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.

References

The generalization capability of the proposed method to other nuclides remains to be systematically verified using sufficient experimental datasets.

— Multi-task deep neural network for predicting both nuclear fission yields and their experimental errors in peak-shaped data  (2603.29100 - Ueno et al., 31 Mar 2026) in Section SUMMARY