Principal Neighbourhood Aggregation for Graph Nets

Benchmark Model Rank Results
graph-classification-on-cifar10-100kPNA#12Accuracy (%): 70.47
graph-classification-on-ddPNA#19Accuracy: 78.992±4.407
graph-classification-on-enzymesPNA#9Accuracy: 73.021±2.512
graph-classification-on-imdb-bPNA#12Accuracy: 78.000±3.808
graph-classification-on-nci1PNA#13Accuracy: 84.964±1.391
graph-classification-on-nci109PNA#10Accuracy: 83.382±1.045
graph-classification-on-proteinsPNA#23Accuracy: 77.679±3.281
graph-property-prediction-on-ogbg-molhivPNA#18Test ROC-AUC: 0.7905 ± 0.0132Ext. data: No
graph-property-prediction-on-ogbg-molpcbaPNA#18Test AP: 0.2838 ± 0.0035Ext. data: NoValidation AP: 0.2926 ± 0.0026
graph-regression-on-esr2PNA#2R2: 0.696±0.000RMSE: 0.486±0.696
graph-regression-on-f2PNA#1R2: 0.891±0.000RMSE: 0.336±0.891
graph-regression-on-kitPNA#1R2: 0.843±0.000RMSE: 0.430±0.843
graph-regression-on-lipophilicityPNA#1RMSE: 0.520±0.011R2: 0.830±0.007
graph-regression-on-parp1PNA#2R2: 0.924±0.000RMSE: 0.346±0.924
graph-regression-on-pgrPNA#2R2: 0.717±0.000RMSE: 0.514±0.717
graph-regression-on-zincPNA#20MAE: 0.142
graph-regression-on-zinc-fullPNA#14Test MAE: 0.057±0.007
molecular-property-prediction-on-esolPNA#2RMSE: 0.493±0.026R2: 0.942±0.006
molecular-property-prediction-on-freesolvPNA#7RMSE: 0.870±0.081R2: 0.951±0.009
node-classification-on-pattern-100kPNA#4Accuracy (%): 86.567