Kernel feature spaces and nonlinear blind source separation

Stefan Harmeling, Andreas Ziehe, Motoaki Kawanabe, Klaus Muller

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

21 Citations (Scopus)

Abstract

In kernel based learning the data is mapped to a kernel feature space of a dimension that corresponds to the number of training data points. In practice, however, the data forms a smaller submanifold in feature space, a fact that has been used e.g. by reduced set techniques for SVMs. We propose a new mathematical construction that permits to adapt to the intrinsic dimension and to find an orthonormal basis of this submanifold. In doing so, computations get much simpler and more important our theoretical framework allows to derive elegant kernelized blind source separation (BSS) algorithms for arbitrary invertible nonlinear mixings. Experiments demonstrate the good performance and high computational efficiency of our kTDSEP algorithm for the problem of nonlinear BSS.

Original languageEnglish
Title of host publicationAdvances in Neural Information Processing Systems
PublisherNeural information processing systems foundation
ISBN (Print)0262042088, 9780262042086
Publication statusPublished - 2002 Jan 1
Externally publishedYes
Event15th Annual Neural Information Processing Systems Conference, NIPS 2001 - Vancouver, BC, Canada
Duration: 2001 Dec 32001 Dec 8

Other

Other15th Annual Neural Information Processing Systems Conference, NIPS 2001
CountryCanada
CityVancouver, BC
Period01/12/301/12/8

Fingerprint

Blind source separation
Computational efficiency
Experiments

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Harmeling, S., Ziehe, A., Kawanabe, M., & Muller, K. (2002). Kernel feature spaces and nonlinear blind source separation. In Advances in Neural Information Processing Systems Neural information processing systems foundation.

Kernel feature spaces and nonlinear blind source separation. / Harmeling, Stefan; Ziehe, Andreas; Kawanabe, Motoaki; Muller, Klaus.

Advances in Neural Information Processing Systems. Neural information processing systems foundation, 2002.

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

Harmeling, S, Ziehe, A, Kawanabe, M & Muller, K 2002, Kernel feature spaces and nonlinear blind source separation. in Advances in Neural Information Processing Systems. Neural information processing systems foundation, 15th Annual Neural Information Processing Systems Conference, NIPS 2001, Vancouver, BC, Canada, 01/12/3.
Harmeling S, Ziehe A, Kawanabe M, Muller K. Kernel feature spaces and nonlinear blind source separation. In Advances in Neural Information Processing Systems. Neural information processing systems foundation. 2002
Harmeling, Stefan ; Ziehe, Andreas ; Kawanabe, Motoaki ; Muller, Klaus. / Kernel feature spaces and nonlinear blind source separation. Advances in Neural Information Processing Systems. Neural information processing systems foundation, 2002.
@inproceedings{69c524100f4542b791babe74e92efcc4,
title = "Kernel feature spaces and nonlinear blind source separation",
abstract = "In kernel based learning the data is mapped to a kernel feature space of a dimension that corresponds to the number of training data points. In practice, however, the data forms a smaller submanifold in feature space, a fact that has been used e.g. by reduced set techniques for SVMs. We propose a new mathematical construction that permits to adapt to the intrinsic dimension and to find an orthonormal basis of this submanifold. In doing so, computations get much simpler and more important our theoretical framework allows to derive elegant kernelized blind source separation (BSS) algorithms for arbitrary invertible nonlinear mixings. Experiments demonstrate the good performance and high computational efficiency of our kTDSEP algorithm for the problem of nonlinear BSS.",
author = "Stefan Harmeling and Andreas Ziehe and Motoaki Kawanabe and Klaus Muller",
year = "2002",
month = "1",
day = "1",
language = "English",
isbn = "0262042088",
booktitle = "Advances in Neural Information Processing Systems",
publisher = "Neural information processing systems foundation",

}

TY - GEN

T1 - Kernel feature spaces and nonlinear blind source separation

AU - Harmeling, Stefan

AU - Ziehe, Andreas

AU - Kawanabe, Motoaki

AU - Muller, Klaus

PY - 2002/1/1

Y1 - 2002/1/1

N2 - In kernel based learning the data is mapped to a kernel feature space of a dimension that corresponds to the number of training data points. In practice, however, the data forms a smaller submanifold in feature space, a fact that has been used e.g. by reduced set techniques for SVMs. We propose a new mathematical construction that permits to adapt to the intrinsic dimension and to find an orthonormal basis of this submanifold. In doing so, computations get much simpler and more important our theoretical framework allows to derive elegant kernelized blind source separation (BSS) algorithms for arbitrary invertible nonlinear mixings. Experiments demonstrate the good performance and high computational efficiency of our kTDSEP algorithm for the problem of nonlinear BSS.

AB - In kernel based learning the data is mapped to a kernel feature space of a dimension that corresponds to the number of training data points. In practice, however, the data forms a smaller submanifold in feature space, a fact that has been used e.g. by reduced set techniques for SVMs. We propose a new mathematical construction that permits to adapt to the intrinsic dimension and to find an orthonormal basis of this submanifold. In doing so, computations get much simpler and more important our theoretical framework allows to derive elegant kernelized blind source separation (BSS) algorithms for arbitrary invertible nonlinear mixings. Experiments demonstrate the good performance and high computational efficiency of our kTDSEP algorithm for the problem of nonlinear BSS.

UR - http://www.scopus.com/inward/record.url?scp=84898966891&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84898966891&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84898966891

SN - 0262042088

SN - 9780262042086

BT - Advances in Neural Information Processing Systems

PB - Neural information processing systems foundation

ER -