Which characteristics my neural network (NN) model should have to be considered as a graph convolutional network (GCN) instead of a graph neural network (GNN)? I know that GCN is a variant of GNN, but what is the difference between these two types of NNs?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
I think it's a reasonable claim that all graph convolutional networks are graph neural networks, since they operate on graphs, and are NNs. However, there are graph neural networks which don't use graph convolutions.
For example, graphRNN is a generative neural network for graphs where an RNN is given all the previous nodes and edges, and decides whether or not to add a new node/edges to the existing graph, or to terminate the generation process.
