Recipe for a General, Powerful, Scalable Graph Transformer

Benchmark Model Rank Results
graph-classification-on-cifar10-100kGPS#11Accuracy (%): 72.298
graph-classification-on-enzymesGraphGPS#2Accuracy: 78.667±4.625
graph-classification-on-imdb-bGraphGPS#9Accuracy: 79.250±3.096
graph-classification-on-malnet-tinyGPS#4Accuracy: 93.36 ± 0.6
graph-classification-on-mnistGPS#11Accuracy: 98.05
graph-classification-on-nci1GraphGPS#11Accuracy: 85.110±1.423
graph-classification-on-nci109GraphGPS#17Accuracy: 81.256±0.501
graph-classification-on-peptides-funcGPS#29AP: 0.6535±0.0041
graph-classification-on-proteinsGraphGPS#29Accuracy: 77.143±1.494
graph-property-prediction-on-ogbg-code2GPS#5Test F1 score: 0.1894Ext. data: NoValidation F1 score: 0.1739 ± 0.001
graph-property-prediction-on-ogbg-molhivGPS#21Test ROC-AUC: 0.7880Ext. data: NoValidation ROC-AUC: 0.8255 ± 0.0092
graph-property-prediction-on-ogbg-molpcbaGPS#15Test AP: 0.2907Ext. data: NoValidation AP: 0.3015 ± 0.0038
graph-property-prediction-on-ogbg-ppaGPS#4Ext. data: NoTest Accuracy: 0.8015
graph-regression-on-lipophilicityGraphGPS#8RMSE: 0.579±0.006R2: 0.790±0.004
graph-regression-on-pcqm4mv2-lscGPS#9Validation MAE: 0.0852Test MAE: 0.0862
graph-regression-on-peptides-structGPS#21MAE: 0.2500±0.0005
graph-regression-on-zincGPS#9MAE: 0.070 ± 0.002
graph-regression-on-zincGINE#10MAE: 0.070 ± 0.004
graph-regression-on-zinc-500kGPS#11MAE: 0.070
graph-regression-on-zinc-fullGraphGPS#8Test MAE: 0.024±0.007
link-prediction-on-pcqm-contactGPS#16MRR: 0.3337±0.0006
molecular-property-prediction-on-esolGraphGPS#9RMSE: 0.613±0.010R2: 0.911±0.003
molecular-property-prediction-on-freesolvGraphGPS#12RMSE: 1.462±0.188R2: 0.861±0.037
node-classification-on-clusterGPS#9Accuracy: 77.95
node-classification-on-coco-spGPS#7macro F1: 0.3412±0.0044
node-classification-on-pascalvoc-sp-1GPS#8macro F1: 0.3748±0.0109
node-classification-on-patternGPS#8Accuracy: 86.685