Institute of Information Science Academia Sinica
Local, Global, and Hybrid Learning

     The use of classifiers permeates in various fields of 
engineering and science disciplines.  When constructing 
classifiers, there is a dichotomy in choosing whether to 
use local vs. global characteristics of the input data.  In 
this talk, we will describe our work on combining the local 
and global learning in the Maxi-Min Margin Machine (M^4).  
M^4 presents a unifying theory that subsumes the Support 
Vector Machine (SVM), the Minimax Probability Machine 
(MPM), and the Linear Discriminant Analysis (LDA).  While 
LDA and MPM focus on building the decision plane using 
global information and SVM focuses on building the decision 
plane in a local manner, M^4 incorporates these two 
seemingly different yet complementary characteristics in a 
unifying framework that achieves good classification 
accuracy.  We will present the formulation of M^4 and also 
experimental results to show the advantage of our novel 


     Irwin King received the BSc degree in Engineering and 
Applied Science from California Institute of Technology, 
Pasadena, in 1984. He received his MSc and PhD degree in 
Computer Science from the University of Southern 
California, Los Angeles, in 1988 and 1993 respectively.  He 
joined the Chinese University of Hong Kong in 1993.  His 
research interests include content-based multimedia 
information retrieval and statistical learning theory.

     He is a member of ACM, IEEE Computer Society, 
International Neural Network Society (INNS), and a 
governing board member of the Asian Pacific Neural Network 
Assembly (APNNA).  He has served as program and/or 
organizing member in international conferences, e.g., ACM 
Multimedia, WWW, ICASSP, IJCANN, ICONIP, etc.  He has also 
served as reviewer for many international journals, e.g.,