A question about implementation of graph convolution

[tomato18463]

Hi,

Thanks for sharing your code! I have got a question about the implementaion of graph convolution in this code.

According to here, your graph is normalized column-wise. But the matrix multiplication here seems to use each row of your graph to weight the features of different node and sum it together. I am not sure if it is correct. It seems the right way is to use each column of your graph to weight-and-sum the features, so that the sum of weights is one whatever the number of neighbours? Is there something wrong in my understanding here?

Thank you!

[zhysora]

Thanks for your good question.

These parts of codes are borrowed from this repo. You can make an issue there for a better answer.

[Sfadingz]

The normalized graph A is added to x1 in the matrix multiplication you mentioned. Based on what I learn about torch.einsum( ), it seems to use each column of x1 to multiply each column of x3, so I think it is correct.