Pith. sign in

REVIEW 3 major objections 5 minor 28 references

Assessing Convergence Patterns Across Modern Nucleon-Nucleon Potentials

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The angular correlation length of chiral-EFT expansion coefficients falls as one over the relative momentum, and warping the angle axis by momentum restores the BUQEYE truncation-error model's statistical consistency.

desk verdict ℓ_θ ~ 1/p is likely right; the warping validation is confounded and the abstract oversells it, but the paper is honest, reproducible, and deserves refereeing. read the letter →

arxiv 2508.17558 v1 pith:VTAWNHBW submitted 2025-08-25 nucl-th hep-phphysics.data-an

P. J. Millican , R. J. Furnstahl , J. A. Melendez , D. R. Phillips This is my paper · ORCID
classification nucl-thhep-phphysics.data-an PACS 13.75.Cs21.30.-x
keywords chiraleffectivefieldtheorynucleon-nucleonpotentialsEFTtruncationerrorsGaussianprocessBUQEYEmodelnonstationarityinput-spacewarpingbreakdownscale
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

The paper asks whether a Bayesian model for correlated effective field theory truncation errors — the BUQEYE model, which treats the dimensionless coefficients of an order-by-order observable expansion as draws from one Gaussian process — survives contact with sixteen modern chiral nucleon-nucleon potentials from four regulator families. It does, but only after one modification: the correlation length in the scattering-angle direction is not constant but falls roughly as $1/p_{\rm rel}$, the inverse of the relative momentum, so a stationary Gaussian process is misspecified. Warping the angle coordinate by a single power of $p_{\rm rel}$ restores stationarity and makes the model's coverage and consistency diagnostics pass for the potentials that behave. The authors also show that soft regulators produce an even-odd staggering in coefficient sizes that no choice of the breakdown scale $\Lambda_b$ can repair, and that order-stable values for $\Lambda_b$ emerge only for the SMS potential at 450 or 500 MeV. Why this matters: if the modified model is right, truncation-error bars on chiral EFT predictions can be computed from lower orders alone, for the potentials where the convergence pattern is regular.

What carries the argument

The load-bearing object is the BUQEYE model itself: a zero-mean Gaussian process with a squared-exponential kernel, $\kappa(x,x') = \bar{c}^2 \exp[-(x_E-x'_E)^2/(2\ell_E^2) - (x_\theta-x'_\theta)^2/(2\ell_\theta^2)]$, in the input space $x_E = p_{\rm rel}$, $x_\theta = -\cos\theta$, whose draws model the coefficient functions $c_n(x)$. The mechanism that carries the argument is the warping transformation $w_\theta(x_E,x_\theta) = x_\theta\,(p_{\rm rel}/405\,\mathrm{MeV})^B$ with $B = 1$, which shrinks the angular input at low momentum so that a stationary length scale becomes an accurate description; it is exactly equivalent to a nonstationary kernel whose angular length scale depends on momentum, and it guides the downsampling of low-momentum training points, where the long correlation length makes each point carry less information.

What would settle it

Extract the angular length scale $\ell_\theta$ in fine momentum bins from order-by-order predictions of a chiral potential outside the fitted set — for example a $\Delta$-full interaction, or the same potentials extended beyond $p_{\rm rel} = 405$ MeV. If the fitted exponent in $\ell_\theta = a\,(p_{\rm rel}/450\,\mathrm{MeV})^{-b}$ departs significantly from $b = 1$, or if the warped-model diagnostics degrade once the low-momentum region is sampled without downsampling, the proposed warping is the wrong remedy rather than a validated one.

Watch

Extended reading notes

Core claim

On the paper's own terms: the assumption that the dimensionless coefficients $c_n(x)$ extracted from order-by-order observable predictions are draws from a single Gaussian process fails for modern chiral NN potentials as currently implemented, and the specific failure is identified. The GP's angular length scale obeys $\ell_\theta = a\,(p_{\rm rel}/450\,\mathrm{MeV})^{-b}$ with $b \approx 1$ for all six observables tested, a semi-classical reflection of the number of partial waves accessible at momentum $p_{\rm rel}$; the variance $\bar{c}^2$ and the momentum-direction length scale $\ell_E$, by contrast, show no systematic trend and can be treated as stationary. Making $\ell_\theta$ momentum-dependent restores the model: the authors implement this as a warping of the angular input, $w_\theta = x_\theta\,(p_{\rm rel}/405\,\mathrm{MeV})^B$ with $B = 1$, show it is equivalent to a nonstationary kernel with a momentum-dependent length scale, and find that after warping with low-momentum downsampling the re-extracted exponent is consistent with $b = 0$ and all three diagnostics (credible-interval coverage, Mahalanobis distance, pivoted Cholesky decomposition) improve substantially. The modified model then yields statistically consistent truncation errors for the regular potentials, provided $\Lambda_b$ is set near 600–750 MeV; cross-order consistency of the $\Lambda_b$ posterior is the exception, holding only for SMS 450 and 500 MeV.

Load-bearing premise

Everything rests on one power-law assumption: that the angular correlation length follows $\ell_\theta = a\,(p_{\rm rel}/450\,\mathrm{MeV})^{-b}$ with the exponent fixed at $b = 1$ for all observables and potentials; if the true momentum dependence is not this single-exponent form, the warped Gaussian process is misspecified and the improved diagnostics do not validate the model.

Editorial extensions

If this is right

  • For potentials with regular convergence (SMS 450–550 MeV, SCS 0.9–1.0 fm, EMN 500 MeV), order-by-order error bars computed with the warped model cover the next order at the claimed rate, so those truncation errors can be propagated into nuclear-structure predictions.
  • Soft potentials (SMS 400 MeV, SCS 1.1 and 1.2 fm) are diagnosably irregular: the even-odd coefficient staggering persists for any choice of $\Lambda_b$ or $m_{\rm eff}$, so they should be excluded from this kind of uncertainty quantification.
  • Reported values of the chiral breakdown scale depend on the truncation order used to extract them; only SMS 450 and 500 MeV give $\Lambda_b$ posteriors that are stable across orders, so single-order $\Lambda_b$ claims for other potentials carry unquantified order dependence.
  • Treating both $\Lambda_b$ and $m_{\rm eff}$ as free parameters overfits: the joint posteriors do not narrow as orders are added, which points to the parametrization of $Q$ rather than the GP kernel as the model's limiting assumption.
  • Because warping and a nonstationary kernel give identical results under an equivalent train-test split, the practical lesson transfers: the split, and specifically low-momentum downsampling, matters more than which implementation is chosen.

Reading between the lines

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

  • The $1/p_{\rm rel}$ law for the angular correlation length is a partial-wave counting argument, so the same warping with $B = 1$ could plausibly serve as a default prior for other scattering observables — pp with Coulomb, $\Delta$-full potentials, and 3N observables — without re-deriving the exponent, an extension the paper leaves to future work.
  • A testable stress point: the semi-classical argument predicts the exponent should fail in regions where a single partial wave dominates, so measuring $\ell_\theta$ near thresholds or in kinematically constrained configurations would probe whether $B = 1$ is a universal feature or a fitted coincidence.
  • The even-odd staggering pattern gives a cheap quantitative diagnostic of regulator artifacts: alternating rms coefficient sizes beyond roughly a factor of two between consecutive orders signal that pion exchange has been pushed into even-order contact terms, connecting directly to pionless-EFT expectations for these potentials.
  • The correlation between the warping exponent $B$ and the extracted $\Lambda_b$ is not explored in the paper; fitting both together would reveal whether the quoted 600–750 MeV range carries an additional systematic error from fixing $B = 1$.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper assesses the BUQEYE correlated-truncation-error model across multiple modern chiral EFT nucleon-nucleon potentials. Using six scattering observables as functions of relative momentum and scattering angle, the authors first identify an even-odd staggering in coefficient sizes for soft potentials and exclude the softest cases from further analysis. They then report evidence that the Gaussian-process angular length scale ℓ_θ is approximately inversely proportional to p_rel, fit this as a power law, and introduce a warping transformation with fixed exponent B=1 to restore stationarity. After warping and low-momentum downsampling they report improved diagnostics (DCI, D2_MD, DPC) for SMS 500 MeV and EMN 500 MeV, and then extract posterior distributions for the breakdown scale Λ_b at fixed meff and with meff variable. They find statistically consistent Λ_b posteriors across orders only for SMS 450 and 500 MeV, and they interpret the variable-meff results as an overfitting problem. The paper includes a publicly available Jupyter notebook for reproduction.

Significance. If the central claim is established, the paper would provide an important practical modification of the BUQEYE model for NN scattering, with a physically motivated treatment of nonstationarity in the angular correlation structure and a systematic survey of regulator schemes and scales. The paper is commendably honest about the limitations of cross-order Λ_b consistency, and it ships reproducible code. However, the validation of the proposed warping is currently under-supported: the diagnostics comparison in Sec. IVD changes two things at once (warping and downsampling), and the b=0 result after warping is a consistency check on data used to fix B=1 rather than an independent test. The central claim therefore needs a controlled validation before the abstract's 'validating' conclusion can be accepted.

major comments (3)
  1. [Sec. IVD, Figs. 6-9] The comparison that is claimed to validate the warping is not controlled. Figure 6 uses 30 training and 69 testing points with no warping and no downsampling, while Figure 7 uses 14 training and 33 testing points with warping and downsampling; the same is true of Figs. 8 and 9 for EMN 500 MeV. The improvement in DCI, D2_MD, and DPC could therefore be due to the removal of low-momentum test points or to the smaller training set rather than to the restoration of stationarity in the warped space. To support the abstract's claim that the diagnostics 'validate' the modification, the authors should provide a controlled comparison that isolates warping from downsampling, for example by applying the warp without changing the training/testing points, by applying the downsampling without warping, or by testing on a fixed held-out set. A quantitative significance statement for the diagnostics would also strengthen the claim.
  2. [Sec. IIIB1, Eq. (14); Sec. IVA, Eq. (16); Sec. IVD] The validation logic is partially circular. The power law ℓ_θ = a (p_rel/450 MeV)^{-b} is extracted from the data, the warping in Eq. (16) is chosen as its inverse with B=1, and the re-fit in Sec. IVD then reports that b is consistent with 0 after warping. This confirms that the warp removes the fitted trend, but it is not independent evidence that the 1/p_rel law is the correct description of the nonstationarity. The authors should test the warping on held-out data or observables not used to fit Eq. (14), or compare the warped-space GP against alternative nonstationarity models (e.g., other exponents or a nonlinear ℓ_θ(p_rel)) using predictive diagnostics. Without such a test, the claim that the modification is 'validated' goes beyond what the current evidence supports.
  3. [Sec. VA and Sec. VB, Table II] All Λ_b and meff results in Sec. V are obtained with the fixed warp exponent B=1 and the accompanying downsampling. If the assumed power-law form or the fixed exponent is misspecified for one of the potentials, the subsequent Λ_b posteriors inherit that misspecification, so the finding that only SMS 450 and 500 MeV show cross-order consistency is conditional on the B=1 assumption. The paper would be stronger if it reported a sensitivity analysis of the Λ_b posteriors to B (or to the choice of downsampling scheme), at least for one potential, so that the robustness of the central 'statistically consistent only for SMS' conclusion can be assessed.
minor comments (5)
  1. [Table I caption] The caption contains typographical errors: 'Note thesimilarvaluesacrossforSMS500MeVandandacrossSCS 0.9 fm' should be cleaned up, and the sentence about even-order and odd-order values is grammatically incomplete.
  2. [Sec. VA, text after Fig. 13] The phrase 'There are are notable jumps between orders' contains a duplicated word, and the sentence beginning 'It may also be the case that how potentials are fitted matters here' would benefit from a smoother transition.
  3. [Sec. IVD, Fig. 10 discussion] The phrase 'the sy-setmatic separation of the curves' contains a typo, and the qualitative statement about the DCI trend could be complemented by a numerical measure, such as the coverage deficit or the slope of the DCI curves.
  4. [Sec. IIIB1] The presentation states that 'the values of b are not far from 1' and then fixes B=1; it would be helpful to report the fitted b values and their uncertainties numerically, at least in a table, rather than only showing them graphically in Fig. 3.
  5. [Sec. IVC] The discussion of the equivalence between warping and the nonstationary kernel is clear, but it should be explicitly noted that the practical claim of equivalence in the paper applies only to the diagnostics computed with the specific downsampled training and testing sets, since the paper does not present a head-to-head comparison with identical point sets for both methods.

Circularity Check

1 steps flagged · score 4.0 of 10

The post-warp b≈0 check is a mechanical consequence of choosing the warp as the inverse of the fitted ℓθ power law; the rest of the model comparison is largely self-contained.

  1. fitted input called prediction [Sec. IIIB1 (Eq. 14), Sec. IVA (Eq. 16), Sec. IVD]
    "After warping and accompanying downsampling with B = 1, we re-fitted the momentum-dependence of the angular length scale using Eq. (14), thus effectively turning the power of p with which the space is warped into one of the set of global parameters, βg. Now, when b is extracted across all observables for a given potential, the resulting posterior probability distribution function (pdf) is consistent with b = 0."

    Eq. (16) sets wθ = xθ (prel/405 MeV)^B, i.e., it multiplies the angular coordinate by p^B. If the raw ℓθ is ℓθ = a (prel/450 MeV)^(-b), then in the warped coordinate the effective angular length scale is ℓθ (prel/405 MeV)^B, whose exponent is -(b-B) (up to the 405/450 normalization). The paper fit b ≈ 1 from the same SMS 500 data and then fixed B = 1 in Eqs. (14) and (16); re-fitting Eq. (14) after the warp therefore algebraically forces b ≈ 0. The b = 0 result confirms only that the warp is the inverse of the fitted trend, not an independent datum for ℓθ ∼ 1/prel.

full rationale

The paper's BUQEYE analysis is mostly self-contained: coefficient curves are extracted from published NN potential calculations, GP hyperparameters are re-estimated, and the diagnostics are computed from held-out test points and next-order results. The nonstationarity finding is an empirical fit supported by a semiclassical partial-wave argument, not a renamed consensus. Self-citations to Refs. [2,4,13] provide methodology and prior choices that are re-tested here, so they are not load-bearing circularity. The one genuine circular step is the Sec. IVD re-fit of b after warping: since the warp in Eq. (16) is the inverse of the power law fitted in Eq. (14) with B set to the fitted value 1, obtaining b≈0 is a consistency restatement. A further non-circular concern is that Figs. 6–7 change warping and downsampling simultaneously (30/69 vs 14/33 train/test points), so the diagnostic improvement is not cleanly attributable to the ℓθ(p) remediation; this is an experimental-design weakness rather than definitional circularity. The paper itself limits its strongest claim by reporting that cross-order Λb consistency is established only for SMS 450 and 500 MeV, which tempers the score. Overall: one fitted-input-as-validation step, with the central model comparison retaining independent content.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The analysis imports the BUQEYE GP framework (zero-mean SERBF kernel, Q_sum, y_ref, priors) from the authors' own Refs. [2,4]. On top of that it fits Λ_b, the GP length scales, and the power-law parameters a and b; b is then fixed to B=1 for the warping. These fitted quantities carry most of the load. No new physical entities are introduced.

free parameters (7)
  • Λ_b (EFT breakdown scale) = 430-810 MeV depending on potential and order (Table II)
    Inferred from order-by-order coefficient data; the central fitted parameter.
  • m_eff (soft scale) = Fixed at 138 or 200 MeV; when treated as random variable, 44-193 MeV (Table II)
    Held fixed in most analyses, but Sec VB treats it as a free parameter and finds posteriors inconsistent across orders.
  • ℓ_θ (GP angular length scale in warped space) = Observable-specific random variables, not tabulated explicitly
    Fitted per observable/potential as part of kernel hyperparameters.
  • ℓ_E (GP momentum length scale) = Observed in range 40-80 MeV (Fig. 21)
    Fitted per observable, treated as stationary.
  • Power-law prefactor a in Eq. (14) = Per-observable fits shown in Fig. 3 red lines
    Fit observable by observable to describe ℓ_θ(p_rel).
  • Power-law exponent b, fixed to B = 1 = b ≈ 1 from fits; B = 1 adopted
    Initial fits give b near 1; the warp uses B=1, so the exponent becomes a fitted global constant.
  • Training and testing grid sizes after downsampling = 14 training, 33 testing points
    Chosen by hand in Sec IVC; low-momentum downsampling is motivated by longer ℓ_θ there.
assumptions (6)
  • domain assumption Coefficient functions c_n(x) are i.i.d. draws from a single zero-mean Gaussian process with a squared-exponential kernel (Eqs. 3-6).
    This is the BUQEYE hypothesis under test, imported from Refs. [2,4]; all inferences in the paper assume it.
  • domain assumption The expansion parameter is Q_sum(p_rel, m_eff) = (p_rel + m_eff)/(Λ_b + m_eff) (Eq. 2).
    Adopted from the authors' earlier work Ref. [2] without re-validation across the potential families; it controls the extraction of coefficients and all Λ_b posteriors.
  • domain assumption Reference scales y_ref are chosen as in Ref. [2] Table I.
    Affects the extracted coefficients through Eq. (1); the paper does not test alternative y_ref choices.
  • ad hoc to paper The angular length scale obeys ℓ_θ = a (p_rel/450 MeV)^{-b}, and b can be fixed to B = 1 via a semi-classical argument.
    The power-law form is fit to the same data used in the validation, and the exponent B=1 is fixed from that fit; the semi-classical argument is invoked post hoc.
  • domain assumption The GP length scale in momentum ℓ_E and the variance ¯c² are stationary over the input space.
    The authors check stationarity of ℓ_E and ¯c² and conclude they can be treated as stationary; this is an empirical conclusion, not an a priori truth.
  • standard math Bayesian priors: Heaviside prior for length scales and scaled inverse-chi-squared prior for ¯c² (Eq. 12).
    Standard weakly-informative priors from Ref. [4]; not fit to the data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Assessing Convergence Patterns Across Modern Nucleon-Nucleon Potentials." pith.science (2026). https://pith.science/paper/VTAWNHBW

@misc{pith2026250817558,
  author       = {Pith},
  title        = {Pith review of: Assessing Convergence Patterns Across Modern Nucleon-Nucleon Potentials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VTAWNHBW}},
  note         = {Machine review of arXiv:2508.17558}
}
abstract

The BUQEYE model for correlated effective field theory (EFT) truncation errors assumes a regular pattern of dimensionless coefficients extracted from order-by-order observable calculations. This enables results from lower orders to inform statistical predictions for error estimates of omitted higher orders. We test the model for multiple chiral EFT ($\chi$EFT) nucleon-nucleon (NN) potentials using a suite of six common NN scattering observables represented as functions of relative momentum and scattering angle. First, we flag irregularity in the convergence patterns of potentials with so-called "soft" regulator scales, namely that the sizes of the coefficients in the observables' expansions are mismatched between the even and odd orders. Second, we test the BUQEYE model's assumption of Gaussian process (GP) stationarity against the data and find that the GP's correlation structure as encoded in its length scale is nonstationary: The GP length scale in the angular dimension $\ell_{\theta}$ is approximately inversely proportional to the relative momentum of the scattering process. After we remediate this issue by allowing $\ell_{\theta}$ to vary with momentum, diagnostics show significant improvement, validating the application of the BUQEYE model after this modification. Third, we find that good performance of the BUQEYE model relies on properly choosing the $\chi$EFT breakdown scale $\Lambda_{b}$. With $m_{\text{eff}}$ held fixed at the physical pion mass we find $\Lambda_{b}=600$--$750$ MeV for various N$^{3}$LO interactions. Statistically consistent distributions for $\Lambda_{b}$ across orders are only found for the SMS potential at a regulator scale of 450 or 500 MeV. All our results can be reproduced using a publicly available Jupyter notebook, which can be straightforwardly modified to analyze other $\chi$EFT NN potentials.

Figures

Figures reproduced from arXiv: 2508.17558 by the authors.

Figure 1
Figure 1. FIG. 1. Order-by-order coefficient functions for [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Order-by-order coefficients for [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Plots of [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (19 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Plots of [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. Unwarped (left) and warped (right) coefficients for [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. D [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. D [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8. D [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9. D [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10. Diagnostics for the differential cross section [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: FIG. 11. Diagnostics for the differential cross section [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: FIG. 12. Diagnostics for the differential cross section [PITH_FULL_IMAGE:figures/full_fig_p015_12.png]
Figure 13
Figure 13. Figure 13: FIG. 13 [PITH_FULL_IMAGE:figures/full_fig_p017_13.png]
Figure 14
Figure 14. Figure 14: FIG. 14 [PITH_FULL_IMAGE:figures/full_fig_p018_14.png]
Figure 15
Figure 15. Figure 15: FIG. 15. Diagnostics for the differential cross section [PITH_FULL_IMAGE:figures/full_fig_p018_15.png]
Figure 16
Figure 16. Figure 16: FIG. 16. Plot of the joint [PITH_FULL_IMAGE:figures/full_fig_p019_16.png]
Figure 17
Figure 17. Figure 17: FIG. 17. Order-by-order coefficients for [PITH_FULL_IMAGE:figures/full_fig_p022_17.png]
Figure 18
Figure 18. Figure 18: FIG. 18. Order-by-order coefficients for [PITH_FULL_IMAGE:figures/full_fig_p023_18.png]
Figure 19
Figure 19. Figure 19: FIG. 19. Order-by-order coefficients for [PITH_FULL_IMAGE:figures/full_fig_p024_19.png]
Figure 20
Figure 20. Figure 20: FIG. 20. Order-by-order coefficients for [PITH_FULL_IMAGE:figures/full_fig_p025_20.png]
Figure 21
Figure 21. Figure 21: FIG. 21. Plots of [PITH_FULL_IMAGE:figures/full_fig_p026_21.png]
Figure 22
Figure 22. Figure 22: FIG. 22. Plots of [PITH_FULL_IMAGE:figures/full_fig_p027_22.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 11 canonical work pages

  1. [1]

    Figure 2 shows the coefficients extracted from both hard and soft potentials’ calculations ofdσ/dΩ at fixed values ofxE = prel plot- ted againstxθ =− cos(θ)

    Length scale We begin with the length scale. Figure 2 shows the coefficients extracted from both hard and soft potentials’ calculations ofdσ/dΩ at fixed values ofxE = prel plot- ted againstxθ =− cos(θ). Low-momentum coefficients are shown in purple and high-momentum in bright yel- low. The flat, almost linear shape of the low-momentum curves gives way to ...

    work page

  2. [2]

    low, high, low

    Variance Now we will move on to consideration of the variance ¯c2. Section IV of Ref. [2] pointed out that ¯c2 shows some signs of nonstationarity, and that the spread of the variance is significantly lower when training points below roughly prel = 125 MeVare omitted. Here, we proceed as for the GP length scales in the previous section. We examine results...

    work page

  3. [3]

    deformed,

    Conclusion We showed in Sec. IIIB1 that the RBF kernel length scale ℓθ has a 1/prel dependence onxE = prel, while in Sec. IIIB2 we found no systematic trend of variance non- stationarity. We can implement a two-dimensional GP for the coefficientscn that includes this behavior of the GP length scale. The extraction of probability distribu- tions for GP hyp...

    work page

  4. [4]

    Machleidt and F

    R. Machleidt and F. Sammarruca, Prog. Part. Nucl. Phys. 137, 104117 (2024), arXiv:2402.14032 [nucl-th]

    arXiv 2024

  5. [5]

    P. J. Millican, R. J. Furnstahl, J. A. Melendez, D. R. Phillips, and M. T. Pratola, Phys. Rev. C110, 044002 (2024), arXiv:2402.13165 [nucl-th]

    arXiv 2024

  6. [6]

    R. J. Furnstahl, N. Klco, D. R. Phillips, and S. Wesolowski, Phys. Rev. C 92, 024005 (2015), arXiv:1506.01343

    arXiv 2015

  7. [7]

    J. A. Melendez, R. J. Furnstahl, D. R. Phillips, M. T. Pratola, and S. Wesolowski, Phys. Rev. C100, 044001 (2019), arXiv:1904.10581

    arXiv 2019

  8. [8]

    Reinert, H

    P. Reinert, H. Krebs, and E. Epelbaum, Eur. Phys. J. A 54, 86 (2018), arXiv:1711.08821

    arXiv 2018

Show all 28 references
  1. [9]

    Epelbaum, H

    E. Epelbaum, H. Krebs, and U. G. Meißner, Eur. Phys. J. A 51, 53 (2015), arXiv:1412.0142

    2015 arXiv

  2. [10]

    D. R. Entem, R. Machleidt, and Y. Nosyk, Phys. Rev. C 96, 024004 (2017), arXiv:1703.05454

    2017 arXiv

  3. [11]

    Gezerlis, I

    A. Gezerlis, I. Tews, E. Epelbaum, M. Freunek, S. Gan- dolfi, K. Hebeler, A. Nogga, and A. Schwenk, Phys. Rev. C 90, 054323 (2014), arXiv:1406.0454

    2014 arXiv

  4. [12]

    B. D. Carlsson, A. Ekström, C. Forssén, D. F. Ström- berg, G. R. Jansen, O. Lilja, M. Lindby, B. A. Matts- son, and K. A. Wendt, Phys. Rev. X6, 011019 (2016), arXiv:1506.02466

    2016 arXiv

  5. [13]

    Ekström, G

    A. Ekström, G. R. Jansen, K. A. Wendt, G. Hagen, T. Papenbrock, B. D. Carlsson, C. Forssén, M. Hjorth- Jensen, P. Navrátil, and W. Nazarewicz, Phys. Rev. C 91, 051301(R) (2015), arXiv:1502.04682

    2015 arXiv

  6. [14]

    Piarulli, L

    M. Piarulli, L. Girlanda, R. Schiavilla, R. Navarro Pérez, J. E. Amaro, and E. Ruiz Arriola, Phys. Rev. C 91, 024003 (2015), arXiv:1412.6446

    2015 arXiv

  7. [15]

    Ekström, G

    A. Ekström, G. Hagen, T. D. Morris, T. Papenbrock, and P. D. Schwartz, Phys. Rev. C 97, 024332 (2018), arXiv:1707.09028

    2018 arXiv

  8. [16]

    J. A. Melendez, S. Wesolowski, and R. J. Furnstahl, Phys. Rev. C96, 024003 (2017), arXiv:1704.03308

    2017 arXiv

  9. [17]

    J. M. Bub, M. Piarulli, R. J. Furnstahl, S. Pastore, and D. R. Phillips, Phys. Rev. C 111, 034005 (2025), arXiv:2408.02480 [nucl-th]

    2025 arXiv

  10. [18]

    J.-W. Chen, G. Rupak, and M. J. Savage, Nucl. Phys. A 653, 386 (1999), arXiv:nucl-th/9902056

    1999 arXiv

  11. [19]

    Weinberg, Nucl

    S. Weinberg, Nucl. Phys. B363, 3 (1991)

    1991

  12. [20]

    J. A. Melendez, R. J. Furnstahl, H. W. Grießhammer, J. A. McGovern, D. R. Phillips, and M. T. Pratola, Eur. Phys. J. A57, 81 (2021), arXiv:2004.11307 [nucl-th]

    2021 arXiv

  13. [21]

    Zhang and D.-Y

    Y. Zhang and D.-Y. Yeung, in2010 IEEE Computer So- ciety Conference on Computer Vision and Pattern Recog- nition (2010) pp. 2622–2629. 20

    2010

  14. [22]

    Lally and B

    N. Lally and B. Hartman, Insurance: Mathematics and Economics 82, 124 (2018)

    2018

  15. [23]

    V. D. Agou, A. Pavlides, and D. T. Hristopulos, Entropy 24 (2022), 10.3390/e24030321

    2022 doi

  16. [24]

    P. Kou, F. Gao, and X. Guan, Applied Energy108, 410 (2013)

    2013

  17. [25]

    Snelson, Z

    E. Snelson, Z. Ghahramani, and C. Rasmussen, in Advances in Neural Information Processing Systems , Vol. 16, edited by S. Thrun, L. Saul, and B. Schölkopf (MIT Press, 2003)

    2003

  18. [26]

    Pedregosa, G

    F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R.Weiss, V.Dubourg, J.Vanderplas, A.Passos, D.Cour- napeau, M. Brucher, M. Perrot, and E. Duchesnay, Jour- nal of Machine Learning Research12, 2825 (2011)

    2011

  19. [27]

    Epelbaum, PoSCD2018, 006 (2019)

    E. Epelbaum, PoSCD2018, 006 (2019)

    2019

  20. [28]

    Bayesian Analysis of Nuclear Dynamics (BAND) Frame- work project (2020) https://bandframework.github. io/. 21 Appendix A: Supplemental Material The content of this Supplemental Material sheds ad- ditional light on the coefficients not displayed in Figs. 1 and 2 (see Figs. 17–2...

    2020

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.