Canonical Correlation Analysis with Common Graph Priors

Jia Chen, Gang Wang, Yanning Shen, Georgios B Giannakis

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

1 Citation (Scopus)

Abstract

Canonical correlation analysis (CCA) is a well-appreciated linear subspace method to leverage hidden sources common to two or more datasets. CCA benefits are documented in various applications, such as dimensionality reduction, blind source separation, classification, and data fusion. However, the standard CCA does not exploit the geometry of common sources, which may be deduced from (cross-) correlations, or, inferred from the data. In this context, the prior information provided by the common source is encoded here through a graph, and is employed as a CCA regularizer. This leads to what is termed here as graph CCA (gCCA), which accounts for the graph-induced knowledge of common sources, while maximizing the linear correlation between the canonical variables. When the dimensionality of data vectors is high relative to the number of vectors, the dual formulation of the novel gCCA is also developed. Tests on two real datasets for facial image classification showcase the merits of the proposed approaches relative to their competing alternatives.

Original languageEnglish (US)
Title of host publication2018 IEEE Statistical Signal Processing Workshop, SSP 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages463-467
Number of pages5
ISBN (Print)9781538615706
DOIs
StatePublished - Aug 29 2018
Event20th IEEE Statistical Signal Processing Workshop, SSP 2018 - Freiburg im Breisgau, Germany
Duration: Jun 10 2018Jun 13 2018

Publication series

Name2018 IEEE Statistical Signal Processing Workshop, SSP 2018

Other

Other20th IEEE Statistical Signal Processing Workshop, SSP 2018
CountryGermany
CityFreiburg im Breisgau
Period6/10/186/13/18

Fingerprint

Blind source separation
Image classification
Data fusion
Geometry
fusion
image classification
multisensor fusion
cross correlation
formulations
geometry

Keywords

  • Canonical correlations
  • dimensionality reduction
  • generalized eigenvalue
  • signal processing over graphs

Cite this

Chen, J., Wang, G., Shen, Y., & Giannakis, G. B. (2018). Canonical Correlation Analysis with Common Graph Priors. In 2018 IEEE Statistical Signal Processing Workshop, SSP 2018 (pp. 463-467). [8450749] (2018 IEEE Statistical Signal Processing Workshop, SSP 2018). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SSP.2018.8450749

Canonical Correlation Analysis with Common Graph Priors. / Chen, Jia; Wang, Gang; Shen, Yanning; Giannakis, Georgios B.

2018 IEEE Statistical Signal Processing Workshop, SSP 2018. Institute of Electrical and Electronics Engineers Inc., 2018. p. 463-467 8450749 (2018 IEEE Statistical Signal Processing Workshop, SSP 2018).

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

Chen, J, Wang, G, Shen, Y & Giannakis, GB 2018, Canonical Correlation Analysis with Common Graph Priors. in 2018 IEEE Statistical Signal Processing Workshop, SSP 2018., 8450749, 2018 IEEE Statistical Signal Processing Workshop, SSP 2018, Institute of Electrical and Electronics Engineers Inc., pp. 463-467, 20th IEEE Statistical Signal Processing Workshop, SSP 2018, Freiburg im Breisgau, Germany, 6/10/18. https://doi.org/10.1109/SSP.2018.8450749
Chen J, Wang G, Shen Y, Giannakis GB. Canonical Correlation Analysis with Common Graph Priors. In 2018 IEEE Statistical Signal Processing Workshop, SSP 2018. Institute of Electrical and Electronics Engineers Inc. 2018. p. 463-467. 8450749. (2018 IEEE Statistical Signal Processing Workshop, SSP 2018). https://doi.org/10.1109/SSP.2018.8450749
Chen, Jia ; Wang, Gang ; Shen, Yanning ; Giannakis, Georgios B. / Canonical Correlation Analysis with Common Graph Priors. 2018 IEEE Statistical Signal Processing Workshop, SSP 2018. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 463-467 (2018 IEEE Statistical Signal Processing Workshop, SSP 2018).
@inproceedings{bd306d794d12414f89d442f05a5c9b85,
title = "Canonical Correlation Analysis with Common Graph Priors",
abstract = "Canonical correlation analysis (CCA) is a well-appreciated linear subspace method to leverage hidden sources common to two or more datasets. CCA benefits are documented in various applications, such as dimensionality reduction, blind source separation, classification, and data fusion. However, the standard CCA does not exploit the geometry of common sources, which may be deduced from (cross-) correlations, or, inferred from the data. In this context, the prior information provided by the common source is encoded here through a graph, and is employed as a CCA regularizer. This leads to what is termed here as graph CCA (gCCA), which accounts for the graph-induced knowledge of common sources, while maximizing the linear correlation between the canonical variables. When the dimensionality of data vectors is high relative to the number of vectors, the dual formulation of the novel gCCA is also developed. Tests on two real datasets for facial image classification showcase the merits of the proposed approaches relative to their competing alternatives.",
keywords = "Canonical correlations, dimensionality reduction, generalized eigenvalue, signal processing over graphs",
author = "Jia Chen and Gang Wang and Yanning Shen and Giannakis, {Georgios B}",
year = "2018",
month = "8",
day = "29",
doi = "10.1109/SSP.2018.8450749",
language = "English (US)",
isbn = "9781538615706",
series = "2018 IEEE Statistical Signal Processing Workshop, SSP 2018",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "463--467",
booktitle = "2018 IEEE Statistical Signal Processing Workshop, SSP 2018",

}

TY - GEN

T1 - Canonical Correlation Analysis with Common Graph Priors

AU - Chen, Jia

AU - Wang, Gang

AU - Shen, Yanning

AU - Giannakis, Georgios B

PY - 2018/8/29

Y1 - 2018/8/29

N2 - Canonical correlation analysis (CCA) is a well-appreciated linear subspace method to leverage hidden sources common to two or more datasets. CCA benefits are documented in various applications, such as dimensionality reduction, blind source separation, classification, and data fusion. However, the standard CCA does not exploit the geometry of common sources, which may be deduced from (cross-) correlations, or, inferred from the data. In this context, the prior information provided by the common source is encoded here through a graph, and is employed as a CCA regularizer. This leads to what is termed here as graph CCA (gCCA), which accounts for the graph-induced knowledge of common sources, while maximizing the linear correlation between the canonical variables. When the dimensionality of data vectors is high relative to the number of vectors, the dual formulation of the novel gCCA is also developed. Tests on two real datasets for facial image classification showcase the merits of the proposed approaches relative to their competing alternatives.

AB - Canonical correlation analysis (CCA) is a well-appreciated linear subspace method to leverage hidden sources common to two or more datasets. CCA benefits are documented in various applications, such as dimensionality reduction, blind source separation, classification, and data fusion. However, the standard CCA does not exploit the geometry of common sources, which may be deduced from (cross-) correlations, or, inferred from the data. In this context, the prior information provided by the common source is encoded here through a graph, and is employed as a CCA regularizer. This leads to what is termed here as graph CCA (gCCA), which accounts for the graph-induced knowledge of common sources, while maximizing the linear correlation between the canonical variables. When the dimensionality of data vectors is high relative to the number of vectors, the dual formulation of the novel gCCA is also developed. Tests on two real datasets for facial image classification showcase the merits of the proposed approaches relative to their competing alternatives.

KW - Canonical correlations

KW - dimensionality reduction

KW - generalized eigenvalue

KW - signal processing over graphs

UR - http://www.scopus.com/inward/record.url?scp=85051170977&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85051170977&partnerID=8YFLogxK

U2 - 10.1109/SSP.2018.8450749

DO - 10.1109/SSP.2018.8450749

M3 - Conference contribution

AN - SCOPUS:85051170977

SN - 9781538615706

T3 - 2018 IEEE Statistical Signal Processing Workshop, SSP 2018

SP - 463

EP - 467

BT - 2018 IEEE Statistical Signal Processing Workshop, SSP 2018

PB - Institute of Electrical and Electronics Engineers Inc.

ER -