Learning to Classify Networked Entities

Dr Dell Zhang
School of Computer Science and Information Systems (SCSIS), Birkbeck, University of London

Date: 7th October 2008 (Tuesday)
Time: 14:00 - 15:00
Venue: MB564

Statistical machine learning techniques for classification usually assume that all data instances are independent and identically distributed. However, real-life entities are often interconnected with each other by explicit or implicit relationships. It is promising to exploit the links among entities to enhance their classification. This talk first reviews the spectral methods for graph partitioning and community detection, and then presents several ways in which they can be extended for classification of networked entities.