Skip to content

🛣[Deep Learning]Stanford CS224w:Machine Learning with Graphs

想说的话🎇

🔝课程网站：http://web.stanford.edu/class/cs224w/

👀一些资源: B站精讲：https://www.bilibili.com/video/BV1pR4y1S7GA/?spm_id_from=333.337.search-card.all.click&vd_source=280e4970f2995a05fdeab972a42bfdd0

https://github.com/TommyZihao/zihao_course/tree/main/CS224W

Slides: http://web.stanford.edu/class/cs224w/slides

Prediction with GNN

  • Node-level prediction

After GNN computation, we can get node embeddings:\(\{ \mathbf{h}_v^{l} \in \mathbb{R}^d, \forall v \in G \}\)

\[ \hat{y}^v = Head_{node} (\mathbf{h}_v^{l}) = \mathbf{W}^{H} \mathbf{h}_v^{l} \]

其中\(\hat{y}^v \in \mathbb{R}^k\)来表示要分类的k类

  • Edge-level prediction
\[ \hat{y}^{uv} = Head_{edge} (\mathbf{h}_v^{l}, \mathbf{h}_v^{l}) \]

(1)Concatenation + Linear

\[ \hat{y}^{uv} = Linear(Concat(\mathbf{h}_v^{l}, \mathbf{h}_u^{l})) \]

(2) Dot product

\[ \hat{y}^{uv} = (\mathbf{h}_u^{l})^T \mathbf{h}_v^{l} \]

This approach only applies to 𝟏-way prediction (e.g., link prediction: predict the existence of an edge)

  • Graph-level prediction
\[ \hat{y}^{G} = Head_{graph} ( \{\mathbf{h}_v^{l}, \forall v \in G \}) \]

Dataset Split:Fixed split/Random split