Injecting noise for analysing the stability of ICA components

Stefan Harmeling; Frank Meinecke; Klaus Robert Müller

doi:10.1016/j.sigpro.2003.10.009

Injecting noise for analysing the stability of ICA components

Stefan Harmeling^*
, Frank Meinecke
, Klaus Robert Müller

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

16 Citations (Scopus)

Abstract

Usually, noise is considered to be destructive. We present a new method that constructively injects noise to assess the reliability and the grouping structure of empirical ICA component estimates. Our method can be viewed as a Monte-Carlo-style approximation of the curvature of some performance measure at the solution. Simulations show that the true root-mean-squared angle distances between the real sources and the source estimates can be approximated well by our method. In a toy experiment, we see that we are also able to reveal the underlying grouping structure of the extracted ICA components. Furthermore, an experiment with fetal ECG data demonstrates that our approach is useful for exploratory data analysis of real-world data.

Original language	English
Pages (from-to)	255-266
Number of pages	12
Journal	Signal Processing
Volume	84
Issue number	2
DOIs	https://doi.org/10.1016/j.sigpro.2003.10.009
Publication status	Published - 2004 Feb
Externally published	Yes

Bibliographical note

Funding Information:
The authors wish to thank the reviewers, furthermore also Gilles Blanchard, Benjamin Blankertz, Motoaki Kawanabe, and Andreas Ziehe for valuable discussions. This work was partly supported by the EU project (IST-1999-14190 – BLISS) and the support from the BMBF under contract FKZ 01IBB02A.

Keywords

ICA
Noise injection
Stability

ASJC Scopus subject areas

Control and Systems Engineering
Software
Signal Processing
Computer Vision and Pattern Recognition
Electrical and Electronic Engineering

Access to Document

10.1016/j.sigpro.2003.10.009

Cite this

@article{fd004054b8da48c5bde3304e3b0b7332,

title = "Injecting noise for analysing the stability of ICA components",

abstract = "Usually, noise is considered to be destructive. We present a new method that constructively injects noise to assess the reliability and the grouping structure of empirical ICA component estimates. Our method can be viewed as a Monte-Carlo-style approximation of the curvature of some performance measure at the solution. Simulations show that the true root-mean-squared angle distances between the real sources and the source estimates can be approximated well by our method. In a toy experiment, we see that we are also able to reveal the underlying grouping structure of the extracted ICA components. Furthermore, an experiment with fetal ECG data demonstrates that our approach is useful for exploratory data analysis of real-world data.",

keywords = "ICA, Noise injection, Stability",

author = "Stefan Harmeling and Frank Meinecke and M{\"u}ller, \{Klaus Robert\}",

note = "Funding Information: The authors wish to thank the reviewers, furthermore also Gilles Blanchard, Benjamin Blankertz, Motoaki Kawanabe, and Andreas Ziehe for valuable discussions. This work was partly supported by the EU project (IST-1999-14190 – BLISS) and the support from the BMBF under contract FKZ 01IBB02A. ",

year = "2004",

month = feb,

doi = "10.1016/j.sigpro.2003.10.009",

language = "English",

volume = "84",

pages = "255--266",

journal = "Signal Processing",

issn = "0165-1684",

publisher = "Elsevier B.V.",

number = "2",

}

TY - JOUR

T1 - Injecting noise for analysing the stability of ICA components

AU - Harmeling, Stefan

AU - Meinecke, Frank

AU - Müller, Klaus Robert

N1 - Funding Information: The authors wish to thank the reviewers, furthermore also Gilles Blanchard, Benjamin Blankertz, Motoaki Kawanabe, and Andreas Ziehe for valuable discussions. This work was partly supported by the EU project (IST-1999-14190 – BLISS) and the support from the BMBF under contract FKZ 01IBB02A.

PY - 2004/2

Y1 - 2004/2

N2 - Usually, noise is considered to be destructive. We present a new method that constructively injects noise to assess the reliability and the grouping structure of empirical ICA component estimates. Our method can be viewed as a Monte-Carlo-style approximation of the curvature of some performance measure at the solution. Simulations show that the true root-mean-squared angle distances between the real sources and the source estimates can be approximated well by our method. In a toy experiment, we see that we are also able to reveal the underlying grouping structure of the extracted ICA components. Furthermore, an experiment with fetal ECG data demonstrates that our approach is useful for exploratory data analysis of real-world data.

AB - Usually, noise is considered to be destructive. We present a new method that constructively injects noise to assess the reliability and the grouping structure of empirical ICA component estimates. Our method can be viewed as a Monte-Carlo-style approximation of the curvature of some performance measure at the solution. Simulations show that the true root-mean-squared angle distances between the real sources and the source estimates can be approximated well by our method. In a toy experiment, we see that we are also able to reveal the underlying grouping structure of the extracted ICA components. Furthermore, an experiment with fetal ECG data demonstrates that our approach is useful for exploratory data analysis of real-world data.

KW - ICA

KW - Noise injection

KW - Stability

UR - https://www.scopus.com/pages/publications/0347354863

U2 - 10.1016/j.sigpro.2003.10.009

DO - 10.1016/j.sigpro.2003.10.009

M3 - Article

AN - SCOPUS:0347354863

SN - 0165-1684

VL - 84

SP - 255

EP - 266

JO - Signal Processing

JF - Signal Processing

IS - 2

ER -

Injecting noise for analysing the stability of ICA components

Abstract

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this