题 目（TITLE）：Feature Optimisation for Support Vector Classifier Learning
讲 座 人（SPEAKER）：Dr. Zhouyu Fu,（University of Sydney,Australia）
主 持 人 (CHAIR)：Prof. Weiming Hu
时 间 (TIME)：August 5, 2013 (Monday), 14:00-16:00
地 点 (VENUE)：No.3 Conferece Room (3rd floor), Intelligence Building
In this talk, I'll discuss the issue of feature optimisation for supervised learning with the support vector classifiers (SVM). The key to this problem is how to effectively associate classifier learning with feature optimisation. I then present a formulation which involves the use of only a single optimisation problem for solving both feature and classifier variables. Under some weak assumptions, we can show that the objective function of the optimisation problem is differentiable and thus can be efficiently solved by gradient descent. Finally I'll show two applications of the proposed framework on multiple instance learning and audio classification respectively.
Zhouyu Fu is a lecturer at the School of Computing, Engineering and Mathematics of University of Western Sydney (UWS). Before joining UWS in 2012, he had been a research fellow at the Gippsland School of Information Technology of Monash University since 2012. He did his PhD at the Australian National University and obtained his doctoral degree in Information Engineering in October 2009. He was also affiliated with and sponsored by National ICT Australia during his PhD studies at ANU. He obtained his master's degree in Pattern Recognition and Intelligent Systems from the National Lab of Pattern Recognition, Institute of Automation, the Chinese Academy of Sciences and bachelor's degree in Information Engineering from Zhejiang University in China.
Zhouyu's research interests are mainly in machine learning and pattern recognition, focusing on supervised learning techniques with applications to computer vision, image/audio analysis, and multimedia information retrieval.