whether FCMLC can replace or supplement Feldman-Cousins analyses.
minor comments (5)
  1. [I, Introduction] There are typos such as 'frequentest' for 'frequentist' and 'uncetainties' for 'uncertainties'; the paper would benefit from a careful proofread.
  2. [II B] The sentence 'We alter replaced the conditional KDE with a normalizing flow' contains a grammatical error; it should read 'We therefore replaced the conditional KDE with a normalizing flow'.
  3. [II C 2] The description of the coverage-check prior states that parameters are sampled 'uniformly in the logarithm of each parameter'; it would be helpful to state explicitly that this matches the training-grid prior used in Sec. II A, since that prior is central to the interpretation of the coverage results.
  4. [IV B] The sentence 'we MC samples were generated from grid points uniformly separated in log10 Ue4 and log10 ∆m2_41 space' is awkwardly phrased and should be rewritten for clarity.
  5. [Fig. 2] The notation 'R² = 0' in the discussion of Fig. 2 is not defined; a brief explanation of what R² refers to in this context would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: FCMLC is a supervised posterior surrogate whose grid-prior and sufficiency assumptions are stated limitations, not definitional reductions of the claimed result to its own inputs.

full rationale

Walking the derivation chain, the FCMLC posterior is built from Eq. 2 via the explicit approximation p(y|ŷ,x) ≈ p(y|ŷ). The term p(y|ŷ) is estimated once from training pairs (y,ŷ), and p(ŷ|x) is obtained from a KDE over repeated dropout-network outputs for the data x. This is a standard simulation-based posterior surrogate, mathematically distinct from its inputs: it is a learned conditional density, not a restatement of the training grid or the simulator. The paper's own Sec. IV B concedes that the training grid acts as an implicit prior and that exclusion curves depend strongly on that choice; this is a prior-dependence and calibration limitation, not circularity, because the posterior still depends on the data through the network and KDE. The coverage checks in Sec. III C reuse the same sblmc generator for both training and validation, but this is standard calibration practice and is supplemented by external benchmarks: the fcmlc contours are compared to published STEREO and BEST exclusion/allowed regions and to a Wilks-based global fit from Ref. [30]. Self-citations to Refs. [13,20,30] supply motivation and software, but the central claims — a roughly 200x runtime reduction and qualitative agreement with published results — are not forced by those citations; the runtime comparison in Table II is an honest accounting of measured compute time, and the physics comparisons are against independent experimental curves. No equation in the paper reduces by construction to its own input, no fitted parameter is renamed as a prediction, and no uniqueness theorem or ansatz is imported from the authors' prior work to forbid alternatives. The unvalidated sufficiency assumption and the grid-induced prior are genuine correctness risks, but they are limitations acknowledged in the manuscript rather than circular reasoning.

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

The central claim rests on the choice of the training grid as an implicit prior, on the sufficiency of the neural network summary, and on the accuracy of the experimental covariance/pull inputs. No new physical entities are introduced; the free parameters are method hyperparameters and the prior grid.

free parameters (5)
  • Training grid prior over (|Ue4|, Delta m2_41) = uniform in log10 over [0.001, 1/sqrt(2)] x [0.01, 10] eV2
    The posterior estimate is defined relative to this implicit prior; the paper states in Sec. IV B that exclusion curves depend strongly on this choice.
  • MC dropout rate for neural networks = 0.02
    Chosen by hand, listed in Table III for each fit.
  • Normalizing flow dropout rate = 0.10 or 0.15 depending on fit
    Chosen per experiment, Table IV.
  • Number of MC realizations per grid point = 50
    Fixed by hand; total 510,050 samples, Sec. II A.
  • Neural network layer sizes = (100,100,50,25,10) or (150,100,50,25,25)
    Chosen without optimization; Tables III and IV.
assumptions (5)
  • domain assumption The 3+1 sterile neutrino survival probability P = 1 - 4|Ue4|^2(1 - |Ue4|^2) sin^2(Delta_41 L/E) (Eq. 1) is the true model.
    The whole fit assumes this minimal sterile neutrino model; from Sec. I A.
  • domain assumption Experimental covariance matrices and pull parameters supplied by the experiments accurately describe the real data statistics.
    Data generation in Sec. II A relies on these cost functions; if inaccurate, all realizations are wrong.
  • ad hoc to paper y is conditionally independent of x given the neural network summary y_hat, i.e., p(y|y_hat,x) approximately equals p(y|y_hat).
    This is the key simplifying assumption in Sec. II B that makes Eq. 2 tractable; it is not validated.
  • ad hoc to paper The training distribution over parameters acts as the Bayesian prior.
    Sec. IV B notes the sampling grid is dual to prior selection and affects exclusion curves.
  • standard math Standard calculus of probability (law of total probability) and normalizing flow density estimation are valid.
    Background math used in Eq. 2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Feldman-Cousins' ML Cousin: Sterile Neutrino Global Fits using Simulation-Based Inference." pith.science (2026). https://pith.science/paper/KFB3BLNK

@misc{pith2026250108988,
  author       = {Pith},
  title        = {Pith review of: Feldman-Cousins' ML Cousin: Sterile Neutrino Global Fits using Simulation-Based Inference},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KFB3BLNK}},
  note         = {Machine review of arXiv:2501.08988}
}
read the original abstract

For many small-signal particle physics analyses, Wilks' theorem, a simplifying assumption that presumes log-likelihood asymptotic normality, does not hold. The most common alternative approach applied in particle physics is a highly computationally expensive procedure put forward by Feldman and Cousins. When many experiments are combined for a global fit to data, deviations from Wilks' theorem are exacerbated, and Feldman-Cousins becomes computationally intractable. We present a novel, machine learning-based procedure that can approximate a full-fledged Bayesian analysis 200 times faster than the Feldman-Cousins method. We demonstrate the utility of this novel method by performing a joint analysis of electron neutrino/antineutrino disappearance data within a single sterile neutrino oscillation framework. Although we present a prototypical simulation-based inference method for a sterile neutrino global fit, we anticipate that similar procedures will be useful for global fits of all kinds, especially those in which Feldman-Cousins is too computationally expensive to use.

Figures

Figures reproduced from arXiv: 2501.08988 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. ), we expect the fcmlc to identify closed contours (that may extend off the plot at high ∆m2 ) around the red dot. In the cases where the ground truth is outside of the computed sensitivity (to the left of the sensitivity line), the fcmlc fit not only draw open contours indicat￾ing that an allowed region is not identified, but contours which exclude regions of parameter space to which the experiment should be sensit… view at source ↗
Figure 3
Figure 3. FIG. 3 [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 [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 [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 39 canonical work pages

  1. [1]

    However, for global fits, it’s crucial that our networks can quantify predictive uncer- tainties meaningfully

    Neural Networks and Prediction Uncertainties Most neural networks (NNs) are limited to making deterministic predictions. However, for global fits, it’s crucial that our networks can quantify predictive uncer- tainties meaningfully. Bayesian neural networks (BNNs) [24, 25] address this challenge by placing distributions on the nodes in the hidden layers, e...

    work page

  2. [2]

    interpolation sample

    Normalizing Flows Normalizing flows (NFs) fall into a class of neural net- works that learn a set of invertible transformations of probability distributions to sample from and estimate the density of a complex posterior distribution. NFs for den- sity estimation were popularized in Ref. [28], and the notation for the operational principle of NFs from this...

    work page

  3. [3]

    I as inputs to the fcmlc procedure and report credibil- ity regions in the model parameter space

    Fits on Observed Experimental Data Once we have established that fcmlc gives reasonable posterior density distribution estimates, we use reported experimental data from the sources outlined in Tab. I as inputs to the fcmlc procedure and report credibil- ity regions in the model parameter space. We compare these with the published exclusion curve generated...

    work page

  4. [4]

    Net- works used for the estimation of the underlying oscilla- tion parameters were created and trained using Python’s tensorflow API (Ref

    We denote this prediction ˆy. Net- works used for the estimation of the underlying oscilla- tion parameters were created and trained using Python’s tensorflow API (Ref. [31]). The neural network archi- tecture is shown in Fig. 6. By calling this neural network m times to predict the underlying oscillation parame- ters, m independent samples of ˆy are made...

    work page

  5. [5]

    Feldman-Cousins’ ML Cousin

    The procedure described in this section is summarized in Fig. 1. Though not a direct ap- proximation of the Feldman-Cousins method, estimating model parameters from a posterior distribution is still an invaluable task for analyzing and understanding exper- imental data; because this procedure satisfies a similar role to the Feldman-Cousins method for unde...

    work page

  6. [6]

    Neural Network Performance To quantify the predictive performance of the fully- connected NNs predicted in this study, we report the training and test-set mean squared error (MSE), binned in each unique Ue4, ∆m2 41, and plotted in two dimensions with respect to the oscillation parameters. It is mislead- ing to report the MSE for the entire training and te...

    work page

  7. [7]

    The procedure used for performing these coverage checks is inspired by Ref

    Coverage Checks Given that the Bayesian CRs we draw on the estimated posterior distribution are used for fit interpretation, it is important that these CRs are reasonable, recovering the true underlying model parameters the correct fraction of the time. The procedure used for performing these coverage checks is inspired by Ref. [36], which defines a corre...

    work page 2000

  8. [8]

    II A), use sblmc to simulate experimental data from said model parameters, and then use fcmlc to build an estimate of the posterior distribution from which CRs can be drawn

    space (uniformly in the logarithm of each pa- rameter, according to the bounds specified in Sec. II A), use sblmc to simulate experimental data from said model parameters, and then use fcmlc to build an estimate of the posterior distribution from which CRs can be drawn. We then compare observed versus expected coverages

    work page

Show all 49 references
  1. [9]

    Barinov V and Gorbunov D 2022 Phys. Rev. D 105(5) L051703 URL https://link.aps.org/doi/10. 1103/PhysRevD.105.L051703

    2022

  2. [10]

    ground truth

    and, the allowed region (assuming Wilks’) computed using the sblmc fitting framework [20, 30]. Finally, we will use fcmlc to draw allowed regions for the oscillation parameters for a complete global fit of the νe disappear- ance data considered in this work. III. RESUL TS A. N...

  3. [11]

    good enough

    = (0 .35, 9.3 eV2) [20]. Devia- tions from Wilks’ assumptions, such as non-Gaussianity in the likelihood or boundary effects in the parameter space drive certainly drive disagreement between the two methods, and further model tuning is likely to improve the SBI fit quality. IV...

  4. [12]

    Almaz´ an H, Bernard L, Blanchet A, Bonhomme A, Buck C, Chalil A, del Amo Sanchez P, El Atmani I, Labit L, Lamblin J, Letourneau A, Lhuillier D, Licciardi M, Lindner M, Materna T, Pessard H, R´ eal J S, Ricol J S, Roca C, Rogly R, Salagnac T, Savu V, Schopp- mann S, Soldner T,...

    2023 doi

  5. [13]

    Andriamirado M, Balantekin A B, Band H R, Bass C D, Bergeron D E, Berish D, Bowden N S, Brodsky J P, 11 Bryan C D, Classen T, Conant A J, Deichert G, Di- wan M V, Dolinski M J, Erickson A, Foust B T, Gaison J K, Galindo-Uribarri A, Gilbert C E, Goddard B W, Hackett B T, Hans S...

    2021

  6. [14]

    org/10.3103/S1068335623120151

    Skrobova N A 2023 Bulletin of the Lebedev Physics In- stitute 50 566–572 ISSN 1934-838X URL https://doi. org/10.3103/S1068335623120151

    2023 doi

  7. [15]

    Atif Z, Choi J H, Han B Y, Jang C H, Jang H I, Jang J S, Jeon E J, Jeon S H, Joo K K, Ju K, Jung D E, Kim B R, Kim H J, Kim H S, Kim J G, Kim J H, Kim J Y, Kim J Y, Kim S B, Kim S Y, Kim W, Kim Y D, Ko Y J, Kwon E, Lee D H, Lee H G, Lee J, Lee J Y, Lee M H, Lim I T, Moon D H, ...

    2022

  8. [16]

    Gavrin V N 2013 Physics of Atomic Nuclei 76 1238– 1243 ISSN 1562-692X URL https://doi.org/10.1134/ S106377881309007X

    2013

  9. [17]

    Kostensalo J, Suhonen J, Giunti C and Srivastava P 2019 Physics Letters B 795 542–547 ISSN 0370- 2693 URL https://www.sciencedirect.com/science/ article/pii/S037026931930440X

    2019

  10. [18]

    Giunti C and Laveder M 2011 Phys. Rev. C 83 065504 (Preprint 1006.3244)

    2011 arXiv

  11. [19]

    Barinov V V, Cleveland B T, Danshin S N, Ejiri H, El- liott S R, Frekers D, Gavrin V N, Gorbachev V V, Gor- bunov D S, Haxton W C, Ibragimova T V, Kim I, Kozlova Y P, Kravchuk L V, Kuzminov V V, Lubsandorzhiev B K, Malyshkin Y M, Massarczyk R, Matveev V A, Mirmov I N, Nico J S...

    2022 doi

  12. [20]

    Hardin J M, Martinez-Soler I, Diaz A, Jin M, Kamp N W, Arg¨ uelles C A, Conrad J M and Shaevitz M H 2023JHEP 09 058 (Preprint 2211.02610)

    arXiv

  13. [21]

    1995 A Combination of preliminary LEP electroweak measurements and constraints on the standard model 17th International Symposium on Lepton Photon Inter- actions

    1995

  14. [22]

    Bernardi G, Carrena M and Junk T 2012 Phys. Rev. D 86 010001 URL https://pdg.lbl.gov/2012/reviews/ rpp2012-rev-higgs-boson.pdf

    2012

  15. [23]

    Gonzalez-Garcia M C, Maltoni M, Salvado J and Schwetz T 2012 JHEP 12 123 (Preprint 1209.3023)

    2012 arXiv

  16. [24]

    Hardin J M 2024 Eur. J. Phys. 45 025806 ( Preprint 2211.06347)

    2024 arXiv

  17. [25]

    Feldman G J and Cousins R D 1998 Phys. Rev. D57(7) 3873–3889 URL https://link.aps.org/doi/10.1103/ PhysRevD.57.3873

    1998

  18. [26]

    Not to be confused with confidence levels, CLs

  19. [27]

    Qian X, Tan A, Ling J J, Nakajima Y and Zhang C 2016 Nucl. Instrum. Meth. A827 63–78 (Preprint 1407.5052)

    2016 arXiv

  20. [28]

    Cranmer K, Brehmer J and Louppe G 2020 Pro- ceedings of the National Academy of Sciences 117 30055–30062 ISSN 1091-6490 URL http://dx.doi.org/ 10.1073/pnas.1912789117

    2020 doi

  21. [29]

    Simulation-based inference https:// simulation-based-inference.org/ accessed: 2024-09- 03

    2024

  22. [30]

    Mishra-Sharma S Awesome neural simulation- based inference https://github.com/smsharma/ awesome-neural-sbi accessed: 2024-09-03

    2024

  23. [31]

    Abadi M, Agarwal A, Barham P, Brevdo E, Chen Z, Citro C, Corrado G S, Davis A, Dean J, Devin M, Ghe- mawat S, Goodfellow I, Harp A, Irving G, Isard M, Jia Y, Jozefowicz R, Kaiser L, Kudlur M, Levenberg J, Man´ e 12 D, Monga R, Moore S, Murray D, Olah C, Schuster M, Shlens J, S...

    2015

  24. [32]

    For fits to electron disappearance experiments only, having |Ue4| ≤1/ √ 2 still fully explores the model param- eter space

    Note that to keep Pνe→νe ∈ [0, 1], we require |Ue4| ∈ [0, 1]. For fits to electron disappearance experiments only, having |Ue4| ≤1/ √ 2 still fully explores the model param- eter space

  25. [33]

    Giunti C, Li Y F, Ternes C A and Xin Z 2022 Phys. Lett. B 829 137054 (Preprint 2110.06820)

    2022 arXiv

  26. [34]

    Elliott S R, Gavrin V N, Haxton W C, Ibragimova T V and Rule E J 2023 Phys. Rev. C 108 035502 ( Preprint 2303.13623)

    2023 arXiv

  27. [35]

    Neal R M 1996 Bayesian Learning for Neural Networks (New York, NY: Springer) ISBN 978-0-387-94724-2

    1996

  28. [36]

    MacKay D J C 1992 Neural Computation4 448–472 ISSN 0899-7667 (Preprint https://direct.mit.edu/neco/article- pdf/4/3/448/812348/neco.1992.4.3.448.pdf) URL https://doi.org/10.1162/neco.1992.4.3.448

    1992 doi

  29. [37]

    Gal Y and Ghahramani Z 2016 Dropout as a bayesian approximation: Representing model uncertainty in deep learning Proceedings of The 33rd International Con- ference on Machine Learning (Proceedings of Machine Learning Research vol 48) ed Balcan M F and Wein- berger K Q (New Yor...

    2016

  30. [38]

    Srivastava N, Hinton G, Krizhevsky A, Sutskever I and Salakhutdinov R 2014 Journal of Machine Learning Re- search 15 1929–1958 URL http://jmlr.org/papers/ v15/srivastava14a.html

    2014

  31. [39]

    Rezende D and Mohamed S 2015 Variational inference with normalizing flows Proceedings of the 32nd Interna- tional Conference on Machine Learning(Proceedings of Machine Learning Researchvol 37) ed Bach F and Blei D (Lille, France: PMLR) pp 1530–1538 URL https: //proceedings.mlr...

    2015

  32. [40]

    Durkan C, Bekasov A, Murray I and Papamakarios G 2019 Neural spline flows Advances in Neural In- formation Processing Systems vol 32 ed Wallach H, Larochelle H, Beygelzimer A, d 'Alch´ e-Buc F, Fox E and Garnett R (Curran Associates, Inc.) URL https: //proceedings.neurips.cc/p...

    2019

  33. [41]

    shape-only

    This probability depends on the ratio of two experimental parameters: L, the distance from the neutrino creation to the observation point in meters (sometimes called the baseline), and E, the en- ergy of the neutrino in MeV. The fit extracts two model parameters: |Ue4|, the am...

  34. [42]

    Diaz A, Arg¨ uelles C A, Collin G H, Conrad J M and Shaevitz M H 2020 Phys. Rept. 884 1–59 ( Preprint 1906.00045)

    2020 arXiv

  35. [43]

    Seabold S and Perktold J 2010 statsmodels: Economet- ric and statistical modeling with python 9th Python in Science Conference

    2010

  36. [44]

    Stimper V, Liu D, Campbell A, Berenz V, Ryll L, Sch¨ olkopf B and Hern´ andez-Lobato J M 2023Journal of Open Source Software8 5361 URL https://doi.org/ 10.21105/joss.05361

    doi

  37. [45]

    Ko Y J, Kim B R, Kim J Y, Han B Y, Jang C H, Jeon E J, Joo K K, Kim H J, Kim H S, Kim Y D, Lee J, Lee J Y, Lee M H, Oh Y M, Park H K, Park H S, Park K S, Seo K M, Siyeon K and Sun G M (NEOS Collaboration) 2017 Phys. Rev. Lett. 118(12) 121802 URL https:// link.aps.org/doi/10.11...

    2017 doi

  38. [46]

    Lyons L 2014 Raster scan or 2-d approach? ( Preprint 1404.7395) URL https://arxiv.org/abs/1404.7395

    2014 arXiv

  39. [47]

    Samantha R Cook A G and Rubin D B 2006 Journal of Computational and Graphical Statistics 15 675–692 (Preprint https://doi.org/10.1198/106186006X136976) URL https://doi.org/10.1198/106186006X136976

    2006 doi

  40. [48]

    Almaz´ an H, Bernard L, Blanchet A, Bonhomme A, Buck C, del Amo Sanchez P, El Atmani I, Haser J, Kandzia F, Kox S, Labit L, Lamblin J, Letourneau A, Lhuillier D, Licciardi M, Lindner M, Materna T, Minotti A, Pes- sard H, R´ eal J S, Roca C, Rogly R, Salagnac T, Savu V, Schoppm...

    2020

  41. [49]

    Hermans J, Delaunoy A, Rozet F, Wehenkel A, Begy V and Louppe G 2022 Transactions on Machine Learning Research ISSN 2835-8856 URL https://openreview. net/forum?id=LHAbHkt6Aq 13 Appendix A: Network architectures Input x DRD(n1, p) DRD(n2, p) DRD(n3, p) DRD(n4, p) DRD(n, p) Dens...

    2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.