Frameworks: Bayesian Analysis for Nuclei in Diverse Theories

Project: Research project

Project Details


The United States’ investment in nuclear physics has yielded tremendous gains in our
understanding of matter as well as many applications that benefit society. Progress in the theory of
nuclei and nuclear matter has produced a multitude of models that describe extant data well. But
a paradigm shift in the analysis of nuclear-physics data is now necessary: predictions and quantified
uncertainties must use the collective wisdom of the best models, constrained by data, and include
a unified treatment of all uncertainties. This will enable reliable predictions for experimentally
inaccessible environments, e.g. the properties and dynamics of matter at the core of neutron stars
or in the first microsecond after the Big Bang. And it will make possible quantitative evaluation of
the impact of new experiments, so facilitating optimal use of taxpayers’ investment in this science.
Bayesian Analysis for Nuclei in Diverse Theories (BANDIT) will address these uncertainty
quantification needs by creating and supporting an extendible software base that is augmented
by human advice. Our Cyberinfrastructure (CI) Framework is Bayesian, and so leverages deep
scientific prior knowledge with experimental data for powerful statistical inference. BANDIT will
provide extrapolations and uncertainty quantification that exploit multiple theoretical models. It
will be built through intensive collaboration between statisticians, computer scientists and experts
in nuclear theory. The suite of codes will be useful for a broad community of nuclear physicists,
and coupled with support that helps users reliably incorporate their expert knowledge to use the
framework intelligently and effectively. Both codes and support will be sustainable.
Effective start/end date7/1/206/30/25


  • Ohio State University (UT21727//OAC-2004601)
  • National Science Foundation (UT21727//OAC-2004601)


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.