Print Email Facebook Twitter Physical pooling functions in graph neural networks for molecular property prediction Title Physical pooling functions in graph neural networks for molecular property prediction Author Schweidtmann, A.M. (TU Delft ChemE/Product and Process Engineering; Rheinisch-Westfälische Technische Hochschule) Rittig, J. (Rheinisch-Westfälische Technische Hochschule) Weber, J.M. (TU Delft Pattern Recognition and Bioinformatics) Grohe, Martin (Rheinisch-Westfälische Technische Hochschule) Dahmen, Manuel (Forschungszentrum Jülich GmbH) Leonhard, Kai (Rheinisch-Westfälische Technische Hochschule) Mitsos, Alexander (Rheinisch-Westfälische Technische Hochschule; Forschungszentrum Jülich GmbH; JARA Center for Simulation and Data Science (CSD)) Date 2023 Abstract Graph neural networks (GNNs) are emerging in chemical engineering for the end-to-end learning of physicochemical properties based on molecular graphs. A key element of GNNs is the pooling function which combines atom feature vectors into molecular fingerprints. Most previous works use a standard pooling function to predict a variety of properties. However, unsuitable pooling functions can lead to unphysical GNNs that poorly generalize. We compare and select meaningful GNN pooling methods based on physical knowledge about the learned properties. The impact of physical pooling functions is demonstrated with molecular properties calculated from quantum mechanical computations. We also compare our results to the recent set2set pooling approach. We recommend using sum pooling for the prediction of properties that depend on molecular size and compare pooling functions for properties that are molecular size-independent. Overall, we show that the use of physical pooling functions significantly enhances generalization. Subject Graph convolutional neural networksPhysics-informed machine learningPooling functionProperty prediction To reference this document use: http://resolver.tudelft.nl/uuid:9081cd83-4687-49f1-bab2-5ec8c1665060 DOI https://doi.org/10.1016/j.compchemeng.2023.108202 Embargo date 2023-08-28 ISSN 0098-1354 Source Computers & Chemical Engineering, 172 Bibliographical note Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public. Part of collection Institutional Repository Document type journal article Rights © 2023 A.M. Schweidtmann, J. Rittig, J.M. Weber, Martin Grohe, Manuel Dahmen, Kai Leonhard, Alexander Mitsos Files PDF 1_s2.0_S0098135423000716_main.pdf 784 KB Close viewer /islandora/object/uuid:9081cd83-4687-49f1-bab2-5ec8c1665060/datastream/OBJ/view