AIAA 5028 Machine Learning on Graphs

Introduction

This course covers recent developments in machine learning on graph-structured data. Topics including network embedding, graph neural networks, knowledge graph embedding, generative models for graphs, scalable graph neural networks, explainable graph neural networks, and their applications. You are expected to finish several lab assignments, a survey on graph learning, a course project, and successfully participate a graph competition.

Course materials on Canvas.

Course Outline

1. Introduction
2. Traditional ML on Graphs
3. Network Embedding
4. Graph Neural Networks I
5. Graph Neural Networks II
6. Graph Neural Networks III
7. GNN on Complex Graphs
8. Knowledge Graph Embedding
9. Scalable GNNs
10. AutoML on Graphs
11. Pretraining on Graphs
12. Adversarial attack on Graphs
13. Explainable GNNs
14. GNN for Data Mining
15. GNN for Smart City
16. GNN for NLP&CV

Lab Exercise

You are required to finish three coding assignments with Python and Jupyter Notebook. Checkout assignments on Canvas.

• Deadline: please Submit your assignments via Canvas by 11:59 PM GMT+8 for each ddl
• Late submission: 10% penalty of total points for every day an assignment is late

Mid-term Exam

This course has a mid-term exam, coveraging topics including basic graph theory, network embedding, graph neural networks, knowledge graphs, etc.

Course Project

You are required to finish a course project with topics related to graph learning.

• Topic: checkout the following project topics or free research, make sure the novelty of your topic - Graph-based recommender system
  • Graph-based traffic prediction
  • Domain-specific knowledge graph
  • Graph-based drug discovery
  • TBD
• Grouping: No more than 3 students
• Mid-term project proposal: submit the title and abstract of your project
• Project report: 7 pages not including references in IJCAI format, should enclosed with code
• Final project presentation: a roadshow in the end of semester
• Consider to submit to IJCAI/ACL/ICML/KDD/SIGIR, and good luck!

Graph Competition

Finally, you are encoruaged to participate the Open Graph Benchmark (OGB) competition with using graph learning techniques.

• Competition: team up for a competition track, e.g.,Node classification, link prediction, Graph classification
• Grouping: no more than 3 students
• Grading formula: (100%-percentage your rank)/2
• Basic requirements:
  • Using graph learning techniques in the solution
  • Submit a brief report with code
  • Official implement