Fission trajectory analysis using ML techniques

Abstract

This research analyzed trajectories of nuclear fission leading to symmetric or assymmetric mass division, obtained by a four-dimensional Langevin-model, using machine learning models. A hybrid neural network, combining Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), both of which were types of Recurrent Neural Networks (RNN), was utilized to classify whether each Langevin trajectory led to symmetric or asymmetric mass division. It was found that the current model could classify fate of these trajectories before reaching to the final destination (symmetric or assymmetric mode) with an accuracy of over 70%, clearly overestimating the asymmetric data.