分类: 物理学 >> 核物理学 提交时间: 2025-01-18
摘要: This study explores the use of machine learning techniques, particularly kernel ridge regression (KRR), to refine the estimation of semi-empirical mass formula (SEMF) coefficients, which represent volume energy, surface energy, Coulomb energy, asymmetry energy, and pairing energy. Traditional regression techniques, such as ordinary least square regression (OLSR), are limited by their parametric nature and susceptibility to over fitting, particularly when modeling complex relationships within high-dimensional data. By incorporating a penalty term and kernel functions, KRR mitigates over fitting, reduces variance, and enhances predictive accuracy. Mass number and atomic number data for 109 nuclei were analyzed, with model performance assessed through root mean square error and R2values. The findings demonstrate the superior robustness and accuracy of KRR in predicting nuclear binding energy and estimating SEMF coefficients. This work underscores the potential of machine learning in addressing longstanding challenges in nuclear physics, offering a pathway for enhanced theoretical and experimental alignment. Moreover, the incorporation of shell corrections into the traditional Liquid Drop Model (LDM) leads to the development of the Generalized Liquid Drop Model (GLDM). This enhanced approach provides a more comprehensive theoretical framework for understanding nuclear binding energy by accounting for proximity energy and shell effects. It offers improved alignment with experimental observations, effectively explaining phenomena such as magic numbers and nuclear stability trends.