search

THUDM / GraphMAE

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

Pls provide the detailed hyperparameters #31

Closed Newiz430 closed 1 year ago

Newiz430 commented 1 year ago

Dear authors,

Awesome job you have done here! I'm interested in your work so I'm here with a request for the detailed specifications of GraphMAE, including encoder types & hyper-parameters of linear probing for each dataset (e.g. lr_f, weight_decay_f and max_epoch_f of node_classification_evaluation for each dataset). BTW I think it's kinda weird to use PReLU for every dataset and downstream task as the paper claimed. So I guess there are some unclarified configurations. I'd be grateful if you provide them :heart:

Newiz430 commented 1 year ago

I haven't noticed configs.yml until now. Sorry for this.