I am a master student at NCCU. I saw your SEAL method on OGB link property prediction task, and I am interested in this method. But I have a trivial question, hope you can solve my doubt.
I have read your code and I know that your SEAL method learn the embedding for subgraphs constructed from each pair of nodes. However, I wanna know that, can this method learn the embedding of individual nodes rather than the embedding of edges? Or it can only be used for learning a pair of nodes (i.e. edges)?
No. Using node labeling methods and enclosing subgraphs can also help node/graph representation learning other than edge learning. See this paper and this one.
Hi @muhanzhang,
I am a master student at NCCU. I saw your SEAL method on OGB link property prediction task, and I am interested in this method. But I have a trivial question, hope you can solve my doubt.
I have read your code and I know that your SEAL method learn the embedding for subgraphs constructed from each pair of nodes. However, I wanna know that, can this method learn the embedding of individual nodes rather than the embedding of edges? Or it can only be used for learning a pair of nodes (i.e. edges)?