题目： Support Vector Machines Made Simpler 报告人：Ivor Tsang
Assistant Professor
School of Computer Engineering
Nanyang Technological University, Singapore 时间：6月23日（星期一）， 10:30-11:30 地点：蒙民伟楼404会议室
摘要：
Core vector machine (CVM) is a recent approach for scaling up kernel methods b
ased on the notion of minimum enclosing ball (MEB). However, an efficient impl
ementation requires sophisticated numerical solvers. I introduce enclosing bal
l (EB) problem where the ball's radius is fixed and thus does not have to be m
inimized. I develop efficient approximation algorithms that are simple to
implement and do not require any sophisticated numerical solver. Experiments s
how that the proposed algorithm has accuracies comparable to other SVMs, but c
an handle very large data sets and is even faster than CVM. Beside this, I pro
pose a multiplicative update of SVM, which can be formulated as a Bregman proj
ection problem. Moreover, this update for SVM can be regarded as boosting Parz
en window classifiers. Motivated by the success of boosting, I then consider
the use of an ensemble of the partially trained SVMs. Experiments show that th
e proposed ensemble has even better accuracy than the best-tuned soft-margin S
VM.
简历：
Dr Ivor Wai-Hung Tsang received his Ph.D. degree in Computer Science from the
Hong Kong University of Science and Technology (HKUST) in 2007. He will join t
he School of Computer Engineering of Nanyang Technological University as an As
sistant Professor. He was awarded the prestigious IEEE Transactions on Neural
Networks Outstanding 2004 Paper Award in 2006. He was also awarded the Microso
ft Fellowship in 2005, the Best Paper Award from the IEEE Hong Kong
Chapter of Signal Processing Postgraduate Forum in 2006, and also the HKUST Ho
nor Outstanding Student in 2001. His scientific interests include machine lear
ning, kernel methods, large scale optimization, semi-supervised learning and p
attern recognitions.
|
