Replies: 3 comments 2 replies
-
Sorry, I don't fully understand yet. |
Beta Was this translation helpful? Give feedback.
-
Thanks, |
Beta Was this translation helpful? Give feedback.
-
I have different type of edges. nodes are the same. May I define the node classification task with different edge type? I mean, I aggregate the different types of edge feature length and add with target node's feature? |
Beta Was this translation helpful? Give feedback.
-
Hello everybody,
How HeteroGNN learn different types of edge with a different feature length?
conv = HeteroConv({ edge_type: SAGEConv((-1, -1), hidden_channels, aggr=aggr) for edge_type in metadata[1]}) self.convs.append(conv)
after
convs.append
, is the heteroconv convert to homogenous graph??if the features length are 3,5,10, after conv.append and aggregations of them the edge/node embedding length is 10? it means the bigest length is the length of embedding? it produce large of zero value?!
Beta Was this translation helpful? Give feedback.
All reactions