To an increasing extent, theoretical nuclear physics involves statistical inference on computationally-demanding theoretical models that often combine heterogeneous datasets. Advanced statistical approaches can enhance the quality of nuclear modeling in many ways. First, the statistical tools of uncertainty quantification can be used to estimate theoretical errors on computed observables. Second, they can help to assess the information content of measured observables with respect to theoretical models and assess the information content of present-day theoretical models with respect to measured observables. Importantly, they can be used to understand a model's structure through parameter estimation and model reduction. Finally, statistical tools can improve predictive capability and optimize knowledge extraction by extrapolating beyond the regions reached by experiments to provide meaningful input to applications and planned measurements.
In this presentation, after presenting a brief summary of machine learning applications to low-energy nuclear theory, I will employ Bayesian machine learning tools to assess the predictive power of global mass models towards more unstable nuclei and provide uncertainty quantification of predictions. The proposed robust statistical approach to extrapolations can be useful for assessing the impact of current and future experiments in the context of model developments.
Zoom link will be available via announcement email, or by contacting: stroberg[at]uw.edu.