If train_test_split_edges is replaced by randomlinksplit

[qbitsofalchemy]

In tutorial 6 if we use the randomlinksplit how does train_pos_edge_index get handled later in the code?

Does this line become: train_pos_edge_index = train_data.edge_index.to(device) ?

And how is the test function redefined? If we don't have pos edges nor neg edges, then can you describe what happens to the test function?

[btarun13]

Hiya, I was just going through the repo. In regards to newer functions, when using randomlinksplit. There is a parameter within split_labels which when =True will give you distinct "pos_edge_label" and "neg_edge_label".

After having a look at tutorial 6 , test function should look something like this:

from torch_geometric.utils import negative_sampling

def test(pos_edge_index, neg_edge_index=None):
    model.eval()
    with torch.no_grad():
        z = model.encode(x, train_pos_edge_index)

    # Generate negative samples dynamically if not provided
    if neg_edge_index is None:
        neg_edge_index = negative_sampling(
            edge_index=train_pos_edge_index,
            num_nodes=data.num_nodes,
            num_neg_samples=pos_edge_index.size(1)
        )

    return model.test(z, pos_edge_index, neg_edge_index)

Note: train_pos_edge_index would be derived as train_data.edge_index. in case negative samples are not present you can use sampling like this.