Prediction of $\upalpha$-decay half-lives for superheavy nuclei based on neural network
Abstract
We trained a set of models based on the fully connected feed-forward neural network (FNN) to predict the $\upalpha$-decay half-lives of the superheavy nuclei (SHN). The model was trained with input data set of $Q$ values of $\upalpha$-decay, neutron and proton numbers ($N$, $Z$), etc. By comparing the predicted half-lives and the experimental values, the accuracy of the trained neural network can be enhanced by about 10%–20% than the general liquid drop model (GLDM2), and in the superheavy regions, the prediction of neural network model is in good agreement with the experimental values.
