Robust independent component analysis

Sajjad H. Baloch; Hamid Krim; Marc G. Genton

Robust independent component analysis

Sajjad H. Baloch^*, Hamid Krim, Marc G. Genton

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

6 Scopus citations

Abstract

Independent Component Analysis (ICA) attempts to separate independent components present in the mixture signals. Several criteria have been suggested for ICA in the past, including kurtosis and negentropy. Kurtosis suffers from a drawback of being outlier sensitive. As a remedy, we propose Robust ICA (RICA), which employs appropriate robust estimators. In this paper, we compare the robustness properties of RICA with kurtosis- and negentropy-based ICA. Since robust estimators are insensitive to outliers in contrast to maximum likelihood estimates (MLE), we demonstrate that in the presence of outliers, RICA works better than kurtosis- and negentropy-based ICA.

Original language	English (US)
Title of host publication	2005 IEEE/SP 13th Workshop on Statistical Signal Processing - Book of Abstracts
Pages	61-64
Number of pages	4
Volume	2005
State	Published - 2005
Externally published	Yes
Event	2005 IEEE/SP 13th Workshop on Statistical Signal Processing - Bordeaux, France Duration: Jul 17 2005 → Jul 20 2005

Other

Other	2005 IEEE/SP 13th Workshop on Statistical Signal Processing
Country	France
City	Bordeaux
Period	07/17/05 → 07/20/05

ASJC Scopus subject areas

Signal Processing

Cite this

@inproceedings{6654dbdb225c45b3acd8d79b681e7331,

title = "Robust independent component analysis",

abstract = "Independent Component Analysis (ICA) attempts to separate independent components present in the mixture signals. Several criteria have been suggested for ICA in the past, including kurtosis and negentropy. Kurtosis suffers from a drawback of being outlier sensitive. As a remedy, we propose Robust ICA (RICA), which employs appropriate robust estimators. In this paper, we compare the robustness properties of RICA with kurtosis- and negentropy-based ICA. Since robust estimators are insensitive to outliers in contrast to maximum likelihood estimates (MLE), we demonstrate that in the presence of outliers, RICA works better than kurtosis- and negentropy-based ICA.",

author = "Baloch, {Sajjad H.} and Hamid Krim and Genton, {Marc G.}",

year = "2005",

language = "English (US)",

isbn = "0780394046",

volume = "2005",

pages = "61--64",

booktitle = "2005 IEEE/SP 13th Workshop on Statistical Signal Processing - Book of Abstracts",

note = "2005 IEEE/SP 13th Workshop on Statistical Signal Processing ; Conference date: 17-07-2005 Through 20-07-2005",

}

TY - GEN

T1 - Robust independent component analysis

AU - Baloch, Sajjad H.

AU - Krim, Hamid

AU - Genton, Marc G.

PY - 2005

Y1 - 2005

N2 - Independent Component Analysis (ICA) attempts to separate independent components present in the mixture signals. Several criteria have been suggested for ICA in the past, including kurtosis and negentropy. Kurtosis suffers from a drawback of being outlier sensitive. As a remedy, we propose Robust ICA (RICA), which employs appropriate robust estimators. In this paper, we compare the robustness properties of RICA with kurtosis- and negentropy-based ICA. Since robust estimators are insensitive to outliers in contrast to maximum likelihood estimates (MLE), we demonstrate that in the presence of outliers, RICA works better than kurtosis- and negentropy-based ICA.

AB - Independent Component Analysis (ICA) attempts to separate independent components present in the mixture signals. Several criteria have been suggested for ICA in the past, including kurtosis and negentropy. Kurtosis suffers from a drawback of being outlier sensitive. As a remedy, we propose Robust ICA (RICA), which employs appropriate robust estimators. In this paper, we compare the robustness properties of RICA with kurtosis- and negentropy-based ICA. Since robust estimators are insensitive to outliers in contrast to maximum likelihood estimates (MLE), we demonstrate that in the presence of outliers, RICA works better than kurtosis- and negentropy-based ICA.

UR - http://www.scopus.com/inward/record.url?scp=33947286875&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:33947286875

SN - 0780394046

SN - 9780780394049

VL - 2005

SP - 61

EP - 64

BT - 2005 IEEE/SP 13th Workshop on Statistical Signal Processing - Book of Abstracts

T2 - 2005 IEEE/SP 13th Workshop on Statistical Signal Processing

Y2 - 17 July 2005 through 20 July 2005

ER -

Robust independent component analysis

Abstract

Other

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this