Studies of different kernel functions in nuclear mass predictions with kernel ridge regression

Xin-rong, WU

doi:10.3389/fphy.2023.1061042

Cited by 5 publications

(3 citation statements)

References 92 publications

(127 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Generally, all machine-learning approaches lose their predictive power when extrapolating to large distances, and their abilities with the increasing extrapolation should be examined. The extrapolation ability of the KRR approach has been carefully examined in several previous works [62,63,65]. These works demonstrate that compared with other machine-learning approaches, the KRR approach loses its extrapolation power in a relatively gentle way with the increasing extrapolation distance.…”

Section: Discussionmentioning

confidence: 99%

“…Recently, a multi-task learning framework for nuclear masses and separation energies was developed by introducing gradient kernel functions to the KRR approach, which improves the predictions of both nuclear masses and separation energies [64]. The successful applications of the KRR approach in nuclear masses [62][63][64][65][66] have stimulated other applications in nuclear physics, including the energy density functionals [71], charge radii [72], and neutron-capture cross-sections [73].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Examination of machine learning for assessing physical effects: Learning the relativistic continuum mass table with kernel ridge regression*

Guo

Xin-rong

et al. 2023

Chinese Phys. C

View full text Add to dashboard Cite

The kernel ridge regression (KRR) method and its extension with odd-even effects (KRRoe) are used to learn nuclear mass table obtained by the relativistic continuum Hartree-Bogoliubov theory. With respect to the binding energies of 9035 nuclei, the KRR method has achieved a root-mean-square deviation of 0.96 MeV and the KRRoe method remarkably reduced the deviation to 0.17 MeV. By investigating the shell effects, one-nucleon and two-nucleon separation energies, odd-even mass differences, and empirical proton-neutron interactions extracted from the learned binding energies, the ability of the machine learning tool to grasp the known physics is discussed. It is found that the shell effects, the evolutions of nucleon separation energies and empirical proton-neutron interactions are well reproduced by both the KRR and KRRoe methods, although the odd-even mass differences can only be reproduced by the KRRoe method.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Examination of machine learning for assessing physical effects: Learning the relativistic continuum mass table with kernel ridge regression*

Guo

Xin-rong

et al. 2023

Chinese Phys. C

View full text Add to dashboard Cite

show abstract

“…应相关性质的研究 [21] , 如原子核的基态性质, 包括核质量(结合能) [22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38] 、电荷半径 [25,[39][40][41][42][43][44][45] 、磁矩 [46] 、分离能 [23,25,33,38] 、稳定性 [47,48] 、密度分布 [49][50][51] 、放射性核衰变分支比 [48,52] 等. 此外还有激发态 [53][54][55][56] 、 a衰变 [29,30,[57][58][59][60] 、b衰变 [52,[61][62][63]…”

Section: 目前机器学习已被广泛地应用于原子核结构和反unclassified

Prediction of ground-state spin in odd-<i>A</i> nuclei within decision tree

Wen,

Shang,

et al. 2023

Acta Phys. Sin.

View full text Add to dashboard Cite

As a fundamental property of nuclei, ground-state spin has always been a hot topic in nuclear data and structure research. This paper extensively investigates the odd mass nuclei (odd-A nuclei) on the nuclide chart using decision trees, including odd proton nuclei (odd-Z nuclei) and odd neutron nuclei (odd-N nuclei), and trains ground-state spin prediction models for odd-Z nuclei and odd-N nuclei, respectively. In the case of randomly dividing the training and validation sets at a ratio of 75\% to 25\%, the accuracy rates of the training and validation sets for odd-Z nuclei reach 98.9\% and 79.3\%, respectively; The accuracy of the training set and validation set for the odd-N nuclei reach 98.6\% and 71.6\%, respectively. At the same time, through 1000 random selections of training and validation sets for repeated validation, the standard error of the accuracy obtained can be less than 5\%, further verifying the reliability and generalization performance of the decision tree; On the other hand, the accuracy of decision trees is much higher than that of theoretical models commonly used in nuclear structure research, such as Skyrme Hartree-Fock-Bogoliubov(SHFB), covariant density functional theory(CDFT), finite range droplet model(FRDM), etc. Next, using all spin-determined odd-Z and odd-N nuclei as the learning set, a total of 254 spin undetermined but recommended odd-Z and 268 corresponding odd-N nuclei were predicted for their ground-state spin values, with predicted set coincidence rates of 68.5\% and 69.0\%, respectively. Finally, four odd mass number chains, i.e., Z=59, Z=77, N=41, and N=59, are selected to compare the learning (prediction) results of the decision tree with the experimental (recommended) values of the corresponding nuclei and the results given by the three theoretical models, further demonstrating the research and application value of the decision tree in the ground-state spin of nuclei.

show abstract

Principal components of nuclear mass models

Wu,

Zhao

2024

Sci. China Phys. Mech. Astron.

View full text Add to dashboard Cite

Studies of different kernel functions in nuclear mass predictions with kernel ridge regression

Cited by 5 publications

References 92 publications

Examination of machine learning for assessing physical effects: Learning the relativistic continuum mass table with kernel ridge regression*

Examination of machine learning for assessing physical effects: Learning the relativistic continuum mass table with kernel ridge regression*

Prediction of ground-state spin in odd-<i>A</i> nuclei within decision tree

Principal components of nuclear mass models

Contact Info

Product

Resources

About