Verify generalization to nuclides beyond major actinides

Establish the generalization capability of the Multi-gate Mixture-of-Experts multi-task deep neural network that simultaneously predicts fission product yields and their evaluated errors, beyond the validated cases of U-235, U-238, Pu-239, and Pu-241, by systematically validating its predictions for other nuclides against sufficient experimental datasets.

Background

The paper proposes a multi-task deep neural network based on the Multi-gate Mixture-of-Experts architecture to simultaneously predict fission product yields (FPY) and their associated evaluated uncertainties (FPY errors). The method is shown to improve peak-structure prediction and error estimation compared to baseline approaches.

However, the validation presented in the study is limited to major actinides (U-235, U-238, Pu-239, Pu-241) where reliable data exist. The authors explicitly note that the generalization of this approach to other nuclides has not yet been systematically verified and requires sufficient experimental datasets for evaluation.