是否可以拓展到有向图的比较？

[WZPBelieveFuture]

张博士，您好！
有个问题想向您请教下，SEAL方法能否拓展到有向网络的链路预测？我看了下您的代码，应该是可以的，但是一个节点是基于指向该节点的邻居来修正自身节点是否合适? 是否将SEAL方法拓展到有向图与其他有向的图卷积方法计较过，该方法是否更优？
期待您的回复！
感谢

[muhanzhang]

Hi! Yes. SEAL can be applied to directed graphs. When working on directed graphs, you should include both incoming and outgoing neighbors when extracting enclosing subgraphs.

I haven't compared with GNNs specifically for directed graphs. But you may check SEAL_OGB, which achieves state-of-the-art result on ogbl-citation2, a very large directed graph. Basically SEAL is a framework for link prediction where you can apply whatever GNN in it (including directed GNNs). You may check that implementation for how SEAL handles directed edges.

[WZPBelieveFuture]

Thank you very much! At 2021-10-31 18:25:06, "Muhan Zhang" ***@***.***> wrote: Hi! Yes. SEAL can be applied to directed graphs. When working on directed graphs, you should include both incoming and outgoing neighbors when extracting enclosing subgraphs. I haven't compared with GNNs specifically for directed graphs. But you may check SEAL_OGB, which achieves state-of-the-art result on ogbl-citation2, a very large directed graph. Basically SEAL is a framework for link prediction where you can apply whatever GNN in it (including directed GNNs). You may check that implementation for how SEAL handles directed edges. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.