Optimizing Machine Learning Code to Classify Neutron Resonances

Understanding the nuclear formation and decay processes of elements is crucial to unravelling the mysteries of our universe.

These processes depend on intrinsic properties of nuclei such as nuclear level densities, decay strength functions, and other

nuclear data. It is thus critical that a reliable nuclear database is produced. The focus of this project is to use current

measurements of resonance states observed in compound nuclei (formed by neutron-induced reactions) and develop a

machine learning algorithm to automate and assess this data and correctly classify neutron resonances according to their spin

groups. Synthetic data was used to train machine learning algorithms to classify resonances according to their spin groups,

considering their widths and spacings. Performance comparisons were run on scikit-learn classifiers (such as Random Forest,

Nearest Neighbors, Neural Network, etc.) to assess accuracies when varying hyper-parameters. Continued optimization

allows for application of transfer learning to predict spin assignments in real nuclei, as compiled in evaluated files or in the

Atlas of Neutron Resonances. Having an accurate nuclear database has many applications within astrophysics and nuclear

energy which promotes future discoveries in physics. This project was supported in part by the U.S. Department of Energy,

Office of Science, Office of Workforce Development for Teachers and Scientists (WDTS) under the Science Undergraduate

Laboratory Internships Program (SULI).