Skip to Main Content U.S. Department of Energy
Deep Science Agile Initiative

News and Highlights

Training neural networks to classify low-background data.

Training neural networks to classify low-background data

Emily Mace presented "Use of Neural Networks to Analyze Pulse Shape Data in Low-Background Detectors" at the Methods & Applications of Radioanalytical Chemistry (MARC) conference. MARC, which is held roughly every three years, focuses on emerging developments in radiochemistry and radioanalytical chemistry, and promoting information exchange within that community.

At the conference, Emily discussed analyzing data that PNNL has collected years using ultra-low-background proportional counters in a shallow underground laboratory. This large dataset of events was exploited to study the impact of using neural networks for data analysis compared to a simple pulse-shape discrimination (PSD) method. The team found that the PSD method introduces false positives and false negatives for certain datasets; however, the neural network is able to correctly classify these events. The presentation described the method used for training the neutral network, the analysis of challenge datasets, and a comparison between the standard approach and a neural network.
July 2018

Deep Learning