Nuclear mass predictions using machine learning models

Abstract

The exploration of nuclear mass or binding energy, a fundamental property of atomic nuclei, remains at the forefront of nuclear physics research due to limitations in experimental studies and uncertainties in model calculations, particularly when moving away from the stability line. In this work, we employ two machine learning (ML) models, Support Vector Regression (SVR) and Gaussian Process Regression (GPR), to assess their performance in predicting nuclear mass excesses using available experimental data and a physics-based feature space. We also examine the extrapolation capabilities of these models using newly measured nuclei from AME2020 and by extending our calculations beyond the training and test set regions. Our results indicate that both SVR and GPR models perform quite well within the training and test regions when informed with a physics-based feature space. Furthermore, these ML models demonstrate the ability to make reasonable predictions away from the available experimental data, offering results comparable to the model calculations. Through further refinement, these models can be used as reliable and efficient ML tools for studying nuclear properties in the future.

arXiv comment12 pages, 8 figures

References in corpus (22)

• Nuclear ground-state masses and deformations: FRDM(2012)
• Surface diffuseness correction in global mass formula
• DIRHB -- a relativistic self-consistent mean-field framework for atomic nuclei
• Global performance of covariant energy density functionals: ground state observables of even-even nuclei and the estimate of theoretical uncertainties
• Machine Learning in Nuclear Physics
• Nuclear mass predictions based on Bayesian neural network approach with pairing and shell effects
• The limits of the nuclear landscape explored by the relativistic continuum Hatree-Bogoliubov theory
• Bayesian approach to model-based extrapolation of nuclear observables
• Neutron drip line in the Ca region from Bayesian model averaging
• Impact of nuclear mass uncertainties on the $r$-process
• The neutron drip line: single-particle degrees of freedom and pairing properties as sources of theoretical uncertainties
• Refining mass formulas for astrophysical applications: a Bayesian neural network approach
• The impact of individual nuclear masses on $r$-process abundances
• Nuclear mass systematics using neural networks
• Covariant energy density functionals: nuclear matter constraints and global ground state properties
• Optimizing the relativistic energy density functional with nuclear ground state and collective excitation properties
• Nuclear masses learned from a probabilistic neural network
• Physically Interpretable Machine Learning for nuclear masses
• Validating neural-network refinements of nuclear mass models
• Nuclear binding energy predictions using neural networks: Application of the multilayer perceptron
• Application Of Support Vector Machines To Global Prediction Of Nuclear Properties
• Unveiling The Factors of Aesthetic Preferences with Explainable AI

Cited by in corpus (7)

• Kolmogorov-Arnold networks in nuclear binding energy prediction
• High Precision Binding Energies from Physics Informed Machine Learning
• Quantifying uncertainty in machine learning on nuclear binding energy
• Further exploration of binding energy residuals using machine learning and the development of a composite ensemble model
• Bayesian model mixing with multi-reference energy density functional
• Product units in gated recurrent units improve nuclear-mass prediction
• Further exploration of the machine-learning-based nuclear mass table