I have 2 questions.
1. Why Graph Contrastive Learning?
I have seen the GCL is used for Self-supervised learning to classify the Graph. But, at the downstream task, it seems that the LABELs of the data should be used even though they are saying that GCL is used for classifying large of the unlabeled data. So, why is GCL used? for what?
2. what is how to build an unsupervised learning model on Graph data with no(never have) label?
I have millions of data that should be classified which cannot be labeled with human effort.
I've tried to find a clear answer a lot and I've tried to build the classification model with GraphCL, infoGraph, etc, but they eventually use labels of data at the downstream task.
I want to know why GCL is used, and what is the algorithm or model for classifying huge unlabeled data.
Plz, show me the way how to.
Thank you.