Robust extraction of local structures by the minimum β-divergence method

Md Nurul Haque Mollah, Nayeema Sultana, Mihoko Minami, Shinto Eguchi

Research output: Contribution to journalArticle

18 Citations (Scopus)

Abstract

This paper discusses a new highly robust learning algorithm for exploring local principal component analysis (PCA) structures in which an observed data follow one of several heterogeneous PCA models. The proposed method is formulated by minimizing β-divergence. It searches a local PCA structure based on an initial location of the shifting parameter and a value for the tuning parameter β. If the initial choice of the shifting parameter belongs to a data cluster, then the proposed method detects the local PCA structure of that data cluster, ignoring data in other clusters as outliers. We discuss the selection procedures for the tuning parameter β and the initial value of the shifting parameter μ in this article. We demonstrate the performance of the proposed method by simulation. Finally, we compare the proposed method with a method based on a finite mixture model.

Original languageEnglish
Pages (from-to)226-238
Number of pages13
JournalNeural Networks
Volume23
Issue number2
DOIs
Publication statusPublished - 2010 Mar

Fingerprint

Principal component analysis
Principal Component Analysis
Tuning
Learning algorithms
Learning

Keywords

  • β-divergence
  • Adaptive selection for the tuning parameter
  • Cross validation
  • Initialization of the parameters
  • Local PCA
  • Sequential estimation

ASJC Scopus subject areas

  • Artificial Intelligence
  • Cognitive Neuroscience

Cite this

Robust extraction of local structures by the minimum β-divergence method. / Nurul Haque Mollah, Md; Sultana, Nayeema; Minami, Mihoko; Eguchi, Shinto.

In: Neural Networks, Vol. 23, No. 2, 03.2010, p. 226-238.

Research output: Contribution to journalArticle

Nurul Haque Mollah, Md ; Sultana, Nayeema ; Minami, Mihoko ; Eguchi, Shinto. / Robust extraction of local structures by the minimum β-divergence method. In: Neural Networks. 2010 ; Vol. 23, No. 2. pp. 226-238.
@article{4b2e4c0c13494887bb7225803a7af5fc,
title = "Robust extraction of local structures by the minimum β-divergence method",
abstract = "This paper discusses a new highly robust learning algorithm for exploring local principal component analysis (PCA) structures in which an observed data follow one of several heterogeneous PCA models. The proposed method is formulated by minimizing β-divergence. It searches a local PCA structure based on an initial location of the shifting parameter and a value for the tuning parameter β. If the initial choice of the shifting parameter belongs to a data cluster, then the proposed method detects the local PCA structure of that data cluster, ignoring data in other clusters as outliers. We discuss the selection procedures for the tuning parameter β and the initial value of the shifting parameter μ in this article. We demonstrate the performance of the proposed method by simulation. Finally, we compare the proposed method with a method based on a finite mixture model.",
keywords = "β-divergence, Adaptive selection for the tuning parameter, Cross validation, Initialization of the parameters, Local PCA, Sequential estimation",
author = "{Nurul Haque Mollah}, Md and Nayeema Sultana and Mihoko Minami and Shinto Eguchi",
year = "2010",
month = "3",
doi = "10.1016/j.neunet.2009.11.011",
language = "English",
volume = "23",
pages = "226--238",
journal = "Neural Networks",
issn = "0893-6080",
publisher = "Elsevier Limited",
number = "2",

}

TY - JOUR

T1 - Robust extraction of local structures by the minimum β-divergence method

AU - Nurul Haque Mollah, Md

AU - Sultana, Nayeema

AU - Minami, Mihoko

AU - Eguchi, Shinto

PY - 2010/3

Y1 - 2010/3

N2 - This paper discusses a new highly robust learning algorithm for exploring local principal component analysis (PCA) structures in which an observed data follow one of several heterogeneous PCA models. The proposed method is formulated by minimizing β-divergence. It searches a local PCA structure based on an initial location of the shifting parameter and a value for the tuning parameter β. If the initial choice of the shifting parameter belongs to a data cluster, then the proposed method detects the local PCA structure of that data cluster, ignoring data in other clusters as outliers. We discuss the selection procedures for the tuning parameter β and the initial value of the shifting parameter μ in this article. We demonstrate the performance of the proposed method by simulation. Finally, we compare the proposed method with a method based on a finite mixture model.

AB - This paper discusses a new highly robust learning algorithm for exploring local principal component analysis (PCA) structures in which an observed data follow one of several heterogeneous PCA models. The proposed method is formulated by minimizing β-divergence. It searches a local PCA structure based on an initial location of the shifting parameter and a value for the tuning parameter β. If the initial choice of the shifting parameter belongs to a data cluster, then the proposed method detects the local PCA structure of that data cluster, ignoring data in other clusters as outliers. We discuss the selection procedures for the tuning parameter β and the initial value of the shifting parameter μ in this article. We demonstrate the performance of the proposed method by simulation. Finally, we compare the proposed method with a method based on a finite mixture model.

KW - β-divergence

KW - Adaptive selection for the tuning parameter

KW - Cross validation

KW - Initialization of the parameters

KW - Local PCA

KW - Sequential estimation

UR - http://www.scopus.com/inward/record.url?scp=73949119816&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=73949119816&partnerID=8YFLogxK

U2 - 10.1016/j.neunet.2009.11.011

DO - 10.1016/j.neunet.2009.11.011

M3 - Article

C2 - 19963342

AN - SCOPUS:73949119816

VL - 23

SP - 226

EP - 238

JO - Neural Networks

JF - Neural Networks

SN - 0893-6080

IS - 2

ER -