https://doi.org/10.1140/epja/s10050-025-01610-9
Regular Article - Theoretical Physics
Performance of various kernel functions for mass prediction with support vector machine
1
Zhejiang Key Laboratory of Quantum State Control and Optical Field Manipulation, Department of Physics, Zhejiang Sci-Tech University, 310018, Hangzhou, China
2
School of Physics, Nankai University, 300071, Tianjin, P.R. China
3
Department of Physics, Liaoning Normal University, 116029, Dalian, P.R. China
4
Department of Physics and Astronomy, Louisiana State University, 70803-4001, Baton Rouge, LA, USA
a jalili@zstu.edu.cn, amir.jalili85@gmail.com
b
ziba_saleki@yahoo.com
Received:
23
January
2025
Accepted:
30
May
2025
Published online:
19
June
2025
This manuscript explores the potential of Support Vector Machines (SVM) in predicting nuclear binding energies, emphasizing the critical role of kernel functions in improving model accuracy. We systematically assess the performance of various SVM kernel functions-linear, polynomial, radial basis function, and sigmoid-through rigorous hyperparameter optimization and cross-validation. The study demonstrates that the radial basis function kernel outperforms other kernels, achieving the lowest root-mean-square deviation of 0.199 MeV, making it the most effective for nuclear mass predictions. By integrating key nuclear physics features, including mass model information, the SVM model is able to capture complex nuclear behaviors across different mass ranges. We present a comprehensive comparison of our SVM model against conventional mass models such as liquid drop model and WS4, where our SVM model shows improved predictive accuracy. This work underscores the significance of kernel selection in SVM and highlights the power of machine learning in advancing nuclear mass spectroscopy, providing a valuable framework for future computational modeling in nuclear physics.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to Società Italiana di Fisica and Springer-Verlag GmbH Germany, part of Springer Nature 2025
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
