Physically Interpretable Machine Learning for nuclear masses

TL;DR

This paper introduces a physics-constrained neural network model for predicting atomic nuclear masses, achieving high accuracy and interpretability by integrating physical knowledge into machine learning.

Contribution

The authors develop a novel physically interpretable machine learning approach that incorporates physics-based features and constraints for nuclear mass prediction.

Findings

Achieved RMS error of ~186 keV on training data

Predicted nuclear masses with RMS error of ~316 keV on test set

Model interpretability via feature importance analysis

Abstract

We present a novel approach to modeling the ground state mass of atomic nuclei based directly on a probabilistic neural network constrained by relevant physics. Our Physically Interpretable Machine Learning (PIML) approach incorporates knowledge of physics by using a physically motivated feature space in addition to a soft physics constraint that is implemented as a penalty to the loss function. We train our PIML model on a random set of $\sim$20\% of the Atomic Mass Evaluation (AME) and predict the remaining $\sim$80\%. The success of our methodology is exhibited by the unprecedented $\sigma_\textrm{RMS}\sim186$ keV match to data for the training set and $\sigma_\textrm{RMS}\sim316$ keV for the entire AME with $Z \geq 20$. We show that our general methodology can be interpreted using feature importance.