a-lucic / cf-gnnexplainer Goto Github PK

@a-lucic thanks for opensourcing the work can this be extended to opensource dataset like cora citeseer and pubmed which are used for nodeclassification task ? if so what are the code changes to be made

src/cf_explanation/gcn_perturb.py

self.gc1 = GraphConvolutionPerturb(nfeat, nhid)
self.gc2 = GraphConvolutionPerturb(nhid, nhid)
self.gc3 = GraphConvolution(nhid, nout)
self.lin = nn.Linear(nhid + nhid + nout, nclass)
self.dropout = dropout

in the GCNSyntheticPerturb class when the model layers are built the first two layers are the perturbed layers, while the 3rd one keeps being a simple convolution layer.

Why is that?

Hi Ana,

Thanks for your work in explaining GNN in advance. It's a novel idea to interpret GNN in a counterfactual way. However, when I was trying to run your code, the environment file specified in your readme is missing. Could you please upload it for reference? The command in your readme file is conda env create --file environment.yml.

Best,
Gui

Hello,
this library doesn't exist in your source.

Hi, when I try to run your code, there seems lack utils.util which has not yet been published on this repository? Looking forward to your update!

Great project!
However, I cannot install it through conda env create --file environment.yml
Can you provide a requirments.txt file for easy installation with pip?