Why did not implement "Graph Contrastive Learning with Adaptive Augmentation"
mhadnanali opened this issue · comments
I am reading your paper "Graph Contrastive Learning with Adaptive Augmentation" and want to compare its results with my work, but I see you did not implement this paper in PyGCL.
Any specific reason?
And Why GRACE have two implementations? While in the original GRACE paper. It would be really nice if you give some comments in implementations to differentiate. It would save some people a lot of time. Thanks in anticipation.
As far as I understand, one is Semi-Supervised learning, and the other is Self-Supervised. can you please verify it.
Hey mhadnanali. Thank you for your interest in PyGCL.
reg. GCA implementation: the implementation of GCA is on our schedule. We are currently trying to make PyGCL easier to use in the next milestone. After this, we will reproduce more models and add them into the library. So please stay tuned. :D
reg. difference between implementation of GRACE in original paper and in PyGCL: they should be functionally equivalent, and they actually follow the same settings (contrastive learning and then employ a LR classifier to evaluate embedding quality). It should be enough to use the implementation in PyGCL.
Thank you for the reply, I figured it out the GRACE, and thanks for the update on GCA,
About GCA as far as I understand, it uses GRACE as an encoder and adaptive augmentation. is not it?