Incompleteness of graph neural networks for points clouds in three dimensions

JSON Export

{
  "revision": 5, 
  "id": "1762", 
  "created": "2023-05-12T06:30:53.327580+00:00", 
  "metadata": {
    "doi": "10.24435/materialscloud:66-mm", 
    "status": "published", 
    "title": "Incompleteness of graph neural networks for points clouds in three dimensions", 
    "mcid": "2023.75", 
    "license_addendum": null, 
    "_files": [
      {
        "description": "Archive containing the counterexample structures in extended xyz format, together with python scripts to analyze them and explanations of the usage.", 
        "key": "gnn_counterexamples.zip", 
        "size": 8700514, 
        "checksum": "md5:c987bec26ce8c6055363dce34ff60bef"
      }
    ], 
    "owner": 108, 
    "_oai": {
      "id": "oai:materialscloud.org:1762"
    }, 
    "keywords": [
      "machine learning", 
      "graph neural networks", 
      "point clouds", 
      "MARVEL"
    ], 
    "conceptrecid": "1761", 
    "is_last": true, 
    "references": [
      {
        "type": "Journal reference", 
        "doi": "10.1088/2632-2153/aca1f8", 
        "url": "https://iopscience.iop.org/article/10.1088/2632-2153/aca1f8", 
        "comment": "Paper describing the construction of this class of counterexamples", 
        "citation": "S. N. Pozdnyakov and M. Ceriotti, Mach. Learn.: Sci. Technol. 3(4), 045020 (2022)."
      }
    ], 
    "publication_date": "May 12, 2023, 10:51:45", 
    "license": "Creative Commons Attribution Share Alike 4.0 International", 
    "id": "1762", 
    "description": "Graph neural networks are a popular deep-learning architecture in applications to materials and molecules, and the most widespread implementations rely on interatomic distances as geometric descriptors. Unfortunately, GNNs based on distances are not complete, i.e. there are geometries, corresponding to molecules and/or periodic structures, that are indistinguishable by the GNN. For these, the corresponding machine-learning models will be unable to learn differences in the properties of the \"degenerate\" structures. This dataset contains a collection of molecular and solid structures that cannot be discriminated by distance-based graph neural networks, together with example code showing how to parse them and use to demonstrate the shortcomings of this class of machine-learning algorithms.", 
    "version": 1, 
    "contributors": [
      {
        "email": "sergey.pozdnyakov@epfl.ch", 
        "affiliations": [
          "Institute of Materials, \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL), Lausanne, Switzerland"
        ], 
        "familyname": "Pozdnyakov", 
        "givennames": "Sergey"
      }, 
      {
        "email": "michele.ceriotti@gmail.com", 
        "affiliations": [
          "Institute of Materials, \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL), Lausanne, Switzerland"
        ], 
        "familyname": "Ceriotti", 
        "givennames": "Michele"
      }
    ], 
    "edited_by": 576
  }, 
  "updated": "2023-05-12T08:51:45.664843+00:00"
}