Sparse Non-linear CCA through Hilbert-Schmidt Independence Criterion

Viivi Uurtio; Sahely Bhadra; Juho Rousu

doi:10.1109/ICDM.2018.00172

Sparse Non-linear CCA through Hilbert-Schmidt Independence Criterion

Viivi Uurtio, Sahely Bhadra, Juho Rousu

Indian Institute of Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

7 Citations (Scopus)

255 Downloads (Pure)

Abstract

We present SCCA-HSIC, a method for finding sparse non-linear multivariate relations in high-dimensional settings by maximizing the Hilbert-Schmidt Independence Criterion (HSIC). We propose efficient optimization algorithms using a projected stochastic gradient and Nyström approximation of HSIC. We demonstrate the favourable performance of SCCA-HSIC over competing methods in detecting multivariate non-linear relations both in simulation studies, with varying numbers of related variables, noise variables, and samples, as well as in real datasets.

Original language	English
Title of host publication	2018 IEEE International Conference on Data Mining, ICDM 2018
Publisher	IEEE
Pages	1278-1283
Number of pages	6
ISBN (Electronic)	9781538691588
ISBN (Print)	9781538691595
DOIs	https://doi.org/10.1109/ICDM.2018.00172
Publication status	Published - 2018
MoE publication type	A4 Conference publication
Event	IEEE International Conference on Data Mining - Singapore, Singapore Duration: 17 Nov 2018 → 20 Nov 2018

Conference

Conference	IEEE International Conference on Data Mining
Abbreviated title	ICDM
Country/Territory	Singapore
City	Singapore
Period	17/11/2018 → 20/11/2018

Keywords

Canonical correlation
Dimensionality reduction
Hilbert-schmidt independence criterion
Kernel methods
Sparsity

Access to Document

10.1109/ICDM.2018.00172

SCI_Uurtio_et.al._Sparse_Non-Linear_CCA_PID5545933Accepted author manuscript, 1.61 MB

Cite this

@inproceedings{2e738235f0cc4a71864243613e0292b0,

title = "Sparse Non-linear CCA through Hilbert-Schmidt Independence Criterion",

abstract = "We present SCCA-HSIC, a method for finding sparse non-linear multivariate relations in high-dimensional settings by maximizing the Hilbert-Schmidt Independence Criterion (HSIC). We propose efficient optimization algorithms using a projected stochastic gradient and Nystr{\"o}m approximation of HSIC. We demonstrate the favourable performance of SCCA-HSIC over competing methods in detecting multivariate non-linear relations both in simulation studies, with varying numbers of related variables, noise variables, and samples, as well as in real datasets.",

keywords = "Canonical correlation, Dimensionality reduction, Hilbert-schmidt independence criterion, Kernel methods, Sparsity",

author = "Viivi Uurtio and Sahely Bhadra and Juho Rousu",

year = "2018",

doi = "10.1109/ICDM.2018.00172",

language = "English",

isbn = "9781538691595",

pages = "1278--1283",

booktitle = "2018 IEEE International Conference on Data Mining, ICDM 2018",

publisher = "IEEE",

address = "United States",

note = "IEEE International Conference on Data Mining, ICDM ; Conference date: 17-11-2018 Through 20-11-2018",

}

TY - GEN

T1 - Sparse Non-linear CCA through Hilbert-Schmidt Independence Criterion

AU - Uurtio, Viivi

AU - Bhadra, Sahely

AU - Rousu, Juho

PY - 2018

Y1 - 2018

N2 - We present SCCA-HSIC, a method for finding sparse non-linear multivariate relations in high-dimensional settings by maximizing the Hilbert-Schmidt Independence Criterion (HSIC). We propose efficient optimization algorithms using a projected stochastic gradient and Nyström approximation of HSIC. We demonstrate the favourable performance of SCCA-HSIC over competing methods in detecting multivariate non-linear relations both in simulation studies, with varying numbers of related variables, noise variables, and samples, as well as in real datasets.

AB - We present SCCA-HSIC, a method for finding sparse non-linear multivariate relations in high-dimensional settings by maximizing the Hilbert-Schmidt Independence Criterion (HSIC). We propose efficient optimization algorithms using a projected stochastic gradient and Nyström approximation of HSIC. We demonstrate the favourable performance of SCCA-HSIC over competing methods in detecting multivariate non-linear relations both in simulation studies, with varying numbers of related variables, noise variables, and samples, as well as in real datasets.

KW - Canonical correlation

KW - Dimensionality reduction

KW - Hilbert-schmidt independence criterion

KW - Kernel methods

KW - Sparsity

U2 - 10.1109/ICDM.2018.00172

DO - 10.1109/ICDM.2018.00172

M3 - Conference article in proceedings

SN - 9781538691595

SP - 1278

EP - 1283

BT - 2018 IEEE International Conference on Data Mining, ICDM 2018

PB - IEEE

T2 - IEEE International Conference on Data Mining

Y2 - 17 November 2018 through 20 November 2018

ER -

Sparse Non-linear CCA through Hilbert-Schmidt Independence Criterion

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this