Hi, I noticed in your paper that you use features from each layer to reconstruct the graph and update the Graph Laplacian. However, I find in the released code, the Graph Laplacian is fixed according to its initial position and feature. Could you share the code or experience with the updated Graph Laplacian? Will the updated graph laplacian lead to better results? Thank you!
Kindly share your thoughts on the question put forward by @liarba. I'm currently working on GNNs and would like to know if the graph laplacian needs to be updated.
Hi, I noticed in your paper that you use features from each layer to reconstruct the graph and update the Graph Laplacian. However, I find in the released code, the Graph Laplacian is fixed according to its initial position and feature. Could you share the code or experience with the updated Graph Laplacian? Will the updated graph laplacian lead to better results? Thank you!