Analysis on supervised neighborhood preserving embedding

Andrew Teoh B.J., Ying Han Pang

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

Neighborhood Preserving Embedding (NPE) is an un-supervised dimensionality reduction technique. Hence, it is lacking of discriminative capability. Zeng and Luo have proposed Supervised Neighborhood Preserving Embedding (SNPE), which uses class infor-mation of training samples to better describe data intrinsic structure. The robustness of SNPE has been demonstrated since it yields promis-ing recognition results. However, there is no theoretical analysis to explain the good performance. Here, we show analytically that the neighborhood discriminant criterion, which manifested in the objective function of SNPE, is close resembled to Fisher discriminant criterion. SNPE is evaluated in ORL and PIE face databases. The inclusion of class information in data learning results superior performance of SNPE to NPE.

Original languageEnglish
Pages (from-to)1631-1637
Number of pages7
Journalieice electronics express
Volume6
Issue number23
DOIs
Publication statusPublished - 2009 Dec 30

Fingerprint

embedding
preserving
data structures
learning
education
inclusions

All Science Journal Classification (ASJC) codes

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Electrical and Electronic Engineering

Cite this

Teoh B.J., Andrew ; Pang, Ying Han. / Analysis on supervised neighborhood preserving embedding. In: ieice electronics express. 2009 ; Vol. 6, No. 23. pp. 1631-1637.
@article{42bb353bf5cc4d059c457350cbcc217c,
title = "Analysis on supervised neighborhood preserving embedding",
abstract = "Neighborhood Preserving Embedding (NPE) is an un-supervised dimensionality reduction technique. Hence, it is lacking of discriminative capability. Zeng and Luo have proposed Supervised Neighborhood Preserving Embedding (SNPE), which uses class infor-mation of training samples to better describe data intrinsic structure. The robustness of SNPE has been demonstrated since it yields promis-ing recognition results. However, there is no theoretical analysis to explain the good performance. Here, we show analytically that the neighborhood discriminant criterion, which manifested in the objective function of SNPE, is close resembled to Fisher discriminant criterion. SNPE is evaluated in ORL and PIE face databases. The inclusion of class information in data learning results superior performance of SNPE to NPE.",
author = "{Teoh B.J.}, Andrew and Pang, {Ying Han}",
year = "2009",
month = "12",
day = "30",
doi = "10.1587/elex.6.1631",
language = "English",
volume = "6",
pages = "1631--1637",
journal = "IEICE Electronics Express",
issn = "1349-2543",
publisher = "The Institute of Electronics, Information and Communication Engineers (IEICE)",
number = "23",

}

Analysis on supervised neighborhood preserving embedding. / Teoh B.J., Andrew; Pang, Ying Han.

In: ieice electronics express, Vol. 6, No. 23, 30.12.2009, p. 1631-1637.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Analysis on supervised neighborhood preserving embedding

AU - Teoh B.J., Andrew

AU - Pang, Ying Han

PY - 2009/12/30

Y1 - 2009/12/30

N2 - Neighborhood Preserving Embedding (NPE) is an un-supervised dimensionality reduction technique. Hence, it is lacking of discriminative capability. Zeng and Luo have proposed Supervised Neighborhood Preserving Embedding (SNPE), which uses class infor-mation of training samples to better describe data intrinsic structure. The robustness of SNPE has been demonstrated since it yields promis-ing recognition results. However, there is no theoretical analysis to explain the good performance. Here, we show analytically that the neighborhood discriminant criterion, which manifested in the objective function of SNPE, is close resembled to Fisher discriminant criterion. SNPE is evaluated in ORL and PIE face databases. The inclusion of class information in data learning results superior performance of SNPE to NPE.

AB - Neighborhood Preserving Embedding (NPE) is an un-supervised dimensionality reduction technique. Hence, it is lacking of discriminative capability. Zeng and Luo have proposed Supervised Neighborhood Preserving Embedding (SNPE), which uses class infor-mation of training samples to better describe data intrinsic structure. The robustness of SNPE has been demonstrated since it yields promis-ing recognition results. However, there is no theoretical analysis to explain the good performance. Here, we show analytically that the neighborhood discriminant criterion, which manifested in the objective function of SNPE, is close resembled to Fisher discriminant criterion. SNPE is evaluated in ORL and PIE face databases. The inclusion of class information in data learning results superior performance of SNPE to NPE.

UR - http://www.scopus.com/inward/record.url?scp=72649096254&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=72649096254&partnerID=8YFLogxK

U2 - 10.1587/elex.6.1631

DO - 10.1587/elex.6.1631

M3 - Article

AN - SCOPUS:72649096254

VL - 6

SP - 1631

EP - 1637

JO - IEICE Electronics Express

JF - IEICE Electronics Express

SN - 1349-2543

IS - 23

ER -