- By:
- Hatton, Conner J; Phathanapirom, Urairisa B; Dayman, Kenneth J
- Journal Name:
- Journal of Radioanalytical and Nuclear Chemistry
- Volume:
- TBD
- Publication Date:
- March 12, 2026
- Conference Name:
- International Conference on Methods and Applications of Radioanalytical Chemistry (MARC)
- Conference Location:
- Kailua-Kona, Hawaii, United States of America
- Conference Sponsor:
- American Nuclear Society
- View DOI Listing:
- https://doi.org/10.1007/s10967-025-10590-5
Abstract
Although neural networks offer cutting-edge predictive power, their deployment in high-consequence nuclear forensic applications is limited, partly because of their black-box nature. Incorporating robust uncertainty quantification methods into the predictive frameworks of neural networks is progress towards their future deployment in such scenarios. This work integrates uncertainty quantification into neural networks for nuclear reactor core-average burnup estimation from simulated environmental samples. We test two regimes (homogeneous and heterogeneous events) on DeepSets and Set Transformer architectures, we find both quantify predictive uncertainty effectively, but Set Transformer excels in partitioning latent events, offering superior predictive power and more informative uncertainty estimates.
