Pith. sign in

REVIEW 3 major objections 4 minor 68 references

Physics-structured cooperative neural network for baseline-free nuclear mass modeling

T0 review · 3 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read A modular neural network predicts nuclear binding energies from proton and neutron counts alone, reaching 0.269 MeV accuracy without a theoretical baseline.

desk verdict A clean architecture-prior result with credible accuracy numbers; the physical-decomposition story is plausible but needs identifiability checks. read the letter →

arxiv 2603.09747 v2 pith:3QTCXKRC submitted 2026-03-10 nucl-th

Peiwen Zai , Wei Cheng , Feng-Shou Zhang This is my paper · ORCID
classification nucl-th
keywords nuclearmassesbindingenergymachinelearningarchitecturepriorsmacroscopic-microscopicdecompositionshelleffectspairingextrapolation
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

Cooperative Neural Network (CoNN) aims to establish that the macroscopic–microscopic picture of nuclear masses—a smooth bulk term plus shell, collective, and pairing corrections—can be built into a neural network's architecture instead of being supplied as hand-crafted features or a theoretical baseline. Taking only proton and neutron numbers as input, the network sums four constrained branches and, on the AME2020 dataset, achieves a reported RMSD of 0.269 MeV across 3558 nuclei, versus 0.836 MeV for a parameter-matched unstructured MLP. If the claim holds, architecture-level priors are a viable route to baseline-free mass prediction, and the learned branches become diagnostics of nuclear structure. The authors further report that shell embeddings develop extrema at magic numbers and the pairing branch produces the expected odd–even staggering without explicit supervision.

What carries the argument

The load-bearing mechanism is the four-branch additive architecture and its cooperative training schedule. The macroscopic branch is a bottlenecked fully connected network that can only represent smooth trends; the shell branch is a sum of two discrete learnable embeddings eZ[Z]+eN[N]; the correlation branch is a 50×60 trainable grid with bilinear interpolation; and the pairing branch is a small MLP whose fixed, non-learnable modulo inputs (Z mod 2, N mod 2) isolate parity. Alternating updates—macroscopic branch first, then microscopic branches on the residual, at a 10:1 learning-rate ratio—push each module toward a distinct function class, which is what lets the total sum separate into phys

What would settle it

Fit the same CoNN to a synthetic dataset generated from a smooth liquid-drop formula with no shell effects. If the decomposition is genuine, the discrete shell embeddings should stay flat; if they develop artificial extrema, the 'emergent' magic-number structure is a by-product of optimization dynamics. A second test: on real data, transfer the macroscopic branch's role to the correlation grid (or vice versa) and compare total RMSD—if it stays about the same while the branch contents change drastically, the additive decomposition is not identifiable.

Watch

Extended reading notes

Core claim

The central claim is that the additive decomposition Bpred = EMacro + EShell + ECor + EPair, enforced by four structurally constrained modules and an alternating training protocol, lets the network capture the entire binding-energy surface from raw nucleon numbers alone. On AME2020 the model reaches 0.269 MeV RMSD on all 3558 nuclei, 0.419 MeV on a held-out interpolation subset, and 0.728 MeV on 122 nuclei newly measured since AME2016. The authors interpret the accuracy gap against a parameter-matched plain MLP (0.836 MeV) as evidence that inductive bias, not capacity, does the work. They also report that the learned components reproduce recognizable physics—shell kinks near magic numbers, d

Load-bearing premise

The training recipe only guarantees the four branches sum to the right total; it does not guarantee that each branch absorbs only its intended physics, since moving smooth trends or constants between branches leaves the sum unchanged, and the paper reports no check that the separation is unique.

Editorial extensions

If this is right

  • Baseline-free mass prediction becomes competitive with feature-engineered models: CoNN's 0.269 MeV RMSD with two raw inputs is close to the reported 0.260 MeV of an eleven-feature model on the same task.
  • The same structural priors can be transferred to other nuclear observables such as charge radii and beta-decay properties, shifting the design question from feature selection to architecture design.
  • The extrapolation result on 122 newly measured nuclei (0.728 MeV) indicates that architecture-level priors alone give better boundary extrapolation than two standard macroscopic–microscopic models (2.444 and 1.295 MeV on the same set), though not as good as feature-augmented networks.
  • Derived quantities such as separation energies and decay Q-values are reproduced at 0.29–0.36 MeV RMSD, meaning the learned mass surface is smooth enough for finite-difference predictions to remain meaningful.

Reading between the lines

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

  • The identifiability of the branch decomposition is a testable assumption the paper does not address; one could quantify it by reparameterizing the sum (for example, learning a smooth function and moving it between EMacro and EShell) and measuring how much each module's output shifts—if the modules trade functions without changing total error, the physical interpretation of individual branches shou
  • The hand-set 10:1 learning-rate asymmetry is a prime suspect for causing the emergent shell peaks; sweeping this ratio and checking whether the embeddings still develop extrema at magic numbers would reveal whether the physics-like outputs are a robust attractor or an artifact of the training schedule.
  • The hard Z ≤ 120, N ≤ 180 boundary of the discrete embeddings is a clear limitation the authors acknowledge; replacing them with continuous but sharply sloped functions is a natural next step that could be validated on superheavy and very neutron-rich nuclei, where the current model has no support.
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 / 4 minor

Summary. The paper proposes a modular neural network, CoNN, for predicting nuclear binding energies directly from (Z,N) without a theoretical mass baseline or hand-crafted physics features. The architecture sums four constrained branches: a smooth macroscopic MLP, discrete shell embeddings, a two-dimensional regional correlation grid, and a parity-aware pairing network, trained with a two-phase alternating protocol. On AME2020 the model is reported to achieve RMSDs of 0.269 MeV on all 3558 nuclei, 0.419 MeV on a held-out interpolation subset, and 0.728 MeV on 122 nuclei newly measured since AME2016, beating a parameter-matched MLP (0.836 MeV) and approaching feature-engineered direct models. The paper further claims that the learned branches recover recognizable physical patterns—shell extrema at magic numbers, regional deformation-like structures, and odd-even staggering—and that the architecture itself constitutes a generalizable physical prior.

Significance. If the main accuracy result holds, the paper makes a valuable contribution to ML-based nuclear mass modeling: it demonstrates a clean recipe for replacing engineered input features with architectural inductive biases, and it provides a genuine temporal extrapolation test on AME2020-new nuclei. The parameter-matched MLP control and the explicit interpolation/extrapolation split are appropriate and strengthen the case that the improvement is due to structure rather than parameter count. The derived-quantity tests (separation energies and Q-values) are a useful addition. However, the paper's second central claim—that the learned branch decomposition is physically meaningful and diagnostic—rests on an unproven identifiability assumption. Because the sum in Eq. (1) is invariant under moving functions between branches, the visual 'emergent' patterns in Figs. 4–6 may be artifacts of initialization or optimization, not stable physical separations. This does not invalidate the numerical accuracy claims, but it substantially weakens the interpretability message, which is presented as a key advantage of the method. The paper would need additional experiments that establish branch stability an

major comments (3)
  1. [Sec. II C, Eq. (1), Sec. III B] The decomposition B = EMacro + EShell + ECor + EPair is not identifiable from the additive structure alone. The modules have overlapping representational capacity: a smooth function of Z and N can be absorbed by the macroscopic MLP or, if separable, by the shell embeddings; a global constant can shift among any branch; and the bilinear grid can represent smooth trends that the macroscopic network can also fit. The alternating training protocol with a 10:1 learning-rate ratio may bias the solution, but it provides no formal or empirical guarantee of separation. Section III B then interprets the learned branches as 'microscopic structures,' and Figures 4–6 are presented as confirmation. This is circular unless the authors show that the decomposition is stable and that the observed patterns are not destroyed by invariant reassignments. I ask for (a) per-seed branch outputs (e.g., the range
  2. [Sec. III A/B, Fig. 3] The ablation analysis is incomplete. The only ablation reported is removal of the pairing module (Fig. 3d); there is no ablation removing the shell embeddings, the correlation grid, or combinations thereof. The claim that each module 'captures a specific type of physical contribution' (Sec. II B) is therefore not tested. In particular, the reader cannot tell whether the grid is responsible for the doubly-magic patches, or whether the shell embeddings already absorb part of those correlations, because the decomposition is not identifiable in the first place. An ablation with each module removed separately, plus a branch-importance metric (e.g., change in validation RMSD), is necessary to support the modular attribution claims. Without such experiments, statements like 'the grid autonomously supplies the additional non-linear binding' (Sec. III B) are not load-bearing.
  3. [Abstract, Sec. III A, Table I] The headline number 'RMSD of 0.269 MeV on all 3558 nuclei' mixes training and validation data; the true held-out interpolation number is 0.419 MeV (688 validation nuclei). While the 0.269 MeV figure may be useful as an overall fit-quality statistic, it is not an out-of-sample accuracy metric and should not be the basis for comparing methods. Table I also mixes evaluation sets and editions: ANN2 is evaluated on 3556 AME2016 nuclei, KAN-2 on 3456 AME2020 nuclei, and CoNN on 3558 AME2020 nuclei, while BML's extrapolation entry is marked as estimated. The central 'most accurate baseline-free' claim should be rephrased to use the validation and extrapolation RMSDs (0.419 MeV and 0.728 MeV) or the comparison should be made on a unified protocol with a fixed data split. This is a presentation issue, but it directly affects how the paper's main claim is read.
minor comments (4)
  1. [Sec. II B] The macroscopic branch is described as an encoder-decoder with three hidden layers of width 128 and a 16-dimensional bottleneck, but the activation functions for the decoder and the final output layer are not specified. Please state the full architecture, including output activation and whether the target standardization is applied to the sum or to each branch.
  2. [Sec. III B, Fig. 4] The text says the embeddings develop 'shell-kink signatures' at magic numbers, but Fig. 4 shows extrema in bZ and bN, not kinks in a mass surface. The terminology should be adjusted to match what is plotted, or the figure should plot the derivative (e.g., separation-energy-like differences) to support the shell-kink claim.
  3. [Sec. III A, Fig. 3] The panel (b) 'CoNN macroscopic branch only' is stated to have RMSD = 2.114 MeV. Please clarify whether this is the warmup-trained branch alone, before cooperative training, or the macroscopic branch extracted from the final trained model. The distinction matters because the final macroscopic branch may have absorbed different content than the warmup-only branch.
  4. [General] The paper contains a number of typographical artifacts (e.g., 'efficently' and 'difficulty'). A careful proofreading pass is needed. No code or data-availability statement is provided; since all data are public, please consider adding a statement about code availability for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the headline accuracy figures are externally benchmarked against a temporal holdout and a parameter-matched MLP, and the physical-interpretation caveat is a robustness/identifiability concern, not a circular derivation.

full rationale

The paper's central numerical claims do not reduce to their inputs. CoNN is trained on the AME2016-overlap nuclei while the 122 AME2020-new nuclei form a temporal holdout used neither in training nor in validation-based model selection, and the parameter-matched plain MLP is trained under identical conditions as a control (Sec. II A, Sec. III A, Table I). Thus the reported 0.269/0.419/0.728 MeV figures are externally anchored predictions, not re-statements of fitted quantities. The physical-interpretation section (Sec. III B, Figs. 4–6) is the only candidate for concern. But the paper does not define EMacro, EShell, ECor, or EPair in terms of the shell/pairing patterns it later displays; Eq. (1) is an additive architectural ansatz, and the learned components are fitted functions inspected post hoc. The absence of an identifiability check (the sum is invariant under moving smooth trends between branches) means the 'physically meaningful decomposition' claim is under-supported, but that is a missing-validation issue, not a derivation-by-construction. No load-bearing self-citation is present: references to previous ML mass models are external comparisons, and no uniqueness theorem is invoked. The paper also states limitations explicitly (uncalibrated ensemble spread, finite grid range, larger extrapolation error) rather than hiding them. Under the circularity rules requiring a specific reduction (e.g., Eq. X = Eq. Y by construction, or a fitted parameter renamed as a prediction), none is exhibited. The central accuracy result would stand even if the branch decomposition were non-unique. Score 0.

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

The central claim rests on a standard nuclear-structure decomposition, accurate AME labels, and several hand-chosen architectural and optimization choices. No new physical entities are introduced; the four modules are ML constructs, not physical postulates.

free parameters (5)
  • All CoNN trainable weights = ≈7.4×10^4 parameters
    Trained by gradient descent to minimize MSE on AME2020 (Sec. II C); the headline accuracy is a fit on the pool that includes training nuclei, with separate held-out subsets.
  • Shell embedding vectors eZ and eN = 302 scalar values (Z=0..120, N=0..180)
    Each per-nucleon-number scalar is learned to absorb per-Z/N structure; the magic-number extrema in Fig. 4 are properties of these fitted values, not independent predictions.
  • Regional correlation grid G = 50×60 = 3000 values
    Bilinear-interpolated learned table; its resolution controls smoothness and is chosen by hand, not from an automated criterion.
  • Pairing MLP weights = One hidden layer, width 32
    Learns odd-even staggering from parity inputs; fitted to the residual targets.
  • Hyperparameters (bottleneck width, grid size, learning-rate ratio, weight decay, ensemble size, split ratio) = 16; 50×60; 10:1; 10^-3; 5; 80:20
    Hand-chosen; the 10:1 learning-rate ratio is described as the key asymmetry enforcing the decomposition, but no sensitivity analysis is provided.
assumptions (5)
  • domain assumption Macroscopic-microscopic additivity of the mass surface (B ≈ EMacro + EShell + ECor + EPair)
    Eq. (1) assumes this additive decomposition; it is a modeling assumption inherited from Strutinsky-type macroscopic-microscopic models, not proven in this paper.
  • domain assumption AME2020/AME2016 tabulated binding energies are ground truth
    All targets and comparisons use AME evaluations without propagating their uncertainties or evaluating systematic errors in the compilation.
  • domain assumption Spectral bias plus bottleneck makes the fully-connected branch smooth enough to represent the bulk term
    Relies on optimization behavior of fully-connected networks (Refs. [38,39]); this is an empirical tendency, not a guaranteed separation property.
  • domain assumption Temporal split has no leakage
    Assumes the 122 AME2020-added nuclei are entirely out-of-sample relative to AME2016 training data; reasonable, but AME revisions can affect earlier values, so perfect separation is not proven.
  • domain assumption Validation-based model selection yields an unbiased interpolation estimate
    The 688-nucleus held-out subset was used to select the best model state (Sec. II C), so 0.419 MeV is not a fully independent test; repeated splits would be required for an unbiased estimate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Physics-structured cooperative neural network for baseline-free nuclear mass modeling." pith.science (2026). https://pith.science/paper/3QTCXKRC

@misc{pith2026260309747,
  author       = {Pith},
  title        = {Pith review of: Physics-structured cooperative neural network for baseline-free nuclear mass modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3QTCXKRC}},
  note         = {Machine review of arXiv:2603.09747}
}
read the original abstract

Machine learning approaches can improve nuclear mass modelling, but the most accurate strategies often depend on a theoretical mass baseline or hand-crafted physics features. We test whether a modular architecture encoding selected nuclear-structure priors improves baseline-free direct prediction and yields informative branch diagnostics. The Cooperative Neural Network (CoNN) implements this approach through four form-constrained branches: a smooth macroscopic network, discrete embeddings, a two-dimensional regional grid, and a parity-aware network. It extracts complementary patterns from (Z, N) through these branches and sums their outputs to predict binding energies without a theoretical mass-model baseline. Thus, the model retains physics priors while reducing its reliance on engineered input features. On AME2020, CoNN reaches a root-mean-square deviation (RMSD) of 0.269 MeV for 3558 nuclei, compared with 0.836 MeV for a parameter-matched unstructured MLP. It also gives RMSDs of 0.419 MeV on a held-out interpolation subset and 0.728 MeV on 122 nuclei newly measured since AME2016. The learned branch outputs show recognizable physical patterns, including embedding shell-kink signatures at major magic numbers and odd-even staggering along isotopic chains. These results identify architecture-level priors as a practical route to baseline-free mass prediction, with learned components that help diagnose both nuclear-structure patterns and extrapolation limits.

Figures

Figures reproduced from arXiv: 2603.09747 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p006_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_p007_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references

  1. [1]

    Macroscopic step : freeze the microscopic modules and train the macroscopic branch on the target yMacro = ( Bexp − E∗ Shell − E∗ Cor − E∗ Pair − µy)/σy, where asterisks denote values held fixed during the current step, with Adam ( αMacro = 2 × 10−4)

  2. [2]

    Baseline

    Microscopic step : freeze the macroscopic branch and train all microscopic modules on the residual yMicro = Bexp − E∗ Macro with AdamW [52] (αMicro = 2 × 10−3, weight decay 10−3). The 10:1 learning-rate ratio αMicro/αMacro is the key asymmetry: it anchors the macroscopic branch to smooth trends while allowing the microscopic modules to adapt more rapidly ...

  3. [3]

    Lunney, J

    D. Lunney, J. M. Pearson, and C. Thibault,Rev. Mod. Phys. 75, 1021 (2003)

    2003

  4. [4]

    M.Mumpower, R.Surman, G.McLaughlin,andA.Apra- hamian, Prog. Part. Nucl. Phys.86, 86 (2016)

    2016

  5. [5]

    Martin, A

    D. Martin, A. Arcones, W. Nazarewicz, and E. Olsen, Phys. Rev. Lett.116, 121101 (2016)

    2016

  6. [6]

    J. J. Cowan, C. Sneden, J. E. Lawler, A. Aprahamian, M. Wiescher, K. Langanke, G. Martínez-Pinedo, and F.- K. Thielemann,Rev. Mod. Phys.93, 015002 (2021)

    2021

  7. [7]

    M. Wang, W. Huang, F. Kondev, G. Audi, and S. Naimi, Chin. Phys. C45, 030003 (2021)

    2021

  8. [8]

    Neufcourt, Y

    L. Neufcourt, Y. Cao, S. A. Giuliani, W. Nazarewicz, E. Olsen, and O. B. Tarasov,Phys. Rev. C101, 044307 (2020)

    2020

Show all 68 references
  1. [9]

    Erler, N

    J. Erler, N. Birge, M. Kortelainen, W. Nazarewicz, E. Olsen, A. M. Perhac, and M. Stoitsov,Nature 486, 509 (2012)

    2012

  2. [10]

    Thoennessen,Rep

    M. Thoennessen,Rep. Prog. Phys.67, 1187 (2004)

    2004

  3. [11]

    Sobiczewski, Y

    A. Sobiczewski, Y. A. Litvinov, and M. Palczewski,At. Data Nucl. Data Tables119, 1 (2018)

    2018

  4. [12]

    M.Bender, P.-H.Heenen,andP.-G.Reinhard, Rev.Mod. Phys. 75, 121 (2003)

    2003

  5. [13]

    J. Meng, H. Toki, S. Zhou, S. Zhang, W. Long, and L. Geng,Prog. Part. Nucl. Phys.57, 470 (2006)

    2006

  6. [14]

    Goriely, S

    S. Goriely, S. Hilaire, M. Girod, and S. Péru,Phys. Rev. Lett. 102, 242501 (2009)

    2009

  7. [15]

    Goriely, N

    S. Goriely, N. Chamel, and J. M. Pearson,Phys. Rev. C 93, 034337 (2016)

    2016

  8. [16]

    Strutinsky,Nucl

    V. Strutinsky,Nucl. Phys. A95, 420 (1967)

    1967

  9. [17]

    Data Tables109–110, 1 (2016)

    P.Möller, A.Sierk, T.Ichikawa,andH.Sagawa, At.Data Nucl. Data Tables109–110, 1 (2016)

    2016

  10. [18]

    N. Wang, M. Liu, X. Wu, and J. Meng,Phys. Lett. B 10 734, 215 (2014)

    2014

  11. [19]

    Brack, J

    M. Brack, J. Damgaard, A. S. Jensen, H. C. Pauli, V. M. Strutinsky, and C. Y. Wong,Rev. Mod. Phys. 44, 320 (1972)

    1972

  12. [20]

    Carleo, I

    G. Carleo, I. Cirac, K. Cranmer, L. Daudet, M. Schuld, N. Tishby, L. Vogt-Maranto, and L. Zdeborová, Rev. Mod. Phys. 91, 045002 (2019)

    2019

  13. [21]

    Boehnlein, M

    A. Boehnlein, M. Diefenthaler, N. Sato, M. Schram, V. Ziegler, C. Fanelli, M. Hjorth-Jensen, T. Horn, M. P. Kuchera, D. Lee, W. Nazarewicz, P. Ostroumov, K. Orginos, A. Poon, X.-N. Wang, A. Scheinker, M. S. Smith, and L.-G. Pang, Rev. Mod. Phys. 94, 031003 (2022)

    2022

  14. [22]

    Utama, J

    R. Utama, J. Piekarewicz, and H. B. Prosper,Phys. Rev. C 93, 014311 (2016)

    2016

  15. [23]

    Utama and J

    R. Utama and J. Piekarewicz,Phys. Rev. C96, 044308 (2017)

    2017

  16. [24]

    Neufcourt, Y

    L. Neufcourt, Y. Cao, W. Nazarewicz, and F. Viens, Phys. Rev. C98, 034318 (2018)

    2018

  17. [25]

    Niu and H

    Z. Niu and H. Liang,Phys. Lett. B778, 48 (2018)

    2018

  18. [26]

    Z. M. Niu and H. Z. Liang,Phys. Rev. C106, L021303 (2022)

    2022

  19. [27]

    X. Wu, Y. Lu, and P. Zhao,Phys. Lett. B834, 137394 (2022)

    2022

  20. [28]

    Yüksel, D

    E. Yüksel, D. Soydaner, and H. Bahtiyar,Phys. Rev. C 109, 064322 (2024)

    2024

  21. [29]

    Y. Lu, T. Shang, P. Du, J. Li, H. Liang, and Z. Niu, Phys. Rev. C111, 014325 (2025)

    2025

  22. [30]

    Huang, J

    Y. Huang, J. Chen, J. Jia, L.-M. Liu, Y.-G. Ma, and C. Zhang,Phys. Rev. C111, 034329 (2025)

    2025

  23. [31]

    Jalili, F

    A. Jalili, F. Pan, A. X. Chen, and J. P. Draayer,Phys. Rev. C 112, 024305 (2025)

    2025

  24. [32]

    A. E. Lovell, A. T. Mohan, T. M. Sprouse, and M. R. Mumpower, Phys. Rev. C106, 014305 (2022)

    2022

  25. [33]

    Athanassopoulos, E

    S. Athanassopoulos, E. Mavrommatis, K. A. Gernoth, and J. W. Clark,Nucl. Phys. A743, 222 (2004)

    2004

  26. [34]

    Y. Y. Huang and X. H. Wu,Phys. Lett. B874, 140262 (2026)

    2026

  27. [35]

    M. R. Mumpower, T. M. Sprouse, A. E. Lovell, and A. T. Mohan, Phys. Rev. C106, L021301 (2022)

    2022

  28. [36]

    L.-X.Zeng, Y.-Y.Yin, X.-X.Dong,andL.-S.Geng, Phys. Rev. C 109, 034318 (2024)

    2024

  29. [37]

    M. Wang, G. Audi, F. G. Kondev, W. Huang, S. Naimi, and X. Xu,Chin. Phys. C41, 030003 (2017)

    2017

  30. [38]

    W. D. Myers and W. J. Swiatecki, Nucl. Phys. 81, 1 (1966)

    1966

  31. [39]

    G. E. Hinton and R. R. Salakhutdinov,Science 313, 504 (2006)

    2006

  32. [40]

    Rahaman, A

    N. Rahaman, A. Baratin, D. Arpit, F. Draxler, M. Lin, F. Hamprecht, Y. Bengio, and A. Courville, Proc. ICML, PMLR 97, 5301 (2019)

    2019

  33. [41]

    Z.-Q. J. Xu, Y. Zhang, T. Luo, Y. Xiao, and Z. Ma, Commun. Comput. Phys.28, 1746 (2020)

    2020

  34. [42]

    Buskirk, K

    L. Buskirk, K. Godbey, W. Nazarewicz, and W. Satuła, Phys. Rev. C109, 044311 (2024)

    2024

  35. [43]

    M. G. Mayer,Phys. Rev. 75, 1969 (1949)

    1969

  36. [44]

    Haxel, J

    O. Haxel, J. H. D. Jensen, and H. E. Suess,Phys. Rev. 75, 1766 (1949)

    1949

  37. [45]

    Federman and S

    P. Federman and S. Pittel,Phys. Lett. B69, 385 (1977)

    1977

  38. [46]

    R. F. Casten,Nucl. Phys. A443, 1 (1985)

    1985

  39. [47]

    A. Bohr, B. R. Mottelson, and D. Pines,Phys. Rev.110, 936 (1958)

    1958

  40. [48]

    H. Liu, J. Lei, and Z. Ren,Phys. Rev. C 111, 024316 (2025)

    2025

  41. [49]

    R. A. Jacobs, M. I. Jordan, S. J. Nowlan, and G. E. Hinton, Neural Comput. 3, 79 (1991)

    1991

  42. [50]

    Piratla, P

    V. Piratla, P. Netrapalli, and S. Sarawagi, Proc. ICML, PMLR 119, 7728 (2020)

    2020

  43. [51]

    S. J. Wright,Math. Program. 151, 3 (2015)

    2015

  44. [52]

    J. Zeng, T. T.-K. Lau, S. Lin, and Y. Yao, Proc. ICML, PMLR 97, 7313 (2019)

    2019

  45. [53]

    D. P. Kingma and J. Ba, inICLR (2015)

    2015

  46. [54]

    Loshchilov and F

    I. Loshchilov and F. Hutter, inICLR (2019)

    2019

  47. [55]

    Loshchilov and F

    I. Loshchilov and F. Hutter, inICLR (2017)

    2017

  48. [56]

    Lakshminarayanan, A

    B. Lakshminarayanan, A. Pritzel, and C. Blundell, Adv. Neural Inf. Process. Syst.30, 6405 (2017)

    2017

  49. [57]

    Hüllermeier and W

    E. Hüllermeier and W. Waegeman,Mach. Learn. 110, 457 (2021)

    2021

  50. [58]

    Izmailov, S

    P. Izmailov, S. Vikram, M. D. Hoffman, and A. G. Wil- son, Proc. ICML, PMLR139, 4629 (2021)

    2021

  51. [59]

    Rahaman and A

    R. Rahaman and A. Thiery, Adv. Neural Inf. Process. Syst. 34, 20063 (2021)

    2021

  52. [60]

    J. Y. Zhang, R. F. Casten, and D. S. Brenner,Phys. Lett. B 227, 1 (1989)

    1989

  53. [61]

    R. B. Cakirli, D. S. Brenner, R. F. Casten, and E. A. Millman, Phys. Rev. Lett.94, 092501 (2005)

    2005

  54. [62]

    D. S. Brenner, R. B. Cakirli, and R. F. Casten,Phys. Rev. C 73, 034315 (2006)

    2006

  55. [63]

    Bohr and B

    A. Bohr and B. R. Mottelson,Nuclear Structure, Vol. II: Nuclear Deformations (W. A. Benjamin, Reading, MA, 1974)

    1974

  56. [64]

    Bohr and B

    A. Bohr and B. R. Mottelson, Nuclear Structure, Vol. I: Single-Particle Motion (W. A. Benjamin, New York, 1969)

    1969

  57. [65]

    Sobiczewski and Y

    A. Sobiczewski and Y. A. Litvinov, Phys. Rev. C 90, 017302 (2014)

    2014

  58. [66]

    694,A180 (2025)

    S.MartinetandS.Goriely, Astron.Astrophys. 694,A180 (2025)

    2025

  59. [67]

    Saito, I

    Y. Saito, I. Dillmann, R. Krücken, M. R. Mumpower, and R. Surman,Phys. Rev. C109, 054301 (2024)

    2024

  60. [68]

    D. J. C. MacKay,Neural Comput. 4, 448 (1992)

    1992

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.