[Help] Why Node ID starts from 1 instead of 0?

[bwdeng20]

[Help] Why Node ID starts from 1 instead of 0?

[yule-BUAA]

Hello,

Thanks for your interest in our work!

We make the node ids and edge ids start from 1 because we use the node/edge with id 0 as the mask node/edge, which is easy to implement. Their features are also set to zero vectors. The mask node/edge is then used for padding by the dynamic graph learning methods.

For example, we pad the neighbors in NeighborSampler here and pad the edges here. Then, the dynamic graph learning methods perform masked computations. For instance, the causal attention here.

I hope this answer has well addressed your issue. And feel free to ask if there are any further questions.

Best,

Le Yu

[bwdeng20]

Thanks for your swift response :) I am actually still confused. I think I need some time to compare the CTDG implementations from PyG, DyGLib, and those official repos so that I can have a comprehensive understanding.

[yule-BUAA]

I hope you are doing well. And welcome to contact me if there are any issues with our DyGLib.