Lower bounds for Bayes error estimation

András Antos, Luc Devroye, L. Györfi

Research output: Contribution to journalArticle

36 Citations (Scopus)

Abstract

We give a short proof of the following result. Let (X, Y) be any distribution on N×{0, 1}, and let (X1, Y1), ..., (Xn, Yn) be an i.i.d. sample drawn from this distribution. In discrimination, the Bayes error L* = infg P{g(X)≠Y} is of crucial importance. Here we show that without further conditions on the distribution of (X, Y), no rate-of-convergence results can be obtained. Let φn(X1, Y1, ..., Xn, Yn) be an estimate of the Bayes error, and let {φn(.)} be a sequence of such estimates. For any sequence {an} of positive numbers converging to zero, a distribution of (X, Y) may be found such that E{|L*-φn(X1, Y1, ..., Xn, Yn)|}≥an infinitely often.

Original languageEnglish
Pages (from-to)643-645
Number of pages3
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume21
Issue number7
DOIs
Publication statusPublished - 1999

Fingerprint

Bayes Estimation
Error Estimation
Error analysis
Lower bound
Bayes
Convergence Results
Estimate
Discrimination
Rate of Convergence
Zero

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Electrical and Electronic Engineering
  • Artificial Intelligence
  • Computer Vision and Pattern Recognition

Cite this

Lower bounds for Bayes error estimation. / Antos, András; Devroye, Luc; Györfi, L.

In: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 21, No. 7, 1999, p. 643-645.

Research output: Contribution to journalArticle

@article{306cdc458edf435c8e4fd93ccd899cf3,
title = "Lower bounds for Bayes error estimation",
abstract = "We give a short proof of the following result. Let (X, Y) be any distribution on N×{0, 1}, and let (X1, Y1), ..., (Xn, Yn) be an i.i.d. sample drawn from this distribution. In discrimination, the Bayes error L* = infg P{g(X)≠Y} is of crucial importance. Here we show that without further conditions on the distribution of (X, Y), no rate-of-convergence results can be obtained. Let φn(X1, Y1, ..., Xn, Yn) be an estimate of the Bayes error, and let {φn(.)} be a sequence of such estimates. For any sequence {an} of positive numbers converging to zero, a distribution of (X, Y) may be found such that E{|L*-φn(X1, Y1, ..., Xn, Yn)|}≥an infinitely often.",
author = "Andr{\'a}s Antos and Luc Devroye and L. Gy{\"o}rfi",
year = "1999",
doi = "10.1109/34.777375",
language = "English",
volume = "21",
pages = "643--645",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE Computer Society",
number = "7",

}

TY - JOUR

T1 - Lower bounds for Bayes error estimation

AU - Antos, András

AU - Devroye, Luc

AU - Györfi, L.

PY - 1999

Y1 - 1999

N2 - We give a short proof of the following result. Let (X, Y) be any distribution on N×{0, 1}, and let (X1, Y1), ..., (Xn, Yn) be an i.i.d. sample drawn from this distribution. In discrimination, the Bayes error L* = infg P{g(X)≠Y} is of crucial importance. Here we show that without further conditions on the distribution of (X, Y), no rate-of-convergence results can be obtained. Let φn(X1, Y1, ..., Xn, Yn) be an estimate of the Bayes error, and let {φn(.)} be a sequence of such estimates. For any sequence {an} of positive numbers converging to zero, a distribution of (X, Y) may be found such that E{|L*-φn(X1, Y1, ..., Xn, Yn)|}≥an infinitely often.

AB - We give a short proof of the following result. Let (X, Y) be any distribution on N×{0, 1}, and let (X1, Y1), ..., (Xn, Yn) be an i.i.d. sample drawn from this distribution. In discrimination, the Bayes error L* = infg P{g(X)≠Y} is of crucial importance. Here we show that without further conditions on the distribution of (X, Y), no rate-of-convergence results can be obtained. Let φn(X1, Y1, ..., Xn, Yn) be an estimate of the Bayes error, and let {φn(.)} be a sequence of such estimates. For any sequence {an} of positive numbers converging to zero, a distribution of (X, Y) may be found such that E{|L*-φn(X1, Y1, ..., Xn, Yn)|}≥an infinitely often.

UR - http://www.scopus.com/inward/record.url?scp=0032663729&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0032663729&partnerID=8YFLogxK

U2 - 10.1109/34.777375

DO - 10.1109/34.777375

M3 - Article

VL - 21

SP - 643

EP - 645

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 7

ER -