search

PetarV- / GAT

Graph Attention Networks (https://arxiv.org/abs/1710.10903)
https://petar-v.com/GAT/
MIT License
3.18k stars 642 forks source link

Training HyperParameters on Pubmed and Citeseer #24

Closed aviczhl2 closed 5 years ago

aviczhl2 commented 5 years ago

I run the code on Citeseer and pubmed for multiple times and the performance is about 1% lower than expected in paper. I guess it's because of hyperparameter on these datasets being different from Cora. I'd like to ask which hyperparameters shall be changed to achieve the reported accuracy, thanks.

PetarV- commented 5 years ago

Hello,

Please refer to issue #14 for Citeseer and #12 for Pubmed.

Thanks, Petar

aviczhl2 commented 5 years ago

Thanks very much for your kind support.