GATConv seems to aggregate the features by default.
So if I give a FxN matrix to a GATConv, the output is a F' vector (where N is the number of nodes in the graph, and F, F' are the dimensions of the features.
Why is that? Is it on purpose?
Because it makes it impossible to have a chain of GATConv for example.
GATConv seems to aggregate the features by default.
So if I give a FxN matrix to a GATConv, the output is a F' vector (where N is the number of nodes in the graph, and F, F' are the dimensions of the features.
Why is that? Is it on purpose?
Because it makes it impossible to have a chain of GATConv for example.