A question about "pytorch_geometric/examples/link_pred.py"

[wxbs]

Thanks for your great work! Here is question about "pytorch_geometric/examples/link_pred.py". I found that if I replace Data.x (graph node features) with some random values ( e.g, torch.randn(num_nodes, num_features) ), the algorithm in link_pred.py still got a very high AUC value(more than 90%). Could you please tell me why it happened? Your early reply would be highly appreciated! Thank you very much!