Zhen Zhang

Hi, actually you can use average feature in x1,x2,x3 to make it work.

Hi, The current version only support sparse matrix. If you want to use dense matrix, the following parts should be modified: 1) use DenseGCNConv in convolution layer; 2) the calculation...

Hi, Sorry for the late reply. 1, The code of NodeInformationScore class indeed perform matrix multiplication in Eq(2), but they are in sparse multiplication form. You need to transform them...

Hi, This is very strange. Let me check it.

Hi, Sorry for the late reply. I found this is caused by the updating of torch-sparse. After the release of torch-sparse 0.4.4 ([reference here](https://github.com/rusty1s/pytorch_sparse/issues/45)), spspmm does not support autograd. However,...

These three options ```--sample_neighbor```, ```--sparse_attention```, ```--structure_learning``` are defined as bool type, so they should be initialized as ```True``` or ```False```. I also find the performance drops when updating pytorch_geometric to...

Hi, I have tested the code. ``` python main.py --sample_neighbor False ``` And, the following code is inserted into the main.py. ``` print(args.sample_neighbor) ``` It'll output the False statement.

Hi, I'm afraid not, since the spspmm do not support auto-grad.

Hi, This is because you use the latest torch-sparse. An older version of torch-sparse is needed, lower than 0.4.4.

Hi, It can be adapted to a regression problem. I'm wondering what your input is and what your output is. By the way, what's your error information?