Bayesian approach to model-based extrapolation of nuclear observables

The Bayesian statistical approaches have a recognized utility in improving the quantified predictions for nuclear masses away from stability that provide key inputs for a variety of astrophysical applications. The present paper is devoted to the methodology of such efforts. It applies Bayesian Gaussian processes and neural networks to two-neutron separation energies of nuclei. The authors find that Gaussian processes deliver a more stable performance.