Nuclear $\beta$-decay half-lives are investigated by the neural network approach combined with a physically motivated $\beta$-decay half-life formula. This physics-information machine learning approach achieves better agreement with experimental half-lives than the sophisticated microscopic nuclear models and the empirical formulas. For the nuclei with half-lives less than 1 s, the neural network can describe the experimental half-lives about 1.62 times. It also effectively eliminates the nonphysical odd-even staggering of half-lives that is typically observed in traditional machine learning approaches and empirical formulas, leading to a more reliable description of nuclear $\beta$-decay half-lives. The extrapolation ability of the neural network is verified by newly measured half-lives as well.

Close

PrevNext

View figure in article

• Beta decay
• Lifetimes & widths
• Artificial neural networks
• Machine learning