Efficient implementation of atom-density representations

Abstract

Physically motivated and mathematically robust atom-centered representations of molecular structures are key to the success of modern atomistic machine learning. They lie at the foundation of a wide range of methods to predict the properties of both materials and molecules and to explore and visualize their chemical structures and compositions. Recently, it has become clear that many of the most effective representations share a fundamental formal connection. They can all be expressed as a discretization of n-body correlation functions of the local atom density, suggesting the opportunity of standardizing and, more importantly, optimizing their evaluation. We present an implementation, named librascal, whose modular design lends itself both to developing refinements to the density-based formalism and to rapid prototyping for new developments of rotationally equivariant atomistic representations. As an example, we discuss smooth overlap of atomic position (SOAP) features, perhaps the most widely used member of this family of representations, to show how the expansion of the local density can be optimized for any choice of radial basis sets. We discuss the representation in the context of a kernel ridge regression model, commonly used with SOAP features, and analyze how the computational effort scales for each of the individual steps of the calculation. By applying data reduction techniques in feature space, we show how to reduce the total computational cost by a factor of up to 4 without affecting the model's symmetry properties and without significantly impacting its accuracy.

Classifying soft self-assembled materials via unsupervised machine learning of defects

“…smoothness, additivity and level of locality. 60,61 SOAP descriptors satisfy all these requirements.…”

“…43 as new bond-order parameters, able to efficiently include radial and angular information of the environment that surrounds atoms or molecules. 36,43,60 The SOAP descriptor of a system of N particles describes the atomic/molecular surroundings of a selected set of M coordinates of the system components, which are referred to as the "centers" of the SOAP vector. These M centers can include the position of every single atom of the system, as well as selections or combinations (as e.g.…”

“…(ii) rotational invariance is enforced by building symmetrized combinations of the expansion coefficients. 43,60 We here employ the second-order SOAP descriptor, also called SOAP power-spectrum (in analogy with the Fourier analysis), which can be written as…”

Unsupervised Learning Methods for Molecular Simulation Data

“…PLUMED ( 25 ) allows computing a very large number of collective variables, and even defining ad hoc functions. The three packages DScribe , 63 PANNA, 391 and Librascal ( 392 ) can be used to compute all the most widely used numerical features for condensed matter systems.…”

A Composition-Transferable Machine Learning Potential for LiCl-KCl Molten Salts Validated by HEXRD

“…Machine learning (ML) has enabled [15][16][17][18] a new generation of low-cost interatomic potentials (IP) that provide access to quantum mechanically accurate manybody potential energy surfaces for condensed phases [19][20][21][22][23][24][25][26][27][28]. These ML-IP can drive mesoscopic simulations of atomic processes with ab initio accuracy, bypassing the length scale limitations imposed by traditional ab initio methods.…”