Is Homophily a Necessity for Graph Neural Networks?

Ma Y., Liu X., Shah N. and Tang J. Is homophily a necessity for graph neural networks? ICLR, 2022.

概

• 探究 Homophily 假设 (即相互连接的结点相似) 对于 GCN 发挥效果是否是必须的.
• 结论是如果图中的同一类的结点具有相似的邻居的分布, 则 Homophily 不是必须的. 例如:

• 上图中, 每种颜色代表一个类别, 显然上图是一种非常异质的 (既然所有的边所连接的点都是不同类的).

• 但是, 对于结点 a, b 而言, 它们经过普通的 GCN 操作之后, 往往依然具有相似的表征 (倘若同类结点的初始表征也是相似的).

本栏目推荐文章
• 基于正则化的图自编码器在推荐算法中的应用 Application of graph auto-encoders based on regularization in recommendation algorithms
• 使用 Microsoft Graph oFFICE365 sendmail C#
• [ARC105E] Keep Graph Disconnected 题解
• pytorch反向传播错误解决：RuntimeError: Trying to backward through the graph a second time, but the buffers have already been freed. Specify retain_graph=True when calling backward the first time.
• CF295B Greg and Graph 题解 floyd性质题
• Unity3D Shader Graph 使用 DDXY 节点达到抗锯齿的原理详解
• 《A Novel Table-to-Graph Generation Approach for Document-Level Joint Entity and Relation Extraction》阅读笔记
• MMGCN: Multi-modal Graph Convolution Network for Personalized Recommendation of Micro-video
• Generalised f-Mean Aggregation for Graph Neural Networks
• Cold Brew: Distilling Graph Node Representations with Incomplete or Missing Neighborhoods

Homophily Necessity Networks Neural Graphhomophily necessity networks neural graph condensation networks neural learning networks neural hello understanding plasticity networks neural networks neural bigdataaiml-ibm-a introduction decoupling networks neural depth powerful networks spectral neural pre-train learning networks neural convolutional networks neural cnn sequence learning networks neural