Taming nuclear complexity with a committee of multilayer neural networks

TL;DR

This paper introduces a committee of deep neural networks trained with active learning to accurately predict nuclear energies, achieving state-of-the-art accuracy with significantly reduced training data and enabling rapid, comprehensive nuclear structure studies.

Contribution

It presents a novel ensemble neural network approach combined with active learning to efficiently predict nuclear energies across the entire chart of nuclides.

Findings

Achieved accuracy comparable to nuclear energy density functionals

Reduced training set to only 210 nuclei

Enabled fast, large-scale nuclear structure predictions

Abstract

We demonstrate that a committee of deep neural networks is capable of predicting the ground-state and excited energies of more than 1800 atomic nuclei with an accuracy akin to the one achieved by state-of-the-art nuclear energy density functionals (EDFs) and a major speed-up. An active learning strategy is proposed to train this algorithm with a minimal set of 210 nuclei. This approach enables future fast studies of the influence of EDFs parametrizations on structure properties over the whole nuclear chart and suggests that for the first time an artificial intelligence successfully encoded the laws of nuclear deformation.