balcilar / gnn-spectral-expressive-power

Code of "Analyzing the Expressive Power of Graph Neural Networks in a Spectral Perspective" paper published in ICLR2021

Hi there,

I’m currently working on Graph Neural Networks and Neural Image/Video Compression.

gnn-spectral-expressive-power's Issues

Python script for generating raw 2Dgrid graph and Filtering operation

Thank you for the great work! I really liked the analysis in this paper. For the appendix I, could you please also public your python script for generating the 2Dgrid graph and the corresponding filtering process? I would like to add different filters，like band-rejection filter. Thanks.

a small question about proof of Theorem 4

About Theorem 4 Proposition 2, you have a assumption D=pI, then you can get L=I-A/p. Why? L=D-A=pI-A, isn't it? Or, you want to row noramlize? Can you give me some explanation? It's a great paper.

Python script for generating raw bandclass matrices

Thank you for the great work! I really liked the analysis in this paper and the comprehensive experiments on different datasets. For the bandclass experiment, could you please also public your python script for generating the raw matrices? I would like to generate a similar dataset but with smaller graph sizes for my work. Thanks.