Transfer Learning in Text, Multimedia and Networks
Today's data mining and machine learning applications must deal with the problem caused by a lack of high quality, annotated data. An effective solution to solving this problem is to borrow knowledge from other related but different domains. In this talk I give an overview of our recent work on transfer learning, which aims to extract knowledge from auxiliary domains that may have different data distribution and feature representations. I will first trace back to some early works in Artificial Intelligence, including learning by analogy. I will then discuss the questions of when, what and how to transfer the knowledge between domains, and present our current work on applying transfer learning to text mining, image classification and link prediction. I will conclude by discussing the potential impact of transfer learning research on some outstanding problems in artificial intelligence.
Prof. Qiang Yang
Date & Time
13 May 2011 (Friday) 10:00
Department of Computer and Information Science
Qiang Yang is a professor in the Department of Computer Science and Engineering, Hong Kong University of Science and Technology. He is an IEEE Fellow. His research interests are data mining and artificial intelligence (automated planning, machine learning and activity recognition). He received his PhD from Computer Science Department of the University of Maryland, College Park in 1989, and had been a faculty member at the University of Waterloo and Simon Fraser University in Canada between 1989 to 2001. His research teams won the 2004 and 2005 ACM KDDCUP international competitions on data mining. He is a vice chair of ACM SIGART and the founding Editor in Chief of the ACM Transactions on Intelligent Systems and Technology (ACM TIST). He is on the editorial board of several international journals and has been chairing several international conferences including ACM KDD 2010, ACM IUI 2010 and ACML 2010. He serves on the IJCAI 2011 advisory committee.