Improving Graph Neural Networks with Simple Architecture Design

Benchmark Model Rank Results
node-classification-on-actorFSGNN (8-hop)#34Accuracy: 35.75 ± 0.96
node-classification-on-chameleonFSGNN (8-hop)#5Accuracy: 78.27±1.28
node-classification-on-chameleonFSGNN (3-hop)#6Accuracy: 78.14±1.25
node-classification-on-cornellFSGNN (8-hop)#2Accuracy: 87.84±6.19
node-classification-on-squirrelFSGNN (8-hop)#4Accuracy: 74.10±1.89
node-classification-on-texasFSGNN#11Accuracy: 87.30 ± 5.55
node-classification-on-wisconsinFSGNN (3-hop)#9Accuracy: 88.43±3.22
node-property-prediction-on-ogbn-papers100mFSGNN#7Test Accuracy: 0.6807 ± 0.0006Ext. data: No