Blind separation of post-nonlinear mixtures using linearizing transformations and temporal decorrelation

Andreas Ziehe, Motoaki Kawanabe, Stefan Harmeling, Klaus Muller

Research output: Contribution to journalArticle

38 Citations (Scopus)

Abstract

We propose two methods that reduce the post-nonlinear blind source separation problem (PNL-BSS) to a linear BSS problem. The first method is based on the concept of maximal correlation: we apply the alternating conditional expectation (ACE) algorithm-a powerful technique from non-parametric statistics-to approximately invert the componentwise nonlinear functions. The second method is a Gaussianizing transformation, which is motivated by the fact that linearly mixed signals before nonlinear transformation are approximately Gaussian distributed. This heuristic, but simple and efficient procedure works as good as the ACE method. Using the framework provided by ACE, convergence can be proven. The optimal transformations obtained by ACE coincide with the sought-after inverse functions of the nonlinearities. After equalizing the nonlinearities, temporal decorrelation separation (TDSEP) allows us to recover the source signals. Numerical simulations testing "ACE-TD" and "Gauss-TD" on realistic examples are performed with excellent results.

Original languageEnglish
Pages (from-to)1319-1338
Number of pages20
JournalJournal of Machine Learning Research
Volume4
Issue number7-8
Publication statusPublished - 2004 Oct 1
Externally publishedYes

Fingerprint

Blind source separation
Conditional Expectation
Statistics
Nonlinearity
Computer simulation
Testing
Nonparametric Statistics
Inverse function
Nonlinear Transformation
Blind Source Separation
Invert
Nonlinear Function
Gauss
Linearly
Heuristics
Numerical Simulation

Keywords

  • Blind Source Separation
  • Gaussianization
  • Maximal Correlation
  • Post-nonlinear Mixture
  • Temporal Decorrelation

ASJC Scopus subject areas

  • Artificial Intelligence

Cite this

Blind separation of post-nonlinear mixtures using linearizing transformations and temporal decorrelation. / Ziehe, Andreas; Kawanabe, Motoaki; Harmeling, Stefan; Muller, Klaus.

In: Journal of Machine Learning Research, Vol. 4, No. 7-8, 01.10.2004, p. 1319-1338.

Research output: Contribution to journalArticle

Ziehe, Andreas ; Kawanabe, Motoaki ; Harmeling, Stefan ; Muller, Klaus. / Blind separation of post-nonlinear mixtures using linearizing transformations and temporal decorrelation. In: Journal of Machine Learning Research. 2004 ; Vol. 4, No. 7-8. pp. 1319-1338.
@article{a58006f47e684060a336ffe29b60c83e,
title = "Blind separation of post-nonlinear mixtures using linearizing transformations and temporal decorrelation",
abstract = "We propose two methods that reduce the post-nonlinear blind source separation problem (PNL-BSS) to a linear BSS problem. The first method is based on the concept of maximal correlation: we apply the alternating conditional expectation (ACE) algorithm-a powerful technique from non-parametric statistics-to approximately invert the componentwise nonlinear functions. The second method is a Gaussianizing transformation, which is motivated by the fact that linearly mixed signals before nonlinear transformation are approximately Gaussian distributed. This heuristic, but simple and efficient procedure works as good as the ACE method. Using the framework provided by ACE, convergence can be proven. The optimal transformations obtained by ACE coincide with the sought-after inverse functions of the nonlinearities. After equalizing the nonlinearities, temporal decorrelation separation (TDSEP) allows us to recover the source signals. Numerical simulations testing {"}ACE-TD{"} and {"}Gauss-TD{"} on realistic examples are performed with excellent results.",
keywords = "Blind Source Separation, Gaussianization, Maximal Correlation, Post-nonlinear Mixture, Temporal Decorrelation",
author = "Andreas Ziehe and Motoaki Kawanabe and Stefan Harmeling and Klaus Muller",
year = "2004",
month = "10",
day = "1",
language = "English",
volume = "4",
pages = "1319--1338",
journal = "Journal of Machine Learning Research",
issn = "1532-4435",
publisher = "Microtome Publishing",
number = "7-8",

}

TY - JOUR

T1 - Blind separation of post-nonlinear mixtures using linearizing transformations and temporal decorrelation

AU - Ziehe, Andreas

AU - Kawanabe, Motoaki

AU - Harmeling, Stefan

AU - Muller, Klaus

PY - 2004/10/1

Y1 - 2004/10/1

N2 - We propose two methods that reduce the post-nonlinear blind source separation problem (PNL-BSS) to a linear BSS problem. The first method is based on the concept of maximal correlation: we apply the alternating conditional expectation (ACE) algorithm-a powerful technique from non-parametric statistics-to approximately invert the componentwise nonlinear functions. The second method is a Gaussianizing transformation, which is motivated by the fact that linearly mixed signals before nonlinear transformation are approximately Gaussian distributed. This heuristic, but simple and efficient procedure works as good as the ACE method. Using the framework provided by ACE, convergence can be proven. The optimal transformations obtained by ACE coincide with the sought-after inverse functions of the nonlinearities. After equalizing the nonlinearities, temporal decorrelation separation (TDSEP) allows us to recover the source signals. Numerical simulations testing "ACE-TD" and "Gauss-TD" on realistic examples are performed with excellent results.

AB - We propose two methods that reduce the post-nonlinear blind source separation problem (PNL-BSS) to a linear BSS problem. The first method is based on the concept of maximal correlation: we apply the alternating conditional expectation (ACE) algorithm-a powerful technique from non-parametric statistics-to approximately invert the componentwise nonlinear functions. The second method is a Gaussianizing transformation, which is motivated by the fact that linearly mixed signals before nonlinear transformation are approximately Gaussian distributed. This heuristic, but simple and efficient procedure works as good as the ACE method. Using the framework provided by ACE, convergence can be proven. The optimal transformations obtained by ACE coincide with the sought-after inverse functions of the nonlinearities. After equalizing the nonlinearities, temporal decorrelation separation (TDSEP) allows us to recover the source signals. Numerical simulations testing "ACE-TD" and "Gauss-TD" on realistic examples are performed with excellent results.

KW - Blind Source Separation

KW - Gaussianization

KW - Maximal Correlation

KW - Post-nonlinear Mixture

KW - Temporal Decorrelation

UR - http://www.scopus.com/inward/record.url?scp=8344255796&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=8344255796&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:8344255796

VL - 4

SP - 1319

EP - 1338

JO - Journal of Machine Learning Research

JF - Journal of Machine Learning Research

SN - 1532-4435

IS - 7-8

ER -