Year: 2017
Journal of Information and Computing Science, Vol. 12 (2017), Iss. 3 : pp. 210–215
Abstract
Locality sensitive discriminant analysis is a typical and very effective graph-based dimensionality reduction method which has been successfully applied in pattern recognition problems. LSDA aims to find a projection which maximizes the margin between data points from different classes at each local area. As a result, it can discover the local geometrical structure of the data samples. However, just as linear discriminant analysis, it has the small sample size (SSS) problem. To overcome this limitation, we propose a novel exponential locality sensitive discriminant analysis algorithm in this paper. The proposed algorithm can make nearby objects with the same labels in the input space also nearby in the new representation; while nearby objects with different labels in the input space should be far apart. In addition, it can also deal with the SSS problem. The experiments on gene expression data sets verify the effectiveness of the proposed algorithm.
Journal Article Details
Publisher Name: Global Science Press
Language: English
DOI: https://doi.org/2024-JICS-22479
Journal of Information and Computing Science, Vol. 12 (2017), Iss. 3 : pp. 210–215
Published online: 2017-01
AMS Subject Headings:
Copyright: COPYRIGHT: © Global Science Press
Pages: 6