Information-Enriched Representation Learning for Recommender System

PhD Thesis Proposal Defence


Title: "Information-Enriched Representation Learning for Recommender System"

by

Miss Yueqi XIE


Abstract:

Recommender Systems (RS) are key components in various online platforms, 
developed to provide personalized item suggestions based on users' historical 
behavior and profiles. The effectiveness of RS hinges on acquiring accurate 
user and item representations to optimize the matching process, thereby 
ensuring recommendations align closely with users' interests. The available 
information in RS encompasses interaction data, representing the past 
interactions between users and items; side information, comprising various 
characteristics of items and user profiles; and universal information, such as 
general textual information and common knowledge, among others. This thesis 
proposal endeavors to advance representation learning in RS by better 
leveraging interaction information, side information, and universal 
information. Firstly, it investigates multi-interest learning to enhance the 
utilization of interaction information. We propose the REMI framework, which 
improves the learned multi-interest representations through both the 
optimization objective and the composition information. Secondly, the thesis 
explores the fusion of side information in RS, aiming to leverage different 
item and user characteristics to enhance representations (DIF-SR). Lastly, it 
examines the application of universal information, including pretrained 
foundation models, to enhance recommendations. Overall, we aim to enrich 
representation learning in RS through innovative model design and training 
strategies, thereby broadening and enhancing the utilization of information for 
more precise and personalized recommendations.


Date:                   Tuesday, 26 March 2024

Time:                   10:00am - 12:00noon

Venue:                  Room 4472
                        Lifts 25/26

Committee Members:      Dr. Sunghun Kim (Supervisor)
                        Dr. Qifeng Chen (Supervisor)
                        Dr. Shuai Wang (Chairperson)
                        Dr. Junxian He