We noticed that NOSMOG showed superior performance and inference speed. However, during the training process, it relies on node label information. Can NOSMOG be used for unsupervised or self-supervised graph representation learning and then node clustering? How to improve it?
We noticed that NOSMOG showed superior performance and inference speed. However, during the training process, it relies on node label information. Can NOSMOG be used for unsupervised or self-supervised graph representation learning and then node clustering? How to improve it?