Please use this identifier to cite or link to this item:
https://hdl.handle.net/2440/109539
Type: | Journal article |
Title: | Multi-class discriminant function based on canonical correlation in high dimension low sample size |
Author: | Tamatani, M. Naito, K. Koch, I. |
Citation: | Bulletin of Informatics and Cybernetics, 2013; 45:67-101 |
Publisher: | Research Association of Statistical Sciences |
Issue Date: | 2013 |
ISSN: | 0286-522X |
Statement of Responsibility: | Mitsuru Tamatani, Kanta Naito and Inge Koch |
Abstract: | In multi-class discriminant analysis for High Dimension Low Sample Size settings it is not possible to dene Fisher's discriminant function, since the sample covariance matrix is singular. For the special case of two-class problems, the naive Bayes rule has been studied, and combined with feature selection, this approach yields good practical results. We show how to extend the naive Bayes rule based on the naive canonical correlation matrix to a general setting for K≥2 classes, and we propose variable ranking and feature selection methods which integrate information from all K-1 eigenvectors. Provided the dimension does not grow too fast, we show that the K-1 sample eigenvectors are consistent estimators of the corresponding population parameters as both the dimension and sample size grow, and we give upper bounds for the misclassification rate. For real and simulated data we illustrate the performance of the new method which results in lower errors and typically smaller numbers of selected variables than existing methods. |
Keywords: | High dimension low sample size; canonical correlations; consistency; naive Bayes rule; misclassification; multi-class linear discriminant analysis |
Rights: | Copyright Status Unknown |
Published version: | https://www.lib.kyushu-u.ac.jp/en/publications_kyushu/bic |
Appears in Collections: | Aurora harvest 8 Mathematical Sciences publications |
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.