神经网络方法对原子核质量的改进研究

Neural network approach to improve the quality of atomic nuclei

原子核的质量对核物理的各个方面以及对物理的其他分支,特别是弱相互作用研究和天体物理学来说是非常重要的.但目前而言,所有描述原子核质量的模型都或多或少地具有某些局限性,需要进行修正.而神经网络方法具有强大的学习能力,通过使用神经网络方法,可以对计算原子核质量的公式进行修正,从而得到更加精确的结果.本文的主要目的是描述几种修正原子核质量公式的神经网络方法,并将其对质量公式的改进结果进行比较.

The mass of the nucleus is very important to all aspects of nuclear physics and other branches of physics,particularly weak interaction studies,and astrophysics.However,all the mass models characterizing the nucleus have more or less certain limitations that need to be revised at present.The neural network approach offers a high learning capacity.By using the neural network method,the formula for calculating atomic mass can be modified to obtain more accurate results.We examine the capacity of multiple neural network methods to reproduce experimental data in this paper,beginning with their computation of nuclear masses and focusing on their extrapolation skills.The findings reveal that multiple neural network methods may increase the accuracy of various mass models greatly,while the extrapolation ability of different methods varies depending on the nuclear mass model utilized.