Hi.
I would like to use graph neural networks (from pytorch geometric) for dealing with a dataset of graphs, where each graph has a different number of nodes and edges.
I would like to test graf convolutional networks (GCN) and graph attention networks (GAT). Are these approaches able to deal with such kind of dataset (with a heterogeneous number of nodes and edges)? This is not clear for me.
I plan to use these two layers, specifically:
https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.GCNConv
https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.GATConv