Updates to the DScribe library: New descriptors and derivatives

Abstract

We present an update of the DScribe package, a Python library for atomistic descriptors. The update extends DScribe’s descriptor selection with the Valle–Oganov materials fingerprint and provides descriptor derivatives to enable more advanced machine learning tasks, such as force prediction and structure optimization. For all descriptors, numeric derivatives are now available in DScribe. For the many-body tensor representation (MBTR) and the Smooth Overlap of Atomic Positions (SOAP), we have also implemented analytic derivatives. We demonstrate the effectiveness of the descriptor derivatives for machine learning models of Cu clusters and perovskite alloys.

The challenge of balancing model sensitivity and robustness in predicting yields: a benchmarking study of amide coupling reactions

“…AEV has been shown to be an effective method for encoding conformers in several applications. 22,29 The smooth overlap of atomic positions (SOAP) descriptor 30 was also used in comparison with the AEV descriptor. Both descriptors describe information about the spatial environment of each atom.…”

Predicting the rates of photocatalytic hydrogen evolution over cocatalyst-deposited TiO2 using machine learning with active photon flux as a unifying feature

“…This practice was taken from well-known molecular structure representation approaches including, Sine matrix, Smooth Overlap of Atomic Positions (SOAP), Many-body Tensor Representation (MBTR), and Local Many-body Tensor Representation (LMBTR), as provided in the DScribe package. 30,31…”

“…This practice was taken from well-known molecular structure representation approaches including, Sine matrix, Smooth Overlap of Atomic Positions (SOAP), Many-body Tensor Representation (MBTR), and Local Many-body Tensor Representation (LMBTR), as provided in the DScribe package. 30,31 Being the driver of photocatalytic reactions, the characteristic of irradiation is an indispensable input in constructing a sensible model. This has been by far the most restrictive feature during the data mining process, where the bulk of published data would report the type of the light sources and their electrical power ratings, but without or with ambiguous information on the intensities irradiating the reaction volume.…”

From Local Atomic Environments to Molecular Information Entropy

“…In the next step, we use the ground state geometries of the 13 molecules mentioned above from the QM9 data set , and calculate the SOAP descriptors for each atom using dscribe , (with r cut = 6 Å, n max = 10, and l max = 6). The similarity matrix and entropy are then calculated via eqs and, respectively, for a given sensitivity exponent ζ.…”