Abstract: Machine learning has the potential to improve the speed and reliability of radioactive material identification. However, acquiring sufficient experimental training data can be expensive, and models trained on simulated data often generalize poorly to real-world conditions. This research addresses this challenge using domain adaptation – a technique that enables models trained in one domain (simulations) to perform effectively in a new domain (real-world measurements). This approach helps safeguard national security by enabling radiation detection models that are both efficient to train and reliable in the field.

Bio: Peter Lalor is a Linus Pauling Postdoctoral Research Fellow in the Applied Radiation & Detection group at Pacific Northwest National Laboratory. His research focuses on developing machine learning methods for material identification and detection. Peter received his PhD in Computational Nuclear Science and Engineering from MIT in 2024 as a Computational Science Graduate Fellow. His dissertation work focused on studying computational methods for predicting the material composition of X-ray scans of cargo containers.