000917552 001__ 917552
000917552 005__ 20240712112854.0
000917552 0247_ $$2doi$$a10.48550/ARXIV.2207.13779
000917552 0247_ $$2Handle$$a2128/33643
000917552 037__ $$aFZJ-2023-00754
000917552 1001_ $$0P:(DE-HGF)0$$aSchweidtmann, Artur M.$$b0
000917552 245__ $$aPhysical Pooling Functions in Graph Neural Networks for Molecular Property Prediction
000917552 260__ $$barXiv$$c2022
000917552 3367_ $$0PUB:(DE-HGF)25$$2PUB:(DE-HGF)$$aPreprint$$bpreprint$$mpreprint$$s1673946218_28073
000917552 3367_ $$2ORCID$$aWORKING_PAPER
000917552 3367_ $$028$$2EndNote$$aElectronic Article
000917552 3367_ $$2DRIVER$$apreprint
000917552 3367_ $$2BibTeX$$aARTICLE
000917552 3367_ $$2DataCite$$aOutput Types/Working Paper
000917552 520__ $$aGraph neural networks (GNNs) are emerging in chemical engineering for the end-to-end learning of physicochemical properties based on molecular graphs. A key element of GNNs is the pooling function which combines atom feature vectors into molecular fingerprints. Most previous works use a standard pooling function to predict a variety of properties. However, unsuitable pooling functions can lead to unphysical GNNs that poorly generalize. We compare and select meaningful GNN pooling methods based on physical knowledge about the learned properties. The impact of physical pooling functions is demonstrated with molecular properties calculated from quantum mechanical computations. We also compare our results to the recent set2set pooling approach. We recommend using sum pooling for the prediction of properties that depend on molecular size and compare pooling functions for properties that are molecular size-independent. Overall, we show that the use of physical pooling functions significantly enhances generalization.
000917552 536__ $$0G:(DE-HGF)POF4-1121$$a1121 - Digitalization and Systems Technology for Flexibility Solutions (POF4-112)$$cPOF4-112$$fPOF IV$$x0
000917552 588__ $$aDataset connected to DataCite
000917552 650_7 $$2Other$$aMachine Learning (cs.LG)
000917552 650_7 $$2Other$$aFOS: Computer and information sciences
000917552 7001_ $$0P:(DE-HGF)0$$aRittig, Jan G.$$b1
000917552 7001_ $$0P:(DE-HGF)0$$aWeber, Jana M.$$b2
000917552 7001_ $$0P:(DE-HGF)0$$aGrohe, Martin$$b3
000917552 7001_ $$0P:(DE-Juel1)172097$$aDahmen, Manuel$$b4$$ufzj
000917552 7001_ $$0P:(DE-HGF)0$$aLeonhard, Kai$$b5
000917552 7001_ $$0P:(DE-Juel1)172025$$aMitsos, Alexander$$b6$$eCorresponding author$$ufzj
000917552 773__ $$a10.48550/ARXIV.2207.13779
000917552 8564_ $$uhttps://juser.fz-juelich.de/record/917552/files/2207.13779.pdf$$yOpenAccess
000917552 909CO $$ooai:juser.fz-juelich.de:917552$$pdnbdelivery$$pdriver$$pVDB$$popen_access$$popenaire
000917552 9101_ $$0I:(DE-588b)36225-6$$6P:(DE-HGF)0$$aRWTH Aachen$$b1$$kRWTH
000917552 9101_ $$0I:(DE-588b)5008462-8$$6P:(DE-Juel1)172097$$aForschungszentrum Jülich$$b4$$kFZJ
000917552 9101_ $$0I:(DE-588b)36225-6$$6P:(DE-HGF)0$$aRWTH Aachen$$b5$$kRWTH
000917552 9101_ $$0I:(DE-588b)5008462-8$$6P:(DE-Juel1)172025$$aForschungszentrum Jülich$$b6$$kFZJ
000917552 9101_ $$0I:(DE-588b)36225-6$$6P:(DE-Juel1)172025$$aRWTH Aachen$$b6$$kRWTH
000917552 9131_ $$0G:(DE-HGF)POF4-112$$1G:(DE-HGF)POF4-110$$2G:(DE-HGF)POF4-100$$3G:(DE-HGF)POF4$$4G:(DE-HGF)POF$$9G:(DE-HGF)POF4-1121$$aDE-HGF$$bForschungsbereich Energie$$lEnergiesystemdesign (ESD)$$vDigitalisierung und Systemtechnik$$x0
000917552 9141_ $$y2022
000917552 915__ $$0StatID:(DE-HGF)0510$$2StatID$$aOpenAccess
000917552 920__ $$lyes
000917552 9201_ $$0I:(DE-Juel1)IEK-10-20170217$$kIEK-10$$lModellierung von Energiesystemen$$x0
000917552 9801_ $$aFullTexts
000917552 980__ $$apreprint
000917552 980__ $$aVDB
000917552 980__ $$aUNRESTRICTED
000917552 980__ $$aI:(DE-Juel1)IEK-10-20170217
000917552 981__ $$aI:(DE-Juel1)ICE-1-20170217