I was reading about different graph machine learning tasks in this book (Chapter 1) here and to learn about node classification and graph classification tasks.

Then I looked at this paper here, which is about semi-supervised classification for node classification.

Question: Why is/can node classification (graph machine learning) be semi-supervised while graph classification is just supervised?

Attempted explanation: Is it because:

  1. It would be harder to chop off parts of the graphs for node classification in order to keep the testing nodes unobservable. On the contrary, graph classification is a process for which it is very easy to ignore the testing ’data’ (i.e. graphs) as they are completely separate from our training graphs; they are not physically connected
  2. Secondly, there is information to be gained from the testing nodes as is mentioned above - for example, the neighborhood of the node

Any help is appreciated. Thanks

  • $\begingroup$ Why do you think that graph classification is just supervised? I don't know much about this specific topic, but it's possible that someone has already applied some semi-supervised technique for graph classification. $\endgroup$
    – nbro
    Nov 3, 2021 at 13:11


You must log in to answer this question.

Browse other questions tagged .