Nuclear Mass Predictions Using a Neural Network with Additive Gaussian Process Regression-Optimized Activation Functions

Abstract

Nuclear masses are machine-learned as a function of proton and neutron numbers. The neural network with additive Gaussian process regression-optimized activation functions (GPR-NN) method is employed for the first time for this purpose. GPR-NN combines the advantages of both neural networks and Gaussian process regression, in that it possesses the expressive power of an NN, in principle allowing modeling any kind of dependence of nuclear mass on the features, and robustness of a linear regression with respect to overfitting. A study of the GPR-NN approach for interpolation and extrapolation in nuclear mass predictions is presented. It is found that the optimal hyperparameters for the GPR-NN approach in interpolation and extrapolation are different. If an appropriate set of hyperparameters is adopted, the GPR-NN approach can achieve good extrapolation performance for nuclear mass prediction, which could potentially help improve the mass predictions of a large number of currently experimentally unknown nuclei.