Machine learning approach to pattern recognition in nuclear dynamics from the <i>ab initio</i> symmetry-adapted no-core shell model

A novel machine learning approach is used to provide further insight into atomic nuclei and detect orderly patterns amidst a vast data of large-scale calculations. The method utilizes neural network that trained on ab initio results from the symmetry-adapted no-core shell model (SA-NCSM) for light nuclei. We show SA-NCSM, which expands applications up medium-mass by using dominant symmetries nuclear dynamics, can reach heavier when coupled with approach. In particular, we find probability amplitudes $s$-and $p$-shell wave functions not only predicts configurations but in addition, tested $^{20}$Ne ground state, it accurately reproduces distribution. nonnegligible predicted an important input SA-NCSM reducing ultra-large spaces manageable sizes be, turn, utilized calculations obtain accurate observables. capable describing deformation track shape evolution along $^{20-42}$Mg isotopic chain, suggesting shape-coexistence more pronounced toward very neutron-rich isotopes. first descriptions structure $^{24}$Si $^{40}$Mg interest x-ray burst nucleosynthesis, even extremely heavy such as $^{166,168}$Er $^{236}$U, build upon principles considerations.