Multi-task learning on nuclear masses and separation energies with the kernel ridge regression

Abstract

A multi-task learning (MTL) framework, called gradient kernel ridge regression, for nuclear masses and separation energies is developed by introducing gradient kernel functions to the kernel ridge regression (KRR) approach. By taking the WS4 mass model as an example, the gradient KRR network is trained with the mass model residuals, i.e., deviations between experimental and theoretical values of masses and one-nucleon separation energies, to improve the accuracy of theoretical predictions. Significant improvements are achieved by the gradient KRR approach in both the interpolation and the extrapolation predictions of nuclear masses and separation energies. This demonstrates the advantage of the present MTL framework that integrates the information of nuclear masses and separation energies and improves the predictions for both of them.

3 pages, 3 figures

References in corpus (11)

• New parametrization for the nuclear covariant energy density functional with point-coupling interaction
• Surface diffuseness correction in global mass formula
• Nuclear mass predictions based on Bayesian neural network approach with pairing and shell effects
• Direct mapping of nuclear shell effects in the heaviest elements
• An artificial neural network application on nuclear charge radii
• Nuclear masses in extended kernel ridge regression with odd-even effects
• Beyond the proton drip line: Bayesian analysis of proton-emitting nuclei
• Improved radial basis function approach with the odd-even corrections
• Sensitivity study of \emph{r}-process abundances to nuclear masses
• The description of giant dipole resonance key parameters with multitask neural networks
• Machine-learning approach to finite-size effects in systems with strongly interacting fermions