Simple Unsupervised Graph Representation Learning

In this paper, we propose a simple unsupervised graph representation learning method to conduct effective and efficient contrastive learning. Specifically, the proposed multiplet loss explores complementary information between structural neighbor enlarge inter-class variation, as well adds an upper bound achieve finite distance positive embeddings anchor for reducing intra-class variation. As result, both enlarging variation result in small generalization error, thereby obtaining model. Furthermore, our removes widely used data augmentation discriminator from previous methods, meanwhile available output low-dimensional embeddings, leading Experimental results on various real-world datasets demonstrate effectiveness efficiency of method, compared state-of-the-art methods. The source codes are released at https://github.com/YujieMo/SUGRL.