Title: Introduction to Graph Neural Networks

Speaker: Jiaxuan You

Abstract

In this lecture, we provide an introduction to the popular field of Graph Neural Networks (GNNs). Graph neural networks have shown extraordinary performance in various tasks, and could tame the complex nature of graphs. We will first discuss the motivation and benefits of GNNs. Then, we will present the core idea of GNNs, which is to conduct message passing computation across nodes, defined by nodes local neighborhood structure. Finally, we will introduce the exciting applications enabled by GNNs in node, edge, subgraph and graph level tasks.

Slides

Bio

Jiaxuan You is a 4th year CS PhD candidate at Stanford University, advised by Prof. Jure Leskovec. Jiaxuan's research aims to empower deep learning with graph/relational structured data, including learning from graphs, generating/optimizing graphs, leveraging graphs as the prior for neural networks, and applying these techniques in various domains. Jiaxuan has published 13 papers on top AI conferences (10 as first-author), has got over 1800 citations, and has got 3000+ Github stars. Jiaxuan has won the Best Student Paper Award in AAAI 2017 (CompSus track), and his research is supported by JPMC PhD fellowship and Baidu Scholarship. For more information, please visit Jiaxuan's homepage.