Masking of adjacency matrix

Diego999 / pyGAT

Pytorch implementation of the Graph Attention Network model by Veličković et. al (2017, https://arxiv.org/abs/1710.10903)

MIT License

2.89k stars 689 forks source link

Masking of adjacency matrix #34

Open kochkinaelena opened 4 years ago

kochkinaelena commented 4 years ago

Hi,

It looks like the code is processing a single adjacency matrix that includes information about connections between graph nodes in training, development and testing sets. I was wondering where in the code the adjacency matrix is masked such that it is not using the parts that reflect the training set during testing?

Many thanks, Elena

jiangxinke commented 2 years ago

I wonder too. I think the author forget to perform "mask" on GAT processing. Then the features for test set nodes will aggregate on train node.