search

PetarV- / GAT

Graph Attention Networks (https://arxiv.org/abs/1710.10903)
https://petar-v.com/GAT/
MIT License
3.15k stars 643 forks source link

Method for getting Standard Deviation for Metrics in Report #36

Closed akashb95 closed 4 years ago

akashb95 commented 4 years ago

Hi Petar,

I was wondering if you could tell me how many times you conducted the experiments to get the scores that you reported in your paper? Also, did you run cross-fold validation, and if so, how many folds did you use?

Many thanks, Akash.

PetarV- commented 4 years ago

Hi Akash,

Thank you for your issue and your interest in GAT!

I can confirm we used a single split (standard train/val/test split as normally used for Cora, Citeseer, Pubmed and PPI).

Transductive datasets (Cora, Citeseer, Pubmed) were executed 100 times, while the inductive dataset (PPI) was executed 10 times.

Hope this helps!

Thanks, Petar