Determination of the optimal number of features for quadratic discriminant analysis via the normal approximation to the discriminant distribution Academic Article

abstract

• The use of the normal approximation to the discriminant distribution to determine the optimal number of features for quadratic discriminant analysis was discussed. The mean variance of the estimated discriminant was derived and feature-size optimization was compared using the normal approximation to the estimated discriminant with optimization obtained by simulating the true distribution of the estimated discriminant. Optimization via the normal approximation to the estimated discriminant provides huge computational savings in comparison to optimization via simulation of the true distribution. The results show that the feature-size optimization via the normal approximation is very accurate when the covariance matrices differ modestly.

authors

• Dougherty, Edward
• Xiong, Zixiang

published proceedings

• PATTERN RECOGNITION

author list (cited authors)

• Hua, J. P., Xiong, Z. X., & Dougherty, E. R.

citation count

• 23

complete list of authors

• Hua, JP||Xiong, ZX||Dougherty, ER

publication date

• March 2005

publisher

• Elsevier  Publisher

published in

• Pattern Recognition  Journal

keywords

• Classifier Error
• Discriminant Distribution
• Feature-set Size
• Quadratic Discriminant Analysis

Digital Object Identifier (DOI)

• 10.1016/j.patcog.2004.08.007

start page

• 403

end page

• 421

volume

• 38

issue

• 3

URL

• http%3A%2F%2Fdx.doi.org%2F10.1016%2Fj.patcog.2004.08.007