Notebook companion to a Chinese graph neural network primer

What it does

This repository houses the runnable code for a Chinese-language introduction to graph neural networks. It translates the book’s chapters on GCN, GraphSage, graph classification, and graph autoencoders into Jupyter notebooks. If you are following the text, it provides the expected worked examples in PyTorch and standard scientific Python.

The interesting bit

The repo ships a formal errata PDF and openly flags conceptual vagueness in its own section 5.4 graph filter discussion. That level of transparency is baked into the repository rather than buried in a blog post.

Key highlights

• Notebooks mapped to specific chapters: node classification with GCN, GraphSage, graph classification, and graph autoencoders
• Includes 勘误.pdf correcting known errors, notably the fuzzy conceptual treatment of graph filters in section 5.4
• Cora dataset now uses local data after upstream download issues
• Built on PyTorch and standard scientific Python stacks

Caveats

• The README is sparse: a checklist of four chapters, dependency versions, and an acknowledgment of mistakes. Do not expect narrative guidance outside the notebooks.
• Coverage is deliberately narrow; only chapters 5, 7, 8, and 9 are present, so the repo is not a comprehensive GNN course on its own.

Verdict

Students following the Chinese text will find the expected scaffolding. Everyone else should treat it as a curated set of introductory GNN notebooks rather than a standalone framework.