In the prognn.py file, the EstimatedAdj holds an * ndense matrix , while you pass it to GCN, which require a sparse matrix. Could you please explain this ? I want to know whether to use sparse matrix or dense matrix, because dense matrix contains N*N parameters to train which is troublesome.
If you use a dense matrix as parameters to train, I want to know the limits of number of nodes in a graph. Since 1e4 nodes will need 1e8 paramaters to train. I wonder how you will process large-scale graphs?
Hi, Thanks for the great work.
I have two questions.
In the prognn.py file, the EstimatedAdj holds a
n * ndense matrix , while you pass it to GCN, which require a sparse matrix. Could you please explain this ? I want to know whether to use sparse matrix or dense matrix, because dense matrix containsN*Nparameters to train which is troublesome.If you use a dense matrix as parameters to train, I want to know the limits of number of nodes in a graph. Since 1e4 nodes will need 1e8 paramaters to train. I wonder how you will process large-scale graphs?
Thanks for your patience.