Graph InfoClust: Leveraging cluster-level node information for unsupervised graph representation learning

Benchmark Model Rank Results
link-prediction-on-citeseerGraph InfoClust (GIC)#2AUC: 97AP: 96.8
link-prediction-on-corasGraphite-VAE#7AUC: 93.7%AP: 93.5%
link-prediction-on-pubmedGraph InfoClust (GIC)#11AUC: 93.7%AP: 93.5%
node-classification-on-amz-compGraph InfoClust (GIC)#7Accuracy: 81.5 ± 1.0
node-classification-on-amz-photoGraph InfoClust (GIC)#14Accuracy: 90.4 ± 1.0
node-classification-on-citeseerGraph InfoClust (GIC)#37Accuracy: 71.9 ± 1.4
node-classification-on-coauthor-csGraph InfoClust (GIC)#22Accuracy: 89.4 ± 0.4
node-classification-on-cora-fixed-20-node-perGraph InfoClust (GIC)#8Accuracy: 81.7 ± 1.5
node-classification-on-pubmedGraph InfoClust (GIC)#47Accuracy: 77.4 ± 1.9