Component ordering in independent component analysis based on data power

A.J. Hendrikse, Lieuwe Jan Spreeuwers

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademic

209 Downloads (Pure)

Abstract

With Independent Component Analysis (ICA) the objective is to separate multidimensional data into independent components. A well known problem in ICA is that since both the independent components and the separation matrix have to be estimated, neither the ordering nor the amplitudes of the components can be determined. One suggested method for solving these ambiguities in ICA is to measure the data power of a component, which indicates the amount of input data variance explained by an independent component. This method resembles the eigenvalue ordering of principle components. We will demonstrate theoretically and with experiments that strong sources can be estimated with higher accuracy than weak components. Based on the selection by data power, a method is developed for estimating independent components in high dimensional spaces. A test with synthetic data shows that the new algorithm can provide higher accuracy than the usual PCA dimension reduction.
Original languageUndefined
Title of host publicationProceedings of the 28th Symposium on Information Theory in the Benelux
EditorsRaymond N.J. Veldhuis, R.N.J. Veldhuis, H.S. Cronie
Place of PublicationEindhoven
PublisherWerkgemeenschap voor Informatie- en Communicatietheorie (WIC)
Pages211-218
Number of pages8
ISBN (Print)978-90-365-2509-1
Publication statusPublished - 24 Jun 2007
Event28th Symposium on Information Theory in the Benelux 2007 - Best Western Dish Hotel, Enschede, Netherlands
Duration: 24 May 200725 May 2007
Conference number: 28

Publication series

Name
PublisherWerkgemeenschap voor Informatie- en Communicatietechniek
NumberLNCS4549

Conference

Conference28th Symposium on Information Theory in the Benelux 2007
CountryNetherlands
CityEnschede
Period24/05/0725/05/07

Keywords

  • EWI-10856
  • METIS-241832
  • IR-64279
  • SCS-Safety

Cite this

Hendrikse, A. J., & Spreeuwers, L. J. (2007). Component ordering in independent component analysis based on data power. In R. N. J. Veldhuis, R. N. J. Veldhuis, & H. S. Cronie (Eds.), Proceedings of the 28th Symposium on Information Theory in the Benelux (pp. 211-218). Eindhoven: Werkgemeenschap voor Informatie- en Communicatietheorie (WIC).
Hendrikse, A.J. ; Spreeuwers, Lieuwe Jan. / Component ordering in independent component analysis based on data power. Proceedings of the 28th Symposium on Information Theory in the Benelux. editor / Raymond N.J. Veldhuis ; R.N.J. Veldhuis ; H.S. Cronie. Eindhoven : Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), 2007. pp. 211-218
@inproceedings{535365103bd443819edc41acc344497d,
title = "Component ordering in independent component analysis based on data power",
abstract = "With Independent Component Analysis (ICA) the objective is to separate multidimensional data into independent components. A well known problem in ICA is that since both the independent components and the separation matrix have to be estimated, neither the ordering nor the amplitudes of the components can be determined. One suggested method for solving these ambiguities in ICA is to measure the data power of a component, which indicates the amount of input data variance explained by an independent component. This method resembles the eigenvalue ordering of principle components. We will demonstrate theoretically and with experiments that strong sources can be estimated with higher accuracy than weak components. Based on the selection by data power, a method is developed for estimating independent components in high dimensional spaces. A test with synthetic data shows that the new algorithm can provide higher accuracy than the usual PCA dimension reduction.",
keywords = "EWI-10856, METIS-241832, IR-64279, SCS-Safety",
author = "A.J. Hendrikse and Spreeuwers, {Lieuwe Jan}",
year = "2007",
month = "6",
day = "24",
language = "Undefined",
isbn = "978-90-365-2509-1",
publisher = "Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)",
number = "LNCS4549",
pages = "211--218",
editor = "Veldhuis, {Raymond N.J.} and R.N.J. Veldhuis and H.S. Cronie",
booktitle = "Proceedings of the 28th Symposium on Information Theory in the Benelux",
address = "Netherlands",

}

Hendrikse, AJ & Spreeuwers, LJ 2007, Component ordering in independent component analysis based on data power. in RNJ Veldhuis, RNJ Veldhuis & HS Cronie (eds), Proceedings of the 28th Symposium on Information Theory in the Benelux. Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), Eindhoven, pp. 211-218, 28th Symposium on Information Theory in the Benelux 2007, Enschede, Netherlands, 24/05/07.

Component ordering in independent component analysis based on data power. / Hendrikse, A.J.; Spreeuwers, Lieuwe Jan.

Proceedings of the 28th Symposium on Information Theory in the Benelux. ed. / Raymond N.J. Veldhuis; R.N.J. Veldhuis; H.S. Cronie. Eindhoven : Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), 2007. p. 211-218.

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademic

TY - GEN

T1 - Component ordering in independent component analysis based on data power

AU - Hendrikse, A.J.

AU - Spreeuwers, Lieuwe Jan

PY - 2007/6/24

Y1 - 2007/6/24

N2 - With Independent Component Analysis (ICA) the objective is to separate multidimensional data into independent components. A well known problem in ICA is that since both the independent components and the separation matrix have to be estimated, neither the ordering nor the amplitudes of the components can be determined. One suggested method for solving these ambiguities in ICA is to measure the data power of a component, which indicates the amount of input data variance explained by an independent component. This method resembles the eigenvalue ordering of principle components. We will demonstrate theoretically and with experiments that strong sources can be estimated with higher accuracy than weak components. Based on the selection by data power, a method is developed for estimating independent components in high dimensional spaces. A test with synthetic data shows that the new algorithm can provide higher accuracy than the usual PCA dimension reduction.

AB - With Independent Component Analysis (ICA) the objective is to separate multidimensional data into independent components. A well known problem in ICA is that since both the independent components and the separation matrix have to be estimated, neither the ordering nor the amplitudes of the components can be determined. One suggested method for solving these ambiguities in ICA is to measure the data power of a component, which indicates the amount of input data variance explained by an independent component. This method resembles the eigenvalue ordering of principle components. We will demonstrate theoretically and with experiments that strong sources can be estimated with higher accuracy than weak components. Based on the selection by data power, a method is developed for estimating independent components in high dimensional spaces. A test with synthetic data shows that the new algorithm can provide higher accuracy than the usual PCA dimension reduction.

KW - EWI-10856

KW - METIS-241832

KW - IR-64279

KW - SCS-Safety

M3 - Conference contribution

SN - 978-90-365-2509-1

SP - 211

EP - 218

BT - Proceedings of the 28th Symposium on Information Theory in the Benelux

A2 - Veldhuis, Raymond N.J.

A2 - Veldhuis, R.N.J.

A2 - Cronie, H.S.

PB - Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)

CY - Eindhoven

ER -

Hendrikse AJ, Spreeuwers LJ. Component ordering in independent component analysis based on data power. In Veldhuis RNJ, Veldhuis RNJ, Cronie HS, editors, Proceedings of the 28th Symposium on Information Theory in the Benelux. Eindhoven: Werkgemeenschap voor Informatie- en Communicatietheorie (WIC). 2007. p. 211-218