Talk Info: Hongwei Wang

Graph Representation Learning: From Knowledge Graphs to Recommender Systems

Abstract: Graphs are ubiquitous in the real world. To facilitate machine learning algorithms making use of graph-structured data, researchers proposed graph representation learning (GRL) methods, which learns a real low-dimensional vector for each node in a graph. In this talk, I will first briefly introduce graph representation learning, graph neural networks (GNNs, a special type of GRL methods), and knowledge graphs (KGs, a special type of graphs). Then my talk will consist of two parts: (1) Knowledge graph completion. I will introduce PathCon, a GNN-based method that combines relational context and relational paths information to predict the relation type of an edge in a KG. (2) Knowledge-graph-aware recommendation. Knowledge graphs can provide additional item-item relationship and thus alleviate the cold start problem in recommender systems. I will introduce two KG-aware recommendation algorithms, including an embedding-based method DKN and two structure-based methods KGCN and KGNN-LS.

Bio:

Hongwei Wang is a postdoctoral researcher at Computer Science Department, University of Illinois Urbana-Champaign. His research interests include machine learning and data mining, particularly in graph representation learning mechanisms, algorithms, and their applications in real-world data mining scenarios such as knowledge graphs, recommender systems, social networks, and sentiment analysis. He received Ph.D. degree from Department of Computer Science, Shanghai Jiao Tong University in 2018, and B.E. degree from ACM Class, Shanghai Jiao Tong University in 2014. He was a postdoctoral researcher at Computer Science Department, Stanford University from 2019 to 2021. He was one of the recipients of 2020 CCF (China Computer Federation) Outstanding Doctoral Dissertation Award and 2018 Google Ph.D. Fellowship.

Video Recording and Slides