开启掘金成长之旅!这是我参与「掘金日新计划 · 2 月更文挑战」的第 19 天,点击查看活动详情
4. GNN和图论
- transductive learning / inductive learning
- 节点分类
- 典型任务
- 生物医药领域:药物发现drug discovery,蛋白质结构预测protein structure prediction[^1]
- 典型任务
- 链路预测(图学习中的链路预测任务(持续更新ing...))
- 图分类
- 图着色graph coloring
- clique是一个点集,在一个无向图中,这个点集中任意两个不同的点之间都是相连的。maximal clique是一个clique,这个clique不可以再加入任何一个新的结点构成新的clique
- graph summarization A Survey on Graph Neural Networks for Graph Summarization
- 子图学习
- subgraph neural networks / subgraph mining(NLP课题入门 | day 20)
- node2vec
- ChebNet Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering
- GCN
- GraphSAGE
- GAT
- GIN How Powerful are Graph Neural Networks?
- GGNN
- SGC
- JK-Net Representation Learning on Graphs with Jumping Knowledge Networks
- APPNP
- C&S
- GAE
- MotifNet: a motif-based Graph Convolutional Network for directed graphs
- SIGN: Scalable Inception Graph Neural Networks
- SSGC或SGC Simple Spectral Graph Convolution
- GBP Scalable Graph Neural Networks via Bidirectional Propagation
- 图扩散卷积graph diffusion convolution (GDC)(仅适用于同配图):怎么说呢,感觉就是用PPR之类的扩散方法重新构建出了一个新图 Diffusion improves graph learning gasteigerjo/gdc: Graph Diffusion Convolution, as proposed in "Diffusion Improves Graph Learning" (NeurIPS 2019) 原博文:Graph Diffusion Convolution - MSRM Blog 中文翻译:图扩散卷积:Graph_Diffusion_Convolution_jialonghao的博客-CSDN博客_图扩散