Official code for "InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization" (ICLR 2020, spotlight)
I'm curious about the utility of PriorDiscriminator and GlobalDiscriminator, it seems that in your default parameter setting the code set prior=False and does not use PriorDiscriminator to cacluate part of the loss
I have removed GlobalDiscriminator. They are both experimental codes that are eventually not used nor mentioned in our paper.
If you are really curious about them, check out https://github.com/rdevon/DIM
I'm curious about the utility of PriorDiscriminator and GlobalDiscriminator, it seems that in your default parameter setting the code set prior=False and does not use PriorDiscriminator to cacluate part of the loss