Bipartite graph's edge orientation

[zcaicaros]

Hi,

Suppose I have a bipartite graph G whose adjacency matrix is as follows,

adj = [[0, 0, 1],
       [1, 1, 0]]

If I am correct, adj[0, 2] = 1 means node 0 from part 1 of G pointing to node 2 in the other part of G. So here, it makes an implicit assumption that the NxM matrix can only model a single orientation. What if I want to model node 2 in the other part of G also points to node 0 from part 1 of G? I have to create another adj like the following.

adj_flip = [[0, 1],
            [0, 1],
            [1, 0]]

So, for an undirected bipartite graph, do we need two small graphs with different adjacency matrices to describe it?

[rusty1s]

Yes, you will need to use two adjacency matrices. Luckily, the reverse adjacency matrix is just the transpose:

edge_index_rev = edge_index.flip([0])