Global geometry of chemical graph neural network representations in terms of chemical moieties
Open Access
- 15 February 2024
- journal article
- research article
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 3 (3) , 544-557
Abstract
The embedding vectors from a Graph Neural Network trained on quantum chemical data allow for a global geometric space with a Euclidean distance metric. Moieties that are close in chemical sense, are also close in Euclidean sense.Keywords
Funding Information
- New Brunswick Innovation Foundation
- Canada Foundation for Innovation (John Evans Leader Fund)
- Natural Sciences and Engineering Research Council of Canada (Discovery Grant)
- Canada Research Chairs
This publication has 76 references indexed in Scilit:
- Interpretable and Explainable Machine Learning for Materials Science and ChemistryAccounts of Materials Research, 2022
- Learning hidden chemistry with deep neural networksComputational Materials Science, 2021
- Transparency and the Black Box Problem: Why We Do Not Trust AIPhilosophy & Technology, 2021
- The message passing neural networks for chemical property prediction on SMILESMethods, 2020
- Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AIInformation Fusion, 2020
- Interpretable Discriminative Dimensionality Reduction and Feature Selection on the Manifold2020
- DScribe: Library of descriptors for machine learning in materials scienceComputer Physics Communications, 2020
- Linear Discriminant Analysis2013
- Principal component analysisWIREs Computational Statistics, 2010
- Representing high-dimensional potential-energy surfaces for reactions at surfaces by neural networksChemical Physics Letters, 2004