Self-Supervised Learning for Recommender Systems: Fundamental and Advances

Neural architecture-based recommenders have demonstrated overwhelming advantages over their traditional counterparts. However, the highly sparse user behavioral data often bottlenecks deep neural recommendation models to take full advantage of their capacity for better performance.

Recently, self-supervised learning (SSL), which can enable training on massive unlabeled data with automatic data annotation, has received tremendous attention across multiple fields including recommender systems. Self-supervised recommendation has now become the latest trend. 

In this tutorial, we will systematically introduce the methodologies of applying SSL to recommendation.

The topics to be covered include:

(1) foundation and overview of self-supervised recommendation;
(2) a comprehensive taxonomy of existing SSL-driven recommendation methods;
(3) how to apply SSL to various recommendation scenarios where different types of data and multiple optimization objectives are involved;
(4) limitations in current research and future research directions;
(5) an open-source toolkit to facilitate empirical comparisons and methodological development of self-supervised recommendation methods.

This session will be conducted online via Zoom: https://uqz.zoom.us/j/89362232168. 

Host

Dr Rocky Chen

Speaker

Junliang YU is a third-year Ph.D. candidate with ITEE, supervised by A/Prof. Hongzhi Yin. He completed his B.E. and M.S. degrees from Chongqing University, China. His research interests lie in recommender systems, social media analytics, deep learning on graphs, and self-supervised learning, with a particular focus on self-supervised recommendation. He has published 10+ papers on top-tier venues including KDD, WWW, CIKM, ICDM, AAAI, etc.