pith. sign in

arxiv: 1907.09764 · v1 · pith:YHPJETXOnew · submitted 2019-07-23 · ⚛️ nucl-th · cs.LG· nucl-ex· stat.ML

Trees and Islands -- Machine learning approach to nuclear physics

Nishchal R. Dwivedi This is my paper

Pith reviewed 2026-05-24 17:14 UTC · model grok-4.3

classification ⚛️ nucl-th cs.LGnucl-exstat.ML
keywords machine learningnuclear datagradient boosted treeslevel density parametersuperheavy elementsshell correction energiesquadrupole deformationpairing gaps
0
0 comments X

The pith

Gradient boosted trees trained on nuclear data predict parameters such as level densities for superheavy elements.

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

The paper applies gradient boosted trees to existing nuclear data to build models that predict multiple properties including damping parameters, shell correction energies, quadrupole deformations, pairing gaps, level densities, and giant dipole resonances. It pays special attention to level density parameters for superheavy elements, where measured values are scarce. A sympathetic reader would care because these predictions offer a way to estimate nuclear behavior in regions that are hard to reach experimentally. The models produce predictions whose standard deviations range from 0.00035 to 0.73 relative to the training data.

Core claim

A purely data-driven gradient boosted trees model trained on existing nuclear measurements generates predictions for damping parameter, shell correction energies, quadrupole deformation, pairing gaps, level densities and giant dipole resonance across large numbers of nuclei, with particular results for level density parameters in superheavy elements.

What carries the argument

Gradient boosted trees algorithm that builds successive decision trees on nuclear data to capture intricate trends and produce predictions.

If this is right

  • Level density parameters for superheavy elements can be estimated without new measurements.
  • Similar predictions become available for other nuclear properties where data remains limited.
  • A data-driven approach supplies estimates for damping parameters, shell corrections, and deformations across many nuclei.
  • The trained model serves as a tool to fill gaps in nuclear tables for nuclei not yet measured.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the predictions hold, experiments on rare isotopes could be prioritized using the model outputs as guides.
  • The same training strategy might be tested on other sparse-data domains such as exotic isotopes or reaction rates.
  • Direct comparison of model outputs against future measurements would test how far the training set can be extrapolated.

Load-bearing premise

The existing nuclear data used for training is representative and sufficient for the model to generalize accurately to nuclei outside the training set, especially superheavy elements where experimental data is sparse.

What would settle it

New experimental measurements of level density parameters in superheavy elements that fall well outside the predicted range with standard deviations larger than 0.73.

Figures

Figures reproduced from arXiv: 1907.09764 by Nishchal R. Dwivedi.

Figure 1
Figure 1. Figure 1: Predictions for the test set for γ values show a standard deviation of 0.00035 and a standard error of order 10−5 . The data is trained and tested on a data set of 290 nuclei. −15 −10 −5 0 5 0 −15 −10 −5 0 5 Predicted Value Actual Value [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The predictions of Shell correction energy give a standard deviation of 0.553 and standard error of 0.0067. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Prediction on 8983 values of calculated Quadrupole deformations, [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Prediction of test set for pairing gaps for proton. The model is trained and tested on 8979 nuclei and give a [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Prediction of test set for pairing gaps for neutron. The model is trained and tested on 8979 nuclei and give a [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: By training and testing Level Density Parameter by the GC model for 289 nuclei, we get the standard [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Training and testing on the temperature dependent Level Density Parameter by Semiclassical trace formula [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Training and testing the experimental first peak energy values for GDR for 180 nuclei. It shows the standard [PITH_FULL_IMAGE:figures/full_fig_p007_8.png] view at source ↗
read the original abstract

We implement machine learning algorithms to nuclear data. These algorithms are purely data driven and generate models that are capable to capture intricate trends. Gradient boosted trees algorithm is employed to generate a trained model from existing nuclear data, which is used for prediction for data of damping parameter, shell correction energies, quadrupole deformation, pairing gaps, level densities and giant dipole resonance for large number of nuclei. We, in particular, predict level density parameter for superheavy elements which is of great current interest. The predictions made by the machine learning algorithm is found to have standard deviation from 0.00035 to 0.73.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper applies gradient boosted trees to existing nuclear data in a purely data-driven manner to train models that predict damping parameters, shell correction energies, quadrupole deformations, pairing gaps, level densities, and giant dipole resonances across many nuclei, with particular emphasis on level-density parameters for superheavy elements; the predictions are stated to exhibit standard deviations ranging from 0.00035 to 0.73.

Significance. If the generalization claims were substantiated, the work would illustrate a practical data-driven route to estimating nuclear observables in data-sparse regions. At present the manuscript supplies no evidence that the reported errors reflect out-of-distribution performance, so the significance cannot yet be assessed.

major comments (3)
  1. [Abstract] Abstract: the quoted standard deviations (0.00035–0.73) are presented without any statement of the train/test split, cross-validation scheme, or whether the metric was computed on held-out nuclei; this information is required to evaluate whether the numbers measure generalization rather than in-sample fit.
  2. [Abstract] Abstract and methods description: the model is characterized as “purely data driven” with no external physical benchmark or comparison to established nuclear models (e.g., liquid-drop or microscopic calculations), so the reliability of extrapolation to superheavy nuclei—where experimental data are absent by definition—remains untested.
  3. [Results] Results section (implied by the abstract claim): no performance metrics are supplied for heavy or deformed systems deliberately withheld from training, which is the regime the headline prediction targets; without such a test the central claim that the algorithm produces “usable predictions” for superheavy elements cannot be evaluated.
minor comments (2)
  1. [Abstract] Abstract: grammatical error—“The predictions made by the machine learning algorithm is found to have” should read “are found to have.”
  2. [Abstract] Abstract: the standard-deviation range is given without units or an indication of which predicted quantities correspond to the lower versus upper end of the interval.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments that identify key areas for improving the clarity and validation of our work. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the quoted standard deviations (0.00035–0.73) are presented without any statement of the train/test split, cross-validation scheme, or whether the metric was computed on held-out nuclei; this information is required to evaluate whether the numbers measure generalization rather than in-sample fit.

    Authors: We agree that this information is essential. The standard deviations were obtained via k-fold cross-validation on the nuclear datasets to estimate out-of-sample performance. In the revised manuscript we will explicitly describe the train/test protocol and cross-validation procedure in both the abstract and methods section. revision: yes

  2. Referee: [Abstract] Abstract and methods description: the model is characterized as “purely data driven” with no external physical benchmark or comparison to established nuclear models (e.g., liquid-drop or microscopic calculations), so the reliability of extrapolation to superheavy nuclei—where experimental data are absent by definition—remains untested.

    Authors: The manuscript intentionally highlights a purely data-driven route. To strengthen the assessment of reliability we will add, in the revised version, direct comparisons of the machine-learning predictions against liquid-drop and microscopic model results for nuclei where independent calculations exist, together with a discussion of implications for the superheavy regime. revision: yes

  3. Referee: [Results] Results section (implied by the abstract claim): no performance metrics are supplied for heavy or deformed systems deliberately withheld from training, which is the regime the headline prediction targets; without such a test the central claim that the algorithm produces “usable predictions” for superheavy elements cannot be evaluated.

    Authors: We acknowledge that the present manuscript lacks this targeted validation. We will perform and report an additional experiment in which data from heavy and deformed nuclei are withheld from training, supplying performance metrics on those held-out cases to directly support the claims for superheavy elements. revision: yes

Circularity Check

0 steps flagged

No circularity: standard supervised ML workflow with no self-referential derivation or fitted-input renaming

full rationale

The paper presents an explicitly data-driven gradient-boosted-trees model trained on existing nuclear datasets to generate predictions for quantities such as level-density parameters. No first-principles derivation, uniqueness theorem, or ansatz is invoked; the workflow is the ordinary supervised-learning pipeline in which a fitted function is applied to new inputs. Reported standard deviations are model-evaluation metrics on the training distribution and do not reduce any claimed physical result to its own inputs by construction. None of the enumerated circularity patterns (self-definitional, fitted-input-called-prediction, self-citation load-bearing, etc.) are exhibited by the quoted abstract or described method.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim depends on the quality and representativeness of the training nuclear data and the generalization ability of the ML model. No specific free parameters or invented entities are detailed in the abstract.

free parameters (1)
  • model hyperparameters
    The gradient boosted trees algorithm has multiple hyperparameters that are typically tuned to the data, though not specified in the abstract.
axioms (1)
  • domain assumption Nuclear properties exhibit patterns that can be learned from existing data without explicit physical modeling.
    The paper relies on this to justify the data-driven approach.

pith-pipeline@v0.9.0 · 5627 in / 1284 out tokens · 30757 ms · 2026-05-24T17:14:57.558612+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages

  1. [1]

    Uncovering social spammers: social honeypots+ machine learn- ing

    Kyumin Lee, James Caverlee, and Steve Webb. Uncovering social spammers: social honeypots+ machine learn- ing. In Proceedings of the 33rd international ACM SIGIR conference on Research and development in informa- tion retrieval, pages 435–442. ACM, 2010

    work page 2010

  2. [2]

    Deepwalk: Online learning of social representations

    Bryan Perozzi, Rami Al-Rfou, and Steven Skiena. Deepwalk: Online learning of social representations. In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining , pages 701–710. ACM, 2014

    work page 2014

  3. [3]

    Forecasting stock market movement direction with support vector machine

    Wei Huang, Yoshiteru Nakamori, and Shou-Yang Wang. Forecasting stock market movement direction with support vector machine. Computers & Operations Research, 32(10):2513–2522, 2005

    work page 2005

  4. [4]

    Personalized news recommendation based on click behavior

    Jiahui Liu, Peter Dolan, and Elin Rønby Pedersen. Personalized news recommendation based on click behavior. In Proceedings of the 15th international conference on Intelligent user interfaces , pages 31–40. ACM, 2010

    work page 2010

  5. [5]

    Smartads: bringing contextual ads to mobile apps

    Suman Nath, Felix Xiaozhu Lin, Lenin Ravindranath, and Jitendra Padhye. Smartads: bringing contextual ads to mobile apps. In Proceeding of the 11th annual international conference on Mobile systems, applications, and services, pages 111–124. ACM, 2013

    work page 2013

  6. [6]

    Collaborative tagging as distributed cognition

    Luc Steels. Collaborative tagging as distributed cognition. Pragmatics & cognition, 14(2):287–292, 2006

    work page 2006

  7. [7]

    Searching for exotic particles in high-energy physics with deep learning

    Pierre Baldi, Peter Sadowski, and Daniel Whiteson. Searching for exotic particles in high-energy physics with deep learning. Nature communications, 5:4308, 2014

    work page 2014

  8. [8]

    Adaptive machine learning framework to accelerate ab initio molecular dynamics

    Venkatesh Botu and Rampi Ramprasad. Adaptive machine learning framework to accelerate ab initio molecular dynamics. International Journal of Quantum Chemistry , 115(16):1074–1083, 2015

    work page 2015

  9. [9]

    Greedy function approximation: a gradient boosting machine

    Jerome H Friedman. Greedy function approximation: a gradient boosting machine. Annals of statistics , pages 1189–1232, 2001

    work page 2001

  10. [10]

    Application of support vector machines to global prediction of nuclear properties

    John W Clark and Haochen Li. Application of support vector machines to global prediction of nuclear properties. International Journal of Modern Physics B , 20(30n31):5015–5029, 2006

    work page 2006

  11. [11]

    Nuclear charge radii: density functional theory meets bayesian neural networks

    Raditya Utama, Wei-Chia Chen, and Jorge Piekarewicz. Nuclear charge radii: density functional theory meets bayesian neural networks. Journal of Physics G: Nuclear and Particle Physics , 43(11):114002, 2016

    work page 2016

  12. [12]

    Energy levels of light nuclei a= 3-15

    F Ajzenberg-Selove. Energy levels of light nuclei a= 3-15. Nucl. Phys. A, 268:1, 1976

    work page 1976

  13. [13]

    Nuclear physics a

    Hans Albrecht Bethe and Robert Fox Bacher. Nuclear physics a. stationary states of nuclei. Reviews of Modern Physics, 8(2):82, 1936

    work page 1936

  14. [14]

    Zur theorie der kernmassen

    CF v Weizs ¨acker. Zur theorie der kernmassen. Zeitschrift f ¨ur Physik A Hadrons and Nuclei , 96(7):431–458, 1935. 8 A PREPRINT -

    work page 1935

  15. [15]

    Fission barriers within the liquid drop model with the surface-curvature term

    K Pomorski and J Dudek. Fission barriers within the liquid drop model with the surface-curvature term. Inter- national Journal of Modern Physics E , 13(01):107–112, 2004

    work page 2004

  16. [16]

    Heavy-ion reactions close to the coulomb barrier

    W Reisdorf. Heavy-ion reactions close to the coulomb barrier. Journal of Physics G: Nuclear and Particle Physics, 20(9):1297, 1994

    work page 1994

  17. [17]

    An attempt to calculate the number of energy levels of a heavy nucleus

    HA Bethe. An attempt to calculate the number of energy levels of a heavy nucleus. Physical Review, 50(4):332, 1936

    work page 1936

  18. [18]

    shells in deformed nuclei

    VM Strutinsky. shells in deformed nuclei. Nuclear Physics A, 122(1):1–33, 1968

    work page 1968

  19. [19]

    Phenomenological description of energy dependence of the level density parameter

    A V Ignatyuk, GN Smirenkin, and AS Tishin. Phenomenological description of energy dependence of the level density parameter. Yadernaya Fizika, 21(3):485–490, 1975

    work page 1975

  20. [20]

    Thermal Properties of Nuclei and Their Level Densities

    Md Al Mamun et al. Thermal Properties of Nuclei and Their Level Densities. PhD thesis, Ohio University, 2015

    work page 2015

  21. [21]

    A composite nuclear-level density formula with shell corrections

    A Gilbert and AGW Cameron. A composite nuclear-level density formula with shell corrections. Canadian Journal of Physics, 43(8):1446–1496, 1965

    work page 1965

  22. [22]

    https://www-nds.iaea.org

    work page

  23. [23]

    Fermi-gas model parametrization of nuclear level density

    Alberto Mengoni and Yutaka Nakajima. Fermi-gas model parametrization of nuclear level density. Journal of Nuclear Science and Technology, 31(2):151–162, 1994

    work page 1994

  24. [24]

    Nuclear structure, volume 2

    Aage Bohr and Ben R Mottelson. Nuclear structure, volume 2. World Scientific, 1975

    work page 1975

  25. [25]

    Atomic data and nuclear data tablesatomic data and nuclear data tables

    Peter M ¨oller, JR Nix, WD Myers, and WJ Swiatecki. Atomic data and nuclear data tablesatomic data and nuclear data tables. International Journal of Modern Physics B , 59:185–383, 1995

    work page 1995

  26. [26]

    Possible analogy between the excitation spectra of nuclei and those of the superconducting metallic state

    ˚A Bohr, BR Mottelson, and D Pines. Possible analogy between the excitation spectra of nuclei and those of the superconducting metallic state. Physical Review, 110(4):936, 1958

    work page 1958

  27. [27]

    Nuclear pairing models

    P M ¨oller and JR Nix. Nuclear pairing models. Nuclear Physics A, 536(1):20–60, 1992

    work page 1992

  28. [28]

    https://t2.lanl.gov/

    work page

  29. [29]

    Statistics of nuclear levels

    JMB Lang and KJ Le Couteur. Statistics of nuclear levels. Proceedings of the Physical Society. Section A , 67(7):586, 1954

    work page 1954

  30. [30]

    Semiclassical theory of melting of shell effects in nuclei with temperature

    Harjeet Kaur and Sudhir R Jain. Semiclassical theory of melting of shell effects in nuclei with temperature. Journal of Physics G: Nuclear and Particle Physics , 42(11):115103, 2015

    work page 2015

  31. [31]

    Analytical tests of gutzwillers trace formula for harmonic-oscillator potentials

    Matthias Brack and Sudhir R Jain. Analytical tests of gutzwillers trace formula for harmonic-oscillator potentials. Physical Review A, 51(5):3462, 1995

    work page 1995

  32. [32]

    Semiclassical trace formulae for systems with spin–orbit interactions: successes and limitations of present approaches

    Ch Amann and Matthias Brack. Semiclassical trace formulae for systems with spin–orbit interactions: successes and limitations of present approaches. Journal of Physics A: Mathematical and General , 35(29):6009, 2002

    work page 2002

  33. [33]

    Funny hills: The shell- correction approach to nuclear shell effects and its applications to the fission process.Reviews of Modern Physics, 44(2):320, 1972

    Matthias Brack, Jens Damgaard, AS Jensen, HC Pauli, VM Strutinsky, and CY Wong. Funny hills: The shell- correction approach to nuclear shell effects and its applications to the fission process.Reviews of Modern Physics, 44(2):320, 1972

    work page 1972

  34. [34]

    The discovery of the heaviest elements.Reviews of Modern Physics, 72(3):733, 2000

    Sigurd Hofmann and Gottfried M ¨unzenberg. The discovery of the heaviest elements.Reviews of Modern Physics, 72(3):733, 2000

    work page 2000

  35. [35]

    Level densities and shell corrections of superheavy nuclei

    AN Bezbakh, TM Shneidman, GG Adamian, and NV Antonenko. Level densities and shell corrections of superheavy nuclei. In Journal of Physics: Conference Series , volume 580, page 012026. IOP Publishing, 2015

    work page 2015

  36. [36]

    Isoscalar giant dipole resonance in 90 zr, 116 sn, and 208 pb

    HL Clark, Y-W Lui, and David H Youngblood. Isoscalar giant dipole resonance in 90 zr, 116 sn, and 208 pb. Physical Review C, 63(3):031301, 2001

    work page 2001

  37. [37]

    Measurements of the giant dipole resonance with monoenergetic photons.Reviews of Modern Physics, 47(3):713, 1975

    B Lo Berman and SoCo Fultz. Measurements of the giant dipole resonance with monoenergetic photons.Reviews of Modern Physics, 47(3):713, 1975

    work page 1975

  38. [38]

    Application of extreme learning machine method for time series analysis

    Rampal Singh and S Balasundaram. Application of extreme learning machine method for time series analysis. International Journal of Intelligent Technology, 2(4):256–262, 2007

    work page 2007

  39. [39]

    Fast and accurate modeling of molecular atomization energies with machine learning

    Matthias Rupp, Alexandre Tkatchenko, Klaus-Robert M ¨uller, and O Anatole V on Lilienfeld. Fast and accurate modeling of molecular atomization energies with machine learning. Physical review letters , 108(5):058301, 2012

    work page 2012

  40. [40]

    A novel method to estimate model uncertainty using machine learning techniques

    Dimitri P Solomatine and Durga Lal Shrestha. A novel method to estimate model uncertainty using machine learning techniques. Water Resources Research, 45(12), 2009. 9 A PREPRINT -

    work page 2009

  41. [41]

    Using machine learning algorithms for housing price prediction: The case of fairfax county, virginia housing data

    Byeonghwa Park and Jae Kwon Bae. Using machine learning algorithms for housing price prediction: The case of fairfax county, virginia housing data. Expert Systems with Applications , 42(6):2928–2934, 2015

    work page 2015

  42. [42]

    Oil price prediction using ensemble machine learning

    Lubna A Gabralla, Rania Jammazi, and Ajith Abraham. Oil price prediction using ensemble machine learning. In 2013 International Conference on Computing, Electrical and Electronic Engineering (ICCEEE), pages 674–679. IEEE, 2013

    work page 2013

  43. [43]

    Application of the extreme learning machine algorithm for the prediction of monthly effective drought index in eastern australia

    Ravinesh C Deo and Mehmet S ¸ahin. Application of the extreme learning machine algorithm for the prediction of monthly effective drought index in eastern australia. Atmospheric Research, 153:512–525, 2015

    work page 2015

  44. [44]

    Random matrices, volume 142

    Madan Lal Mehta. Random matrices, volume 142. Elsevier, 2004. 10

    work page 2004