An efficient renovation on kernel Fisher discriminant analysis and face recognition experiments |
| |
Authors: | Yong Xu [Author Vitae] Jing-yu Yang [Author Vitae]Author Vitae] Dong-jun Yu [Author Vitae] |
| |
Affiliation: | Department of Computer Science, Nanjing University of Science and Technology, Nanjing, Jiangsu 210094, China |
| |
Abstract: | A reformative kernel algorithm, which can deal with two-class problems as well as those with more than two classes, on Fisher discriminant analysis is proposed. In the novel algorithm the supposition that in feature space discriminant vector can be approximated by some linear combination of a part of training samples, called “significant nodes”, is made. If the “significant nodes” are found out, the novel algorithm on kernel Fisher discriminant analysis will be superior to the naive one in classification efficiency. In this paper, a recursive algorithm for selecting “significant nodes”, is developed in detail. Experiments show that the novel algorithm is effective and much efficient in classifying. |
| |
Keywords: | Fisher discriminant analysis Kernel trick Pattern recognition Feature space |
本文献已被 ScienceDirect 等数据库收录! |
|