Recommended by

Indexed by

On the robust extrapolation of high-dimensional machine learning potentials

Claudio Zeni1*, Andrea Anelli2*, Aldo Glielmo3*, Kevin Rossi4*

1 International School for Advanced Studies, Via Bonomea, 265, 34136, Trieste, IT

2 Roche Pharma (Schweiz) AG Gartenstrasse 9, 4052 Basel, Switzerland

3 Banca d’Italia, Italy

4 Laboratory of Nanochemistry, Institute of Chemistry and Chemical Engineering, Ecole Polytechnique Fédérale de Lausanne, Lausanne, CH

* Corresponding authors emails: czeni@sissa.it, andrea.anelli@roche.com, Aldo.Glielmo@bancaditalia.it, kevin.rossi@epfl.ch
DOI10.24435/materialscloud:8w-a7 [version v1]

Publication date: Mar 03, 2022

How to cite this record

Claudio Zeni, Andrea Anelli, Aldo Glielmo, Kevin Rossi, On the robust extrapolation of high-dimensional machine learning potentials, Materials Cloud Archive 2022.34 (2022), doi: 10.24435/materialscloud:8w-a7.


We show that, contrary to popular assumptions, predictions from machine learning potentials built upon high-dimensional atom-density representations almost exclusively occur in regions of the representation space which lie outside the convex hull defined by the training set points. We then propose a perspective to rationalise the domain of robust extrapolation and accurate prediction of atomistic machine learning potentials in terms of the probability density induced by training points in the representation space. The data here contained can be used to reproduce all results and graphs shown in the article. We also include the trajectory files for the Au13 dataset we generate by running molecular dynamics simulations of an Au nanoparticle containing 13 atoms at temperatures of 50K, 100K, 200K, 300K, and 400K. Details regarding the generation of such dataset can be found in the supplementary information file for the article.

Materials Cloud sections using this data

No Explore or Discover sections associated with this archive record.


File name Size Description
1.0 GiB Full description is provided in the README.txt file


Files and data are licensed under the terms of the following license: Creative Commons Attribution 4.0 International.
Metadata, except for email addresses, are licensed under the Creative Commons Attribution Share-Alike 4.0 International license.


molecular dynamics simulation density-functional theory machine learning

Version history:

2022.34 (version v1) [This version] Mar 03, 2022 DOI10.24435/materialscloud:8w-a7