Generalized consistent error estimator of linear discriminant analysis

Amin Zollanvari, Edward R. Dougherty

Research output: Contribution to journalArticle

15 Citations (Scopus)

Abstract

A classifier is epistemologically vacuous without an accurate estimate of its true error rate. In situations where the number of sample points is of the same order of magnitude as the dimension of observations, serious issues arise with respect to the performance of error estimators. In this paper, we place the problem of synthesizing an error rate estimator of a common linear classifier in an asymptotic setting in which the number of sample points is kept comparable in magnitude to the dimension of observations (double asymptotic). We construct a generalized consistent estimator of the true error rate for linear discriminant analysis in the multivariate Gaussian model under the assumption of a common covariance matrix. In other words, the estimator converges to true error rate in the double asymptotic sense. We employ simulations using both synthetic and real data to compare the performance of the new estimator to the classical consistent estimator of the true error (plug-in estimator) as well as other well-known estimators. We observe that the constructed estimator can outperform other estimators of the true error in many situations in terms of bias and root-mean-square (RMS) error.

Original languageEnglish
Article number7078850
Pages (from-to)2804-2814
Number of pages11
JournalIEEE Transactions on Signal Processing
Volume63
Issue number11
DOIs
Publication statusPublished - Jun 1 2015
Externally publishedYes

Fingerprint

Discriminant analysis
Classifiers
Covariance matrix
Mean square error

Keywords

  • Double asymptotics
  • error estimation
  • generalized consistent estimation
  • linear discriminant analysis
  • true error

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Signal Processing

Cite this

Generalized consistent error estimator of linear discriminant analysis. / Zollanvari, Amin; Dougherty, Edward R.

In: IEEE Transactions on Signal Processing, Vol. 63, No. 11, 7078850, 01.06.2015, p. 2804-2814.

Research output: Contribution to journalArticle

@article{0b04edda67b247e68f157a827facf403,
title = "Generalized consistent error estimator of linear discriminant analysis",
abstract = "A classifier is epistemologically vacuous without an accurate estimate of its true error rate. In situations where the number of sample points is of the same order of magnitude as the dimension of observations, serious issues arise with respect to the performance of error estimators. In this paper, we place the problem of synthesizing an error rate estimator of a common linear classifier in an asymptotic setting in which the number of sample points is kept comparable in magnitude to the dimension of observations (double asymptotic). We construct a generalized consistent estimator of the true error rate for linear discriminant analysis in the multivariate Gaussian model under the assumption of a common covariance matrix. In other words, the estimator converges to true error rate in the double asymptotic sense. We employ simulations using both synthetic and real data to compare the performance of the new estimator to the classical consistent estimator of the true error (plug-in estimator) as well as other well-known estimators. We observe that the constructed estimator can outperform other estimators of the true error in many situations in terms of bias and root-mean-square (RMS) error.",
keywords = "Double asymptotics, error estimation, generalized consistent estimation, linear discriminant analysis, true error",
author = "Amin Zollanvari and Dougherty, {Edward R.}",
year = "2015",
month = "6",
day = "1",
doi = "10.1109/TSP.2015.2419190",
language = "English",
volume = "63",
pages = "2804--2814",
journal = "IEEE Transactions on Signal Processing",
issn = "1053-587X",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "11",

}

TY - JOUR

T1 - Generalized consistent error estimator of linear discriminant analysis

AU - Zollanvari, Amin

AU - Dougherty, Edward R.

PY - 2015/6/1

Y1 - 2015/6/1

N2 - A classifier is epistemologically vacuous without an accurate estimate of its true error rate. In situations where the number of sample points is of the same order of magnitude as the dimension of observations, serious issues arise with respect to the performance of error estimators. In this paper, we place the problem of synthesizing an error rate estimator of a common linear classifier in an asymptotic setting in which the number of sample points is kept comparable in magnitude to the dimension of observations (double asymptotic). We construct a generalized consistent estimator of the true error rate for linear discriminant analysis in the multivariate Gaussian model under the assumption of a common covariance matrix. In other words, the estimator converges to true error rate in the double asymptotic sense. We employ simulations using both synthetic and real data to compare the performance of the new estimator to the classical consistent estimator of the true error (plug-in estimator) as well as other well-known estimators. We observe that the constructed estimator can outperform other estimators of the true error in many situations in terms of bias and root-mean-square (RMS) error.

AB - A classifier is epistemologically vacuous without an accurate estimate of its true error rate. In situations where the number of sample points is of the same order of magnitude as the dimension of observations, serious issues arise with respect to the performance of error estimators. In this paper, we place the problem of synthesizing an error rate estimator of a common linear classifier in an asymptotic setting in which the number of sample points is kept comparable in magnitude to the dimension of observations (double asymptotic). We construct a generalized consistent estimator of the true error rate for linear discriminant analysis in the multivariate Gaussian model under the assumption of a common covariance matrix. In other words, the estimator converges to true error rate in the double asymptotic sense. We employ simulations using both synthetic and real data to compare the performance of the new estimator to the classical consistent estimator of the true error (plug-in estimator) as well as other well-known estimators. We observe that the constructed estimator can outperform other estimators of the true error in many situations in terms of bias and root-mean-square (RMS) error.

KW - Double asymptotics

KW - error estimation

KW - generalized consistent estimation

KW - linear discriminant analysis

KW - true error

UR - http://www.scopus.com/inward/record.url?scp=84929118622&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84929118622&partnerID=8YFLogxK

U2 - 10.1109/TSP.2015.2419190

DO - 10.1109/TSP.2015.2419190

M3 - Article

AN - SCOPUS:84929118622

VL - 63

SP - 2804

EP - 2814

JO - IEEE Transactions on Signal Processing

JF - IEEE Transactions on Signal Processing

SN - 1053-587X

IS - 11

M1 - 7078850

ER -