A Survey on Link Prediction Models for Networked Data

PhD Qualifying Examination


Title: "A Survey on Link Prediction Models for Networked Data"

Mr. Wei XIANG


Abstract:

Link prediction for networked data is a fundamental data mining
task in various application domains, including social network
analysis, information retrieval, recommendation systems, record
linkage, marketing and bioinformatics. There are variety of
techniques for the link prediction problem, ranging from graph
theory, metric learning, statistical relational learning to matrix
factorization and probabilistic graphical models. In this survey,
we organize the sparse related literature into a structured
presentation and summarize the recent research works on the link
prediction task. We categorize the current link prediction methods
into three classes: the node-wise similarity based methods aim to
seek an appropriate distance measurement for two objects; the
topological pattern based methods focus on exploiting either local
or global patterns that could well describing the network; graph
structure based methods try to learn a compact model that could
abstracting the networked data best. We will first review these
methods, from detailed approaches to the evolution of the ideas,
and then comment on their relative strengths and weaknesses.
Finally, we give a brief summary on them and discuss some
possible research issues.


Date:     		Friday, 12 December 2008

Time:                   3:00p.m.-5:00p.m.

Venue:                  Room 3501
 			lifts 25-26

Committee Members:      Prof. Qiang Yang (Supervisor)
 			Prof. Dik-Lun Lee (Chairperson)
 			Prof. Dit-Yan Yeung
 			Dr. Nevin Zhang


**** ALL are Welcome ****