Impact of quantum-chemical metrics on the machine learning prediction of electron density

TL;DR

This paper investigates how different quantum-chemical metrics affect machine learning predictions of electron density, demonstrating that the Coulomb metric improves accuracy, especially when combined with a model that enforces electron number conservation.

Contribution

It introduces a modified ML model incorporating electron number directly into the kernel, improving prediction accuracy over previous methods.

Findings

Coulomb metric yields better predictions for electrostatic potential and dipole moments.

Including electron number in the kernel reduces errors on test and out-of-sample sets.

The choice of metric significantly impacts the quality of ML-based electron density predictions.

Abstract

Machine learning (ML) algorithms have undergone an explosive development impacting every aspect of computational chemistry. To obtain reliable predictions, one needs to maintain the proper balance between the black-box nature of ML frameworks and the physics of the target properties. One of the most appealing quantum-chemical properties for regression models is the electron density, and some of us recently proposed a transferable and scalable model based on the decomposition of the density onto an atom-centered basis set. The decomposition, as well as the training of the model, is at its core a minimization of some loss function, which can be arbitrarily chosen and may lead to results of different quality. Well-studied in the context of density fitting (DF), the impact of the metric on the performance of ML models has not been analyzed yet. In this work, we compare predictions obtained using the overlap and the Coulomb-repulsion metrics for both decomposition and training. As expected, the Coulomb metric used as both the DF and ML loss functions leads to the best results for the electrostatic potential and dipole moments. The origin of this difference lies in the fact that the model is not constrained to predict densities that integrate to the exact number of electrons $N$. Since an \textit{a posteriori} correction for the number of electrons decreases the errors, we proposed a modification of the model where $N$ is included directly into the kernel function, which allowed to lower the errors on the test and out-of-sample sets.