You have found an Encouraging-Graph-Neural-Network by a little rookie in the to record his learning procedure on the research road of Graph Neural Network.
Why this course?
For me, I build this course to push myself to study harder and read more papers on Graph Neural Network. Also this can help me to do more presentation practice and improve my English speaking. Moreover, the great ambition is that, I am looking forward to writing a book about Graph Neural Network. I will draft the book gradually as the course goes on. I still know nothing about how to do this well, please feel free to leave the message to me.
For you, why choose this course since there are many good courses about learning on graph? I think there are majorly two ways to gain knowledge from the Graph Neural Network. One is some general course, for example, my favourite Jure’s course link. I think the main purpose for those course is to introduce and involve more people into the research of Graph Neural. It may be too basic for the researcher in graph domain. The other is the paper reading group like the great LoGaG. It aims to read the advanced paper on Graph. However, they require much knowledge in the specific domain, e.g., massive related works and mathematic knowledge. Our course aims to bridge the gap between introduction and advantaged topics. We will:
- Go basic: give more background knowledge introduction, e.g., Spectral Graph Theory, tradtional Manifold Learning on Graph.
- Go deep: introduce the development history of a specific GNN research topic, e.g, the oversmooth issue, the graph pooling operation.
The course information
All the information will be updated on the Slack work space. Feel free to give you valuable suggestion and discussion. You can also contact me via email: haitaoma AT msu DOT com.
I wish to complete this course in two years time, a really long time. Since I am also a Ph.D. student, I only have some spare time on Saturday for me to take effort for this course. Hopefully, I need one week to learn and write the learning material, one week to prepare the slides and record the course. It will be cancelled for paper deadlines.
I will first schedule the first few lecture by myself to find the suitable guidance to prepare the course.Then I think it is good to invite experts in the particular domain to work with me on some topics. Voluteer is also welcomed.
The schedule
| Date | Topics | Lecture | Reading material |
|---|---|---|---|
| 2022/09/09 | Course Introduction | [YouTube] [bilibili] | |
| 2022/09/24 | Introduction on Spectral Graph Theory | [YouTube] [bilibili][slides] | |
| Graph Convolutional Network | |||
| Oversmooth and deeper GNN | |||
| Regularization on Graph | |||
| Heterphoily Graph | |||
| Graph Pooling | |||
| Powerful Readout function in Graph Neural Network | |||
| Knowledge Graph Completion | |||
| Link Prediction | |||
| Matrix Factorization on Graph | |||
| Understanding on Graph Neural Network | |||
| Self-supervised Learning on Graph (1) | |||
| Self-supervised Learning on Graph (2) | |||
| Graph Structural Learning | |||
| Graph Generation |
Course Textbook
It will be a draft continually updated with throughout this course. All kinds of feedback is welcomed to improve this book.
Coming soon!
Prerequisites
I hope you can have a brief reading on the first few chapter on Deep Learning on Graph to have a basic knowledge on deep learning on graph.