How Attentive are Graph Attention Networks?

Benchmark Model Rank Results
graph-classification-on-ddGATv2#40Accuracy: 75.966±2.191
graph-classification-on-enzymesGATv2#5Accuracy: 77.987±2.112
graph-classification-on-imdb-bGATv2#7Accuracy: 80.000±2.739
graph-classification-on-nci1GATv2#27Accuracy: 82.384±1.700
graph-classification-on-nci109GATv2#12Accuracy: 83.092±0.764
graph-classification-on-proteinsGATv2#24Accuracy: 77.679±2.187
graph-regression-on-esr2GATv2#6R2: 0.655±0.000RMSE: 0.518±0.655
graph-regression-on-f2GATv2#6R2: 0.885±0.000RMSE: 0.344±0.885
graph-regression-on-kitGATv2#6R2: 0.826±0.000RMSE: 0.453±0.826
graph-regression-on-lipophilicityGATv2#2RMSE: 0.534±0.014R2: 0.821±0.009
graph-regression-on-parp1GATv2#6R2: 0.919±0.000RMSE: 0.356±0.919
graph-regression-on-pgrGATv2#7R2: 0.666±0.000RMSE: 0.558±0.666
graph-regression-on-zinc-fullGATv2#17Test MAE: 0.079±0.004
molecular-property-prediction-on-esolGATv2#7RMSE: 0.549±0.020R2: 0.928±0.005
molecular-property-prediction-on-freesolvGATv2#3RMSE: 0.676±0.081R2: 0.970±0.007
node-property-prediction-on-ogbn-arxivGIANT+XRT+GATv2#17Test Accuracy: 0.7415 ± 0.0005Ext. data: Yes