THUDM / GraphMAE

GraphMAE: Self-Supervised Masked Graph Autoencoders in KDD'22
478 stars 75 forks source link

Results on graph classification NCI1 dataset #64

Open AliothF opened 7 months ago

AliothF commented 7 months ago

Hi, I encountered a problem when reproducing the graph classification experiment. Except for the NCI1 data set, all other graph data sets can achieve the results in your paper, but the classification accuracy on the NCI1 data set can only reach about 68%, whether using the dgl or pyg framework. I used the hyperparameters you provided, and the experimental environment was basically the same as yours.