Clarify Confused Nodes via Separated Learning

Benchmark Model Rank Results
node-classification-on-actorNCSAGE#1Accuracy: 43.89 ± 1.33
node-classification-on-actorNCGCN#2Accuracy: 43.16 ± 1.32
node-classification-on-amz-computersNCGCN#1Accuracy: 90.81 ± 0.46
node-classification-on-amz-computersNCSAGE#2Accuracy: 90.43 ± 0.72
node-classification-on-amz-photoNCSAGE#1Accuracy: 95.93 ± 0.36
node-classification-on-amz-photoNCGCN#3Accuracy: 95.45 ± 0.45
node-classification-on-coauthor-csNCGCN#1Accuracy: 96.64 ± 0.29
node-classification-on-coauthor-csNCSAGE#2Accuracy: 96.48 ± 0.25
node-classification-on-coauthor-physicsNCSAGE#1Accuracy: 98.69 ± 0.26
node-classification-on-coauthor-physicsNCGCN#2Accuracy: 98.63 ± 0.24
node-classification-on-cora-fullNCSAGE#1Accuracy: 72.58 ± 0.65%
node-classification-on-cora-full-supervisedNCGCN#7Accuracy: 73.42 ± 0.58%
node-classification-on-penn94NCGCN#10Accuracy: 84.74 ± 0.28
node-classification-on-penn94NCSAGE#17Accuracy: 81.77 ± 0.71
node-classification-on-pubmedNCGCN#1Accuracy: 91.64 ± 0.53
node-classification-on-pubmedNCSAGE#2Accuracy: 91.55 ± 0.38