An efficient method for computing orthogonal discriminant vectors

Jinghua Wang, Yong Xu, Dapeng Zhang, Jia You

Research output: Journal article publicationJournal articleAcademic researchpeer-review

17 Citations (Scopus)

Abstract

We propose a linear discriminant analysis method. In this method, every discriminant vector, except for the first one, is worked out by maximizing a Fisher criterion defined in a transformed space which is the null space of the previously obtained discriminant vectors. All of these discriminant vectors are used for dimension reduction. We also propose two algorithms to implement the model. Based on the algorithms, we prove that the discriminant vectors will be orthogonal if the within-class scatter matrix is not singular. The experimental results show that the proposed method is effective and efficient.
Original languageEnglish
Pages (from-to)2168-2176
Number of pages9
JournalNeurocomputing
Volume73
Issue number10-12
DOIs
Publication statusPublished - 1 Jun 2010

Keywords

  • Dimension reduction
  • Fisher discriminant analysis
  • Orthogonal discriminant vectors
  • Pattern recognition

ASJC Scopus subject areas

  • Computer Science Applications
  • Cognitive Neuroscience
  • Artificial Intelligence

Cite this