Tighter mean-squared error bounds on kurtosis-based Fast-ICA

Matthew D. Kleffner; Douglas L. Jones

doi:10.1109/SSP.2007.4301320

Tighter mean-squared error bounds on kurtosis-based Fast-ICA

Matthew D. Kleffner, Douglas L. Jones

Electrical and Computer Engineering

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

Abstract

FastICA is a widely-used independent component analysis technique for blindly separating mixtures of instantaneously-mixed, independent sources recorded with multiple sensors. When using FastICA to estimate one source in interference, the unbiased mean-squared error can be bounded from above by the Schniter-Tong bounds on Shalvi-Weinstein estimators. We derive tighter upper bounds by extending both the Schniter-Tong proof and the Schniter-Johnson proof of upper bounds on constant-modulus estimators. These tighter bounds also exist over a wider range of sources and channels; existence gaps of over an order of magnitude of minimum-mean-squared error have been observed.

Original language	English (US)
Title of host publication	2007 IEEE/SP 14th Workshop on Statistical Signal Processing, SSP 2007, Proceedings
Pages	556-560
Number of pages	5
DOIs	https://doi.org/10.1109/SSP.2007.4301320
State	Published - 2007
Event	2007 IEEE/SP 14th WorkShoP on Statistical Signal Processing, SSP 2007 - Madison, WI, United States Duration: Aug 26 2007 → Aug 29 2007

Publication series

Name	IEEE Workshop on Statistical Signal Processing Proceedings

Other

Other	2007 IEEE/SP 14th WorkShoP on Statistical Signal Processing, SSP 2007
Country/Territory	United States
City	Madison, WI
Period	8/26/07 → 8/29/07

ASJC Scopus subject areas

Signal Processing

Online availability

10.1109/SSP.2007.4301320

Library availability

Discover UIUC Full Text

Cite this

Kleffner, MD & Jones, DL 2007, Tighter mean-squared error bounds on kurtosis-based Fast-ICA. in 2007 IEEE/SP 14th Workshop on Statistical Signal Processing, SSP 2007, Proceedings., 4301320, IEEE Workshop on Statistical Signal Processing Proceedings, pp. 556-560, 2007 IEEE/SP 14th WorkShoP on Statistical Signal Processing, SSP 2007, Madison, WI, United States, 8/26/07. https://doi.org/10.1109/SSP.2007.4301320

@inproceedings{8a3fc876eaeb449d9d4c66e3dccea871,

title = "Tighter mean-squared error bounds on kurtosis-based Fast-ICA",

abstract = "FastICA is a widely-used independent component analysis technique for blindly separating mixtures of instantaneously-mixed, independent sources recorded with multiple sensors. When using FastICA to estimate one source in interference, the unbiased mean-squared error can be bounded from above by the Schniter-Tong bounds on Shalvi-Weinstein estimators. We derive tighter upper bounds by extending both the Schniter-Tong proof and the Schniter-Johnson proof of upper bounds on constant-modulus estimators. These tighter bounds also exist over a wider range of sources and channels; existence gaps of over an order of magnitude of minimum-mean-squared error have been observed.",

author = "Kleffner, {Matthew D.} and Jones, {Douglas L.}",

year = "2007",

doi = "10.1109/SSP.2007.4301320",

language = "English (US)",

isbn = "142441198X",

series = "IEEE Workshop on Statistical Signal Processing Proceedings",

pages = "556--560",

booktitle = "2007 IEEE/SP 14th Workshop on Statistical Signal Processing, SSP 2007, Proceedings",

note = "2007 IEEE/SP 14th WorkShoP on Statistical Signal Processing, SSP 2007 ; Conference date: 26-08-2007 Through 29-08-2007",

}

TY - GEN

T1 - Tighter mean-squared error bounds on kurtosis-based Fast-ICA

AU - Kleffner, Matthew D.

AU - Jones, Douglas L.

PY - 2007

Y1 - 2007

N2 - FastICA is a widely-used independent component analysis technique for blindly separating mixtures of instantaneously-mixed, independent sources recorded with multiple sensors. When using FastICA to estimate one source in interference, the unbiased mean-squared error can be bounded from above by the Schniter-Tong bounds on Shalvi-Weinstein estimators. We derive tighter upper bounds by extending both the Schniter-Tong proof and the Schniter-Johnson proof of upper bounds on constant-modulus estimators. These tighter bounds also exist over a wider range of sources and channels; existence gaps of over an order of magnitude of minimum-mean-squared error have been observed.

AB - FastICA is a widely-used independent component analysis technique for blindly separating mixtures of instantaneously-mixed, independent sources recorded with multiple sensors. When using FastICA to estimate one source in interference, the unbiased mean-squared error can be bounded from above by the Schniter-Tong bounds on Shalvi-Weinstein estimators. We derive tighter upper bounds by extending both the Schniter-Tong proof and the Schniter-Johnson proof of upper bounds on constant-modulus estimators. These tighter bounds also exist over a wider range of sources and channels; existence gaps of over an order of magnitude of minimum-mean-squared error have been observed.

UR - http://www.scopus.com/inward/record.url?scp=47849116690&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=47849116690&partnerID=8YFLogxK

U2 - 10.1109/SSP.2007.4301320

DO - 10.1109/SSP.2007.4301320

M3 - Conference contribution

AN - SCOPUS:47849116690

SN - 142441198X

SN - 9781424411986

T3 - IEEE Workshop on Statistical Signal Processing Proceedings

SP - 556

EP - 560

BT - 2007 IEEE/SP 14th Workshop on Statistical Signal Processing, SSP 2007, Proceedings

T2 - 2007 IEEE/SP 14th WorkShoP on Statistical Signal Processing, SSP 2007

Y2 - 26 August 2007 through 29 August 2007

ER -

Tighter mean-squared error bounds on kurtosis-based Fast-ICA

Abstract

Publication series

Other

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this