A smooth basis for atomistic machine learning

Abstract

Machine learning frameworks based on correlations of interatomic positions begin with a discretized description of the density of other atoms in the neighbourhood of each atom in the system. Symmetry considerations support the use of spherical harmonics to expand the angular dependence of this density, but there is as yet no clear rationale to choose one radial basis over another. Here we investigate the basis that results from the solution of the Laplacian eigenvalue problem within a sphere around the atom of interest. We show that this generates the smoothest possible basis of a given size within the sphere, and that a tensor product of Laplacian eigenstates also provides the smoothest possible basis for expanding any higher-order correlation of the atomic density within the appropriate hypersphere. We consider several unsupervised metrics of the quality of a basis for a given dataset, and show that the Laplacian eigenstate basis has a performance that is much better than some widely used basis sets and is competitive with data-driven bases that numerically optimize each metric. In supervised machine learning tests, we find that the optimal function smoothness of the Laplacian eigenstates leads to comparable or better performance than can be obtained from a data-driven basis of a similar size that has been optimized to describe the atom-density correlation for the specific dataset. We conclude that the smoothness of the basis functions is a key and hitherto largely overlooked aspect of successful atomic density representations.

References in corpus (25)

• Gaussian Approximation Potentials: the accuracy of quantum mechanics, without the electrons
• On representing chemical environments
• E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
• Moment Tensor Potentials: a class of systematically improvable interatomic potentials
• Ab initio Random Structure Searching
• DScribe: Library of Descriptors for Machine Learning in Materials Science
• On-the-fly machine learning force field generation: Application to melting points
• Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point clouds
• Physics-inspired structural representations for molecules and materials
• Directional Message Passing for Molecular Graphs
• Automatic Selection of Atomic Fingerprints and Reference Configurations for Machine-Learning Potentials
• MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields
• Understanding molecular representations in machine learning: The role of uniqueness and target similarity
• Extending the Accuracy of the SNAP Interatomic Potential Form
• Feature Optimization for Atomistic Machine Learning Yields A Data-Driven Construction of the Periodic Table of the Elements
• Cormorant: Covariant Molecular Neural Networks
• Informing Geometric Deep Learning with Electronic Interactions to Accelerate Quantum Chemistry
• Covariant Compositional Networks For Learning Graphs
• Recursive evaluation and iterative contraction of $N$-body equivariant features
• An assessment of the structural resolution of various fingerprints commonly used in machine learning
• Efficient implementation of atom-density representations
• Sensitivity and Dimensionality of Atomic Environment Representations used for Machine Learning Interatomic Potentials
• Manifolds of quasi-constant SOAP and ACSF fingerprints and the resulting failure to machine learn four-body interactions
• Optimal radial basis for density-based atomic representations
• Comment on "Manifolds of quasi-constant SOAP and ACSF fingerprints and the resulting failure to machine learn four body interactions"

Cited by in corpus (3)

• Completeness of Atomic Structure Representations
• Prediction rigidities for data-driven chemistry
• Machine learning of electronic structure and atomistic properties from the external potential