Stationary subspace analysis as a generalized eigenvalue problem

Satoshi Hara; Yoshinobu Kawahara; Takashi Washio; Paul Von Bünau

doi:10.1007/978-3-642-17537-4_52

Stationary subspace analysis as a generalized eigenvalue problem

Satoshi Hara, Yoshinobu Kawahara, Takashi Washio, Paul Von Bünau

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

13 Citations (Scopus)

Abstract

Understanding non-stationary effects is one of the key challenges in data analysis. However, in many settings the observation is a mixture of stationary and non-stationary sources. The aim of Stationary Subspace Analysis (SSA) is to factorize multivariate data into its stationary and non-stationary components. In this paper, we propose a novel SSA algorithm (ASSA) that extracts stationary sources from multiple time series blocks. It has a globally optimal solution under certain assumptions that can be obtained by solving a generalized eigenvalue problem. Apart from the numerical advantages, we also show that compared to the existing method, fewer blocks are required in ASSA to guarantee the identifiability of the solution. We demonstrate the validity of our approach in simulations and in an application to domain adaptation.

Original language	English
Title of host publication	Neural Information Processing
Subtitle of host publication	Theory and Algorithms - 17th International Conference, ICONIP 2010, Proceedings
Pages	422-429
Number of pages	8
Edition	PART 1
DOIs	https://doi.org/10.1007/978-3-642-17537-4_52
Publication status	Published - 2010
Externally published	Yes
Event	17th International Conference on Neural Information Processing, ICONIP 2010 - Sydney, NSW, Australia Duration: Nov 22 2010 → Nov 25 2010

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number	PART 1
Volume	6443 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	17th International Conference on Neural Information Processing, ICONIP 2010
Country	Australia
City	Sydney, NSW
Period	11/22/10 → 11/25/10

All Science Journal Classification (ASJC) codes

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-642-17537-4_52

Fingerprint Dive into the research topics of 'Stationary subspace analysis as a generalized eigenvalue problem'. Together they form a unique fingerprint.

Cite this

Hara, S., Kawahara, Y., Washio, T., & Von Bünau, P. (2010). Stationary subspace analysis as a generalized eigenvalue problem. In Neural Information Processing: Theory and Algorithms - 17th International Conference, ICONIP 2010, Proceedings (PART 1 ed., pp. 422-429). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6443 LNCS, No. PART 1). https://doi.org/10.1007/978-3-642-17537-4_52

Stationary subspace analysis as a generalized eigenvalue problem. / Hara, Satoshi; Kawahara, Yoshinobu; Washio, Takashi; Von Bünau, Paul.

Neural Information Processing: Theory and Algorithms - 17th International Conference, ICONIP 2010, Proceedings. PART 1. ed. 2010. p. 422-429 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6443 LNCS, No. PART 1).

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

Hara, S, Kawahara, Y, Washio, T & Von Bünau, P 2010, Stationary subspace analysis as a generalized eigenvalue problem. in Neural Information Processing: Theory and Algorithms - 17th International Conference, ICONIP 2010, Proceedings. PART 1 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 6443 LNCS, pp. 422-429, 17th International Conference on Neural Information Processing, ICONIP 2010, Sydney, NSW, Australia, 11/22/10. https://doi.org/10.1007/978-3-642-17537-4_52

Hara S, Kawahara Y, Washio T, Von Bünau P. Stationary subspace analysis as a generalized eigenvalue problem. In Neural Information Processing: Theory and Algorithms - 17th International Conference, ICONIP 2010, Proceedings. PART 1 ed. 2010. p. 422-429. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). https://doi.org/10.1007/978-3-642-17537-4_52

Hara, Satoshi ; Kawahara, Yoshinobu ; Washio, Takashi ; Von Bünau, Paul. / Stationary subspace analysis as a generalized eigenvalue problem. Neural Information Processing: Theory and Algorithms - 17th International Conference, ICONIP 2010, Proceedings. PART 1. ed. 2010. pp. 422-429 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).

@inproceedings{05a3f82422f54a41912fa2511b5fb1cc,

title = "Stationary subspace analysis as a generalized eigenvalue problem",

abstract = "Understanding non-stationary effects is one of the key challenges in data analysis. However, in many settings the observation is a mixture of stationary and non-stationary sources. The aim of Stationary Subspace Analysis (SSA) is to factorize multivariate data into its stationary and non-stationary components. In this paper, we propose a novel SSA algorithm (ASSA) that extracts stationary sources from multiple time series blocks. It has a globally optimal solution under certain assumptions that can be obtained by solving a generalized eigenvalue problem. Apart from the numerical advantages, we also show that compared to the existing method, fewer blocks are required in ASSA to guarantee the identifiability of the solution. We demonstrate the validity of our approach in simulations and in an application to domain adaptation.",

author = "Satoshi Hara and Yoshinobu Kawahara and Takashi Washio and {Von B{\"u}nau}, Paul",

year = "2010",

doi = "10.1007/978-3-642-17537-4_52",

language = "English",

isbn = "3642175368",

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

number = "PART 1",

pages = "422--429",

booktitle = "Neural Information Processing",

edition = "PART 1",

note = "17th International Conference on Neural Information Processing, ICONIP 2010 ; Conference date: 22-11-2010 Through 25-11-2010",

}

TY - GEN

T1 - Stationary subspace analysis as a generalized eigenvalue problem

AU - Hara, Satoshi

AU - Kawahara, Yoshinobu

AU - Washio, Takashi

AU - Von Bünau, Paul

PY - 2010

Y1 - 2010

N2 - Understanding non-stationary effects is one of the key challenges in data analysis. However, in many settings the observation is a mixture of stationary and non-stationary sources. The aim of Stationary Subspace Analysis (SSA) is to factorize multivariate data into its stationary and non-stationary components. In this paper, we propose a novel SSA algorithm (ASSA) that extracts stationary sources from multiple time series blocks. It has a globally optimal solution under certain assumptions that can be obtained by solving a generalized eigenvalue problem. Apart from the numerical advantages, we also show that compared to the existing method, fewer blocks are required in ASSA to guarantee the identifiability of the solution. We demonstrate the validity of our approach in simulations and in an application to domain adaptation.

AB - Understanding non-stationary effects is one of the key challenges in data analysis. However, in many settings the observation is a mixture of stationary and non-stationary sources. The aim of Stationary Subspace Analysis (SSA) is to factorize multivariate data into its stationary and non-stationary components. In this paper, we propose a novel SSA algorithm (ASSA) that extracts stationary sources from multiple time series blocks. It has a globally optimal solution under certain assumptions that can be obtained by solving a generalized eigenvalue problem. Apart from the numerical advantages, we also show that compared to the existing method, fewer blocks are required in ASSA to guarantee the identifiability of the solution. We demonstrate the validity of our approach in simulations and in an application to domain adaptation.

UR - http://www.scopus.com/inward/record.url?scp=78650224631&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78650224631&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-17537-4_52

DO - 10.1007/978-3-642-17537-4_52

M3 - Conference contribution

AN - SCOPUS:78650224631

SN - 3642175368

SN - 9783642175367

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

SP - 422

EP - 429

BT - Neural Information Processing

T2 - 17th International Conference on Neural Information Processing, ICONIP 2010

Y2 - 22 November 2010 through 25 November 2010

ER -