Prediction of α-decay half-lives for superheavy nuclei based on neural network

We trained a set of models based on the fully connected feed-forward neural network (FNN) to predict the ff-decay halflives of the superheavy nuclei (SHN). The model was trained with input data set of Q values of ff-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.