A Survey on Multi-Interest Recommender Systems

PhD Qualifying Examination


Title: "A Survey on Multi-Interest Recommender Systems"

by

Miss Yueqi XIE


Abstract:

Recommender Systems (RS) serve as key components in various online 
platforms, which aim to provide personalized suggestions based on users' 
historical behavior and profile. Precise user modeling is crucial in RS to 
handle vast amounts of data and improve user experience and service 
providers' profits. Most deep learning-based RS model a user with one 
overall representation and perform matching with item candidates. However, 
user interests are typically diverse and complex, and modeling user 
interests in one representation may not be sufficient. Recent research 
proposes explicitly modeling users' multiple interests, leveraging the 
multi-interest nature of users to improve recommendation accuracy as well 
as diversity.

In this survey, we give a comprehensive review of multi-interest 
recommendation systems. We first introduce the background and task 
formulation of RS, as well as the key concept of multi-interest 
recommendation. Then, we introduce important contributions for improving 
multiple aspects of multi-interest recommendation, including advanced 
multi-interest extraction techniques, training schemes, information-aware 
feature enhancement, etc. We conclude by discussing several open 
challenges and opportunities in this field.


Date:			Wednesday, 14 December 2022

Time:                  	2:00pm - 4:00pm

Venue:                  Room 5566
                         Lifts 27/28

Committee Members:	Dr. Sunghun Kim (Supervisor)
  			Prof. Nevin Zhang (Chairperson)
 			Dr. Qifeng Chen
 			Prof. Raymond Wong


**** ALL are Welcome ****