Joint block diagonalization algorithms for optimal separation of multidimensional components

Dana Lahat; Jean François Cardoso; Hagit Messer

doi:10.1007/978-3-642-28551-6_20

Joint block diagonalization algorithms for optimal separation of multidimensional components

Dana Lahat, Jean François Cardoso, Hagit Messer

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

Abstract

This paper deals with non-orthogonal joint block diagonalization. Two algorithms which minimize the Kullback-Leibler divergence between a set of real positive-definite matrices and a block-diagonal transformation thereof are suggested. One algorithm is based on the relative gradient, and the other is based on a quasi-Newton method. These algorithms allow for the optimal, in the mean square error sense, blind separation of multidimensional Gaussian components. Simulations demonstrate the convergence properties of the suggested algorithms, as well as the dependence of the criterion on some of the model parameters.

Original language	English
Title of host publication	Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings
Pages	155-162
Number of pages	8
DOIs	https://doi.org/10.1007/978-3-642-28551-6_20
State	Published - 2012
Externally published	Yes
Event	10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012 - Tel Aviv, Israel Duration: 12 Mar 2012 → 15 Mar 2012

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	7191 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012
Country/Territory	Israel
City	Tel Aviv
Period	12/03/12 → 15/03/12

Keywords

Joint block diagonalization
quasi-Newton
relative gradient

Access to Document

10.1007/978-3-642-28551-6_20

Cite this

Lahat, D., Cardoso, J. F., & Messer, H. (2012). Joint block diagonalization algorithms for optimal separation of multidimensional components. In Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings (pp. 155-162). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7191 LNCS). https://doi.org/10.1007/978-3-642-28551-6_20

Lahat, Dana ; Cardoso, Jean François ; Messer, Hagit. / Joint block diagonalization algorithms for optimal separation of multidimensional components. Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. 2012. pp. 155-162 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{f88c460bb70b437783075c3cb203cd80,

title = "Joint block diagonalization algorithms for optimal separation of multidimensional components",

abstract = "This paper deals with non-orthogonal joint block diagonalization. Two algorithms which minimize the Kullback-Leibler divergence between a set of real positive-definite matrices and a block-diagonal transformation thereof are suggested. One algorithm is based on the relative gradient, and the other is based on a quasi-Newton method. These algorithms allow for the optimal, in the mean square error sense, blind separation of multidimensional Gaussian components. Simulations demonstrate the convergence properties of the suggested algorithms, as well as the dependence of the criterion on some of the model parameters.",

keywords = "Joint block diagonalization, quasi-Newton, relative gradient",

author = "Dana Lahat and Cardoso, {Jean Fran{\c c}ois} and Hagit Messer",

note = "Copyright: Copyright 2012 Elsevier B.V., All rights reserved.; 10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012 ; Conference date: 12-03-2012 Through 15-03-2012",

year = "2012",

doi = "10.1007/978-3-642-28551-6_20",

language = "אנגלית",

isbn = "9783642285509",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "155--162",

booktitle = "Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings",

}

Lahat, D, Cardoso, JF & Messer, H 2012, Joint block diagonalization algorithms for optimal separation of multidimensional components. in Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 7191 LNCS, pp. 155-162, 10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012, Tel Aviv, Israel, 12/03/12. https://doi.org/10.1007/978-3-642-28551-6_20

Joint block diagonalization algorithms for optimal separation of multidimensional components. / Lahat, Dana; Cardoso, Jean François; Messer, Hagit.
Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. 2012. p. 155-162 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7191 LNCS).

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

TY - GEN

T1 - Joint block diagonalization algorithms for optimal separation of multidimensional components

AU - Lahat, Dana

AU - Cardoso, Jean François

AU - Messer, Hagit

PY - 2012

Y1 - 2012

N2 - This paper deals with non-orthogonal joint block diagonalization. Two algorithms which minimize the Kullback-Leibler divergence between a set of real positive-definite matrices and a block-diagonal transformation thereof are suggested. One algorithm is based on the relative gradient, and the other is based on a quasi-Newton method. These algorithms allow for the optimal, in the mean square error sense, blind separation of multidimensional Gaussian components. Simulations demonstrate the convergence properties of the suggested algorithms, as well as the dependence of the criterion on some of the model parameters.

AB - This paper deals with non-orthogonal joint block diagonalization. Two algorithms which minimize the Kullback-Leibler divergence between a set of real positive-definite matrices and a block-diagonal transformation thereof are suggested. One algorithm is based on the relative gradient, and the other is based on a quasi-Newton method. These algorithms allow for the optimal, in the mean square error sense, blind separation of multidimensional Gaussian components. Simulations demonstrate the convergence properties of the suggested algorithms, as well as the dependence of the criterion on some of the model parameters.

KW - Joint block diagonalization

KW - quasi-Newton

KW - relative gradient

UR - http://www.scopus.com/inward/record.url?scp=84857282781&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-28551-6_20

DO - 10.1007/978-3-642-28551-6_20

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:84857282781

SN - 9783642285509

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 155

EP - 162

BT - Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings

T2 - 10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012

Y2 - 12 March 2012 through 15 March 2012

ER -

Lahat D, Cardoso JF, Messer H. Joint block diagonalization algorithms for optimal separation of multidimensional components. In Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. 2012. p. 155-162. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-642-28551-6_20

Joint block diagonalization algorithms for optimal separation of multidimensional components

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this