×

Recommended by

Indexed by

Sensitivity benchmarks of structural representations for atomic-scale machine learning

Sergey Pozdnyakov1*, Michele Ceriotti1*

1 Laboratory of Computational Science and Modelling, Institute of Materials, Ecole Polytechnique Federale de Lausanne, Lausanne 1015, Switzerland

* Corresponding authors emails: sergey.pozdnyakov@epfl.ch, michele.ceriotti@epfl.ch
DOI10.24435/materialscloud:7z-g6 [version v1]

Publication date: Sep 17, 2021

How to cite this record

Sergey Pozdnyakov, Michele Ceriotti, Sensitivity benchmarks of structural representations for atomic-scale machine learning, Materials Cloud Archive 2021.149 (2021), https://doi.org/10.24435/materialscloud:7z-g6

Description

This dataset contains three sets of CH4 geometries that are distorted along special directions, to reveal the sensitivity to atomic displacements of structural descriptors used in machine-learning applications. The structures are stored in a format that can be visualized on http://chemiscope.org, and contain also DFT-computed energies, as well as the sensitivity analysis of four different kinds of features.

Materials Cloud sections using this data

No Explore or Discover sections associated with this archive record.

Files

File name Size Description
asymmetric_ch4.chemiscope.json.gz
MD5md5:92e76339ffd8b20ef12d634475910a4e
Visualize on Chemiscope 		62.5 KiB 2d manifold around random asymmetric structure
symmetric_ch4.chemiscope.json.gz
MD5md5:df08fd38bd82b76ab2349ffe78f70c9a
Visualize on Chemiscope 		41.3 KiB 2d manifold around ground state
degenerate_ch4.chemiscope.json.gz
MD5md5:d5bfbe6afd3133945498006f1830cb00
Visualize on Chemiscope 		58.1 KiB 2d manifold around degenerate structure
readme.txt
MD5md5:42ac35280baa1bac75249d65577e966b
1.5 KiB readme

License

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

External references

Preprint
Sergey N. Pozdnyakov, Liwei Zhang, Christoph Ortner, Gabor Csanyi, and Michele Ceriotti, Open Research Europe (submission in preparation)

Keywords

DFT methane sensitivity representations machine learning MARVEL SNSF ERC

Version history:

2021.149 (version v1) [This version] Sep 17, 2021 DOI10.24435/materialscloud:7z-g6