Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering

Benchmark Model Rank Results
skeleton-based-action-recognition-on-sbuChebyNet#3Accuracy: 96.00%
node-classification-on-citeseer-0-5ChebyNet#10Accuracy: 45.3%
node-classification-on-citeseer-1ChebyNet#10Accuracy: 59.4%
node-classification-on-cora-3ChebyNet#14Accuracy: 62.1%
node-classification-on-pubmed-0-03ChebyNet#14Accuracy: 45.3%
node-classification-on-pubmed-0-05ChebyNet#14Accuracy: 48.2%
node-classification-on-pubmed-0-1ChebyNet#14Accuracy: 55.2%
node-classification-on-cora-0-5ChebyNet#15Accuracy: 33.9%
node-classification-on-cora-1ChebyNet#15Accuracy: 44.2%
graph-property-prediction-on-ogbg-molpcbaChebNet#28Test AP: 0.2306 ± 0.0016Ext. data: NoValidation AP: 0.2372 ± 0.0018…
node-classification-on-citeseer-with-public-split-fixed-20-nodes-per-classChebyNet#38Accuracy: 70.1%
node-classification-on-cora-with-public-split-fixed-20-nodes-per-classChebyNet#39Accuracy: 78.0%
node-classification-on-pubmed-with-public-split-fixed-20-nodes-per-classChebyNet#42Accuracy: 69.8%
node-classification-on-citeseerChebNet#50Accuracy: 69.8%
node-classification-on-coraChebNet#50Accuracy: 81.2%
node-classification-on-pubmedChebNet#52Accuracy: 74.4%