search

EdisonLeeeee / MaskGAE

[KDD 2023] What’s Behind the Mask: Understanding Masked Graph Modeling for Graph Autoencoders
https://arxiv.org/abs/2205.10053
76 stars 6 forks source link

Edge-wise random masking and torch_geometric's `RandomLinkSplit` #8

Open ClaudMor opened 3 months ago

ClaudMor commented 3 months ago

Dear authors,

Congratulations for your very interesting work!

I was wondering whether the "Edge-wise random masking" strategy was actually equivalent to torch_geometric.transforms.RandomLinkSplit with disjoint_train_ratio > 0.

Thanks.

EdisonLeeeee commented 3 months ago

Thank you @ClaudMor

"Edge-wise random masking" strategy was actually equivalent to torch_geometric.transforms.RandomLinkSplit with disjoint_train_ratio > 0.

Yes, they indeed share similar insights but differ slightly in their implementations.