ViewMaker Networks Learning Views for Unsupervised Representation Learning

ViewMaker Networks learning views for unsupervised representation learning. The idea here is, you know, rather than having an image and training it to predict a particular label, like is this a cat or a dog? So instead you take an image and you transform it in two different ways. And now that the sort of contrastive learning task is to match those two images together, but push away images in representation space that are derived from different examples. It's a cool paradigm because it doesn't require explicit human defined labels.