GAT
GAT copied to clipboard
Why use (adj.col, adj.row) instead of (adj.row, adj.col), the accuracy is higher?
Hi, Petar
Thanks for your great works!
I have some questions about the function preprocess_adj_bias() in utils/process.py
indices = np.vstack((adj.col, adj.row)).transpose() # This is where I made a mistake, I used (adj.row, adj.col) instead
I did some experiment, I can't understand why use (adj.col, adj.row) instead of (adj.row, adj.col), the accuracy is higher. Could you explain? Thanks a lot!
For anyone who is also confused, all sparse operations assume that the sparse tensor is row-major ordering, while (adj.row, adj.col)
in numpy is column-major ordering. Thus, we need to use (adj.col, adj.row)
.