How Powerful are Graph Neural Networks?

Benchmark Model Rank Results
graph-classification-on-cifar10-100kGIN#19Accuracy (%): 53.28
graph-classification-on-collabGIN-0#12Accuracy: 80.2%
graph-classification-on-ddGIN#33Accuracy: 77.311±2.223
graph-classification-on-enzymesGIN#16Accuracy: 68.303±4.170
graph-classification-on-imdb-bGIN#5Accuracy: 81.250±3.775
graph-classification-on-imdb-bGIN-0#22Accuracy: 75.1%
graph-classification-on-imdb-mGIN-0#10Accuracy: 52.3%
graph-classification-on-mutagGIN-0#24Accuracy: 89.4%
graph-classification-on-nci1GIN#15Accuracy: 84.818±0.936
graph-classification-on-nci1GIN-0#26Accuracy: 82.7%
graph-classification-on-nci109GIN#5Accuracy: 84.155±0.812
graph-classification-on-peptides-funcGIN#37AP: 0.6043±0.0216
graph-classification-on-proteinsGIN-0#46Accuracy: 76.2
graph-classification-on-proteinsGIN#55Accuracy: 75.536±1.851
graph-classification-on-ptcGIN-0#18Accuracy: 64.40%
graph-classification-on-re-m5kGIN-0#1Accuracy: 57.5%
graph-classification-on-reddit-bGIN-0#3Accuracy: 92.4
graph-property-prediction-on-ogbg-code2GIN+virtual node#9Test F1 score: 0.1581 ± 0.0026Ext. data: No
graph-property-prediction-on-ogbg-code2GIN#15Test F1 score: 0.1495 ± 0.0023Ext. data: No
graph-property-prediction-on-ogbg-molhivGIN+virtual node#29Test ROC-AUC: 0.7707 ± 0.0149Ext. data: No
graph-property-prediction-on-ogbg-molhivGIN#34Test ROC-AUC: 0.7558 ± 0.0140Ext. data: No
graph-property-prediction-on-ogbg-molpcbaGIN+virtual node#22Test AP: 0.2703 ± 0.0023Ext. data: NoValidation AP: 0.2798 ± 0.0025
graph-property-prediction-on-ogbg-molpcbaGIN#27Test AP: 0.2266 ± 0.0028Ext. data: NoValidation AP: 0.2305 ± 0.0027
graph-property-prediction-on-ogbg-ppaGIN+virtual node#9Ext. data: NoTest Accuracy: 0.7037 ± 0.0107
graph-property-prediction-on-ogbg-ppaGIN#12Ext. data: NoTest Accuracy: 0.6892 ± 0.0100
graph-regression-on-esr2GIN#4R2: 0.668±0.000RMSE: 0.509±0.668
graph-regression-on-f2GIN#3R2: 0.887±0.000RMSE: 0.342±0.887
graph-regression-on-kitGIN#5R2: 0.833±0.000RMSE: 0.444±0.833
graph-regression-on-lipophilicityGIN#4RMSE: 0.537±0.010R2: 0.819±0.007
graph-regression-on-parp1GIN#3R2: 0.922±0.000RMSE: 0.349±0.922
graph-regression-on-pcqm4mv2-lscGIN#18Validation MAE: 0.1195Test MAE: 0.1218
graph-regression-on-pgrGIN#4R2: 0.696±0.000RMSE: 0.532±0.696
graph-regression-on-zinc-500kGIN#34MAE: 0.526
graph-regression-on-zinc-fullGIN#15Test MAE: 0.068±0.004
molecular-property-prediction-on-esolGIN#3RMSE: 0.509±0.044R2: 0.938±0.011
molecular-property-prediction-on-freesolvGIN#4RMSE: 0.744±0.083R2: 0.964±0.008
node-classification-on-pattern-100kGIN#5Accuracy (%): 85.590