Engineering Commons^™

Application Of Artificial Neural Networks To Elemental Assay Data For Nuclear Forensics Analysis, Jason G. Seik

Theses and Dissertations

There is a need to quickly and accurately determine the likely physical origins of a collected sample for nuclear treaty verification purposes. The objective of this research is to prove there is a means to relate different samples (Q-values) to one another using a 'same versus not-same' artificial neural network called a Siamese network. This would provide the capability of comparing an unknown sample to a database of samples with known physical origins. Using moment transformations on current data has shown to increase the prediction capabilities of a Siamese network, and using a triplet loss function in connection with the …