Using Machine Learning to Determine Proton Cross-Sections From Heavy-Ion Data

Abstract

In this article, a class of machine-learning techniques that implement model-averaging are used for the calculation of proton single-event upset (SEU) cross-sections from heavy-ion data. Calculations are checked using data on proton and heavy-ion cross-sections from the published literature. These techniques improve the accuracy of the calculation. Model averaging techniques were predicted the proton cross-section within an order of magnitude in better than 75% of the test cases. However, these techniques did little to improve the precision of the model predictions, and typically had an error bar of about 300%.