Graph-structured data (GSD) is ubiquitous in real-life applications, which appears in many learning applications such as property prediction for molecular graphs, product recommendations from heterogeneous information networks, and logical queries from knowledge graphs. Recently, learning from graph-structured data has also become a research focus in the machine learning community. However, again due to such diversities in GSD, there are no universal learning models that can perform well and consistently across different learning applications based on graphs. In sharp contrast to this, convolutional neural networks work well on natural images, and transformers are good choices for text data. In this tutorial, we will talk about using automated machine learning (AutoML) as a tool to design learning models for GSD. Specifically, we will elaborate on what is AutoML, what kind of prior information from graphs can be explored by AutoML, and how can insights be generated from the searched models.


Time Event
7:30-8:15 Part 1: Automated Graph Neural Network [Slides]
  Speaker: Huan Zhao
8:15-9:00 Part 2: Hyper-parameter Tuning for Graph-Structured Data. [Slides]
  Speaker: Quanming Yao
9:00-9:45 Part 3: Automated Knowledge Graph Reasoning: from Triplets to Subgraphs [Slides]
  Speaker: Yongqi Zhang
9:45-10:00 Part 4: Discussion


Quanming Yao, EE Department, Tsinghua University. Beijing. China.

Huan Zhao, 4Paradigm Inc. Beijing. China.

Yongqi Zhang, 4Paradigm Inc. Beijing. China.

Past Tutorial


Due to the space limitation, we only list highly-related papers.