Precise neural network predictions of energies and radii from the no-core shell model
Tobias Wolfgruber, Marco Knöll, Robert Roth
Abstract
$A\phantom{\rule{0}{0ex}}b$-$i\phantom{\rule{0}{0ex}}n\phantom{\rule{0}{0ex}}i\phantom{\rule{0}{0ex}}t\phantom{\rule{0}{0ex}}i\phantom{\rule{0}{0ex}}o$ calculations of atomic nuclei have revolutionized nuclear structure physics. Yet challenges remain, not least the reliable calculation of nuclear radii. In a concurrent development, modern machine-learning algorithms have excelled in a variety of computational tasks such as pattern recognition and interpolation. The authors have applied artificial neural networks (ANNs) to the extrapolation of no-core shell model calculations to infinite model spaces, effectively circumventing their computational limitations. In particular, the results show that min-max normalization, a common technique in machine learning, leads to the best results for radii. These advances offer hope that the ANN architecture is capable of handling other observables such as electromagnetic moments and transition strengths.