COLLABORATIVE FILTERING VIA CO-FACTORIZATION OF INDIVIDUALS AND GROUPS

MPhil Thesis Defence


Title: "COLLABORATIVE FILTERING VIA CO-FACTORIZATION OF INDIVIDUALS AND GROUPS"

By

Mr. Yihai HUANG


Abstract

Matrix factorization is one of the most successful collaborative filtering 
methods for recommender systems. Traditionally, matrix factorization only makes 
use of observed user-item feedback information so that predictions of cold 
users/items are difficult. In many real recommender systems, there is also 
available content information that has been successfully used in content-based 
methods. Thus in recent years, there is some work on how to incorporate content 
information into matrix factorization models and Factorization Machine(FM) is 
one of the most powerful integrated models among them. However, FM is 
originally designed as a generalized factorization model that models pairwise 
interactions between all features into a latent feature space. We find some 
issues of applying FM directly in the area of recommender system. In this 
thesis, we propose a novel matrix co-factorization model to solve some key 
limitations of Factorization Machine. Experimental results on benchmark data 
sets show that the method outperforms baseline methods especially for cold 
users and cold items.


Date:			Wednesday, 6 May 2015

Time:			4:30pm - 6:30pm

Venue:			Room 3501
 			Lifts 25/26

Committee Members:	Prof. James Kwok (Supervisor)
 			Prof. Dit-Yan Yeung (Chairperson)
 			Dr. Pan Hui


**** ALL are Welcome ****