报告题目: Mathematical Models for Coding based Image Classification Methods
报告人：Prof. Tong Zhang
In recent years, biologically inspired vision models such as sparse coding have become popular due to their strong empirical performance in image classification tasks. However, for a long time there had been no mathematical theory to explain why these methods are effective. In this talk, I will describe our recent efforts on developing a mathematical framework for these coding based image classification methods, where each image is represented as a distribution over local descriptors. Our mathematical framework has two essential components:
1. nonlinear function approximation using geometric structures of local descriptors;
2. functional learning on probability distributions motivated using a kernel method. The mathematical theory justifies why coding models are effective for image classification, and reveals its limitations.
Tong Zhang received a B.A. in mathematics and computer science from Cornell University in 1994 and a Ph.D. in Computer Science from Stanford University in 1998. After graduation, he worked at IBM T.J. Watson Research Center in Yorktown Heights, New York, and Yahoo Research in New York city.
He is currently a professor of statistics at Rutgers University.
His research interests include machine learning, algorithms for statistical computation, their mathematical analysis and applications.