search

tkipf / gcn

Implementation of Graph Convolutional Networks in TensorFlow
MIT License
7.1k stars 2k forks source link

Different number of nodes? #193

Open minhhoangbui opened 3 years ago

minhhoangbui commented 3 years ago

First of all, thank you very much for your work in GCN. I have spent quite a lot of time to understand how this works for graphs with different number of nodes. How a predefined gcn model can do this given that the number of features of each node stays the same? Hope that you could explain this to me.

l42yc4t commented 2 years ago

In GCN layer, you don't need to fixed your number of nodes. You can check the layer first, and you'll see the weight matrix only does on features. You don't need to worry about different graph with different number of nodes. If I'm wrong plz tell me.