In Tutorial6 of PyG the dataset ends up having 455 edges considered negative edges (eg, edges that don’t exist in the original graph). In an AE are these edges needed in the Test set?
I train the Autoencoder on edges with a “negative” label (or 0). The test set will include a subset of “positive” labeled edges and some that are labeled as “negative”. The use of Autoencoder architecture is to use the reconstruction loss on Pos edges to decide if that they are indeed “negative” if the loss is below a threshold and “pos” otherwise.
Do I need to pass negative edges to the test function?
Note that after splitting the dataset, we end up with this profile:
And notice that in the epochs loop, we pass the data as is:
for epoch in range(1, epochs + 1):
loss = train()
auc, ap = test(data.test_pos_edge_index, data.test_neg_edge_index)
print('Epoch: {:03d}, AUC: {:.4f}, AP: {:.4f}'.format(epoch, auc, ap))
can I just pass: auc, ap = test(data.test_pos_edge_index) ?