36th Annual Student Research Conference
Accuracy Correlation in Neutron Resonance Reclassification
Ian Q. Snider
Dr. Rasanjali Jayathissa and Dr. Gustavo Nobre (Brookhaven National Laboratory), Faculty Mentors
Current methods for determining the resonance quantum numbers associated with angular momenta and spin are difficult, time consuming, and may not be fully reproducible, often leading to incorrect spin assignments. To solve this problem, we have employed a machine learning (ML) based method to train an algorithm for identifying and reclassifying incorrect neutron spin assignments. We build synthetic data that mimics the statistical properties of real resonances to train the algorithm. Then, we validate the trained algorithm with a set of real In-115 polarized data. The goal of this project is to analyze the correlation between the training and validation accuracies. We can then attempt to improve the validation accuracy by adjusting the ML classifier’s hyperparameters. We also explored an iterative method in which, under certain conditions, successive reclassifications could incrementally improve the quality of any misclassified resonance sequence.
Keywords: Nuclear Physics, Nuclear Data Science, Neutron Resonance, Machine Learning, Angular Momenta, Spin, Brookhaven National Laboratory
Topic(s):Physics
Presentation Type: Oral Presentation
Session: 110-4
Location: MG 1096
Time: 9:15