How does PYG process duplicate edges?

[YangxinCs]

Hi,Thank you for providing a convenient tool for deep graph learning.But I have some questions about the use of this tool.If
I accidentally add duplicate edges when processing data, will the graph constructed by PYG become a complex graph instead of a simple graph? Are duplicate edges treated as parallel edges (multiple edges with the same edge between a pair of nodes)?

Looking forward to your answer.
YangXin

[songsong0425]

Hi. If you concern about the duplicated edges in the graph, you can simply run the RemoveDuplicatedEdges function in torch_geometric.transforms.

For example, here is a graph with duplicated edge:

# edge between 0 and 1 is duplicated
edge_index = torch.tensor([[0, 1, 1, 2, 0],
                           [1, 0, 2, 1, 1]], dtype=torch.long)
x= torch.tensor([[-1], [0], [1]], dtype=torch.float)

data = Data(x=x, edge_index=edge_index)
data
# Data(x=[3, 1], edge_index=[2, 5])

You can get a unique edges after run RemoveDuplicatedEdges() as below:

from torch_geometric.transforms import RemoveDuplicatedEdges

transform = RemoveDuplicatedEdges()
out = transform(data)
out
# Data(x=[3, 1], edge_index=[2, 4])

References:
https://pytorch-geometric.readthedocs.io/en/latest/generated/torch_geometric.transforms.RemoveDuplicatedEdges.html#torch_geometric.transforms.RemoveDuplicatedEdges
https://github.com/pyg-team/pytorch_geometric/blob/master/test/transforms/test_remove_duplicated_edges.py

[YangxinCs]

thank you very much!

[rusty1s]

Are duplicate edges treated as parallel edges (multiple edges with the same edge between a pair of nodes)

Yes :)

[YangxinCs]

thank you very much!