General Tensor Discriminant Analysis

Speaker:	Dr. Dacheng TAO
		Department of Computing
		Hong Kong Polytechnic University

Title:		"General Tensor Discriminant Analysis"

Date:		Monday, 7 May 2007

Time:		4:00pm - 5:00pm

Venue:		Lecture Theatre F
		(Leung Yat Sing Lecture Theatre, near lift nos. 25/26)
		HKUST

Abstract:

Linear discriminant analysis (LDA) sheds light on classification tasks in
computer vision. However, classification based on LDA can perform poorly
in applications because LDA has the small sample size (SSS) problem, which
is closely relevant to the overfitting problem. The problem arises when
there are too few training measurements. To address the SSS problem in
LDA, a general tensor discriminant analysis (GTDA) is developed. GTDA, a
multilinear extension of a modified LDA, makes better use of the structure
information of the objects in computer vision research. It involves the
estimation of a series of projection matrices to project an object in the
form of a tensor from a high dimensional feature space to a low
dimensional feature space. Comparing with two dimensional LDA, the
significance of GTDA is its training stage converges as shown by
mathematical proof. Experiments on human gait recognition demonstrate that
GTDA combined with LDA and the nearest neighbour rule outperforms
competing methods.


******************
Biography:

Dacheng TAO received the B.Eng. from the University of Science and
Technology of China (USTC), the M.Phil. from the Chinese University of
Hong Kong (CUHK), and the Ph.D. from the University of London (UoL). He is
currently an assistant professor at the Department of Computing in the
Hong Kong Polytechnic University. His research interests include
artificial intelligence, biometrics, computer vision, data mining, machine
learning, and visual surveillance. He published extensively at IEEE TPAMI,
TKDE, TIP, TMM, TCSVT, CVPR, ICDM, ACM Multimedia, KDD, etc. Previously he
gained several Meritorious Awards from the Int'l Interdisciplinary Contest
in Modeling, which is the highest level mathematical modeling contest in
the world, organized by COMAP. He is an associate editor of Neurocomputing
(Elsevier). He is an editor of the forthcoming book entitled "Semantic
Mining Technologies for Multimedia Databases". He is a guest editor for
the Int'l Journal of Image and Graphics (World Scientific), the
Neurocomputing (Elsevier), the Pattern Recognition (Elsevier) special
issue, the Pattern Recognition Letters (Elsevier), and the Int'l Journal
of Imaging Systems and Technology (Wiley).