超越同质性的多视图图形表示学习（英文版）

英文标题：Multi-View Graph Representation Learning Beyond Homophily中文摘要：该研究提出了一个多视角方法和多样化预文本任务引入的框架，即 Multi-view Graph Encoder (MVGE)，以捕捉图形中的不同信号，并在合成和现实数据集上进行了广泛实验，并显示出显着的性能改进。英文摘要：Unsupervised graph representation learning(GRL) aims to distill diverse graphinformation into task-agnostic embeddings without label supervision. Due to alack of support from labels, recent representation learning methods usuallyadopt self-supervised learning, and embeddings are learned by solving ahandcrafted auxiliary t