Unbiased risk estimates for singular value thresholding and spectral estimators

Emmanuel J. Candes, Carlos A. Sing-Long, Joshua D Trazasko

Research output: Contribution to journalArticle

80 Citations (Scopus)

Abstract

In an increasing number of applications, it is of interest to recover an approximately low-rank data matrix from noisy observations. This paper develops an unbiased risk estimate - holding in a Gaussian model - for any spectral estimator obeying some mild regularity assumptions. In particular, we give an unbiased risk estimate formula for singular value thresholding (SVT), a popular estimation strategy that applies a soft-thresholding rule to the singular values of the noisy observations. Among other things, our formulas offer a principled and automated way of selecting regularization parameters in a variety of problems. In particular, we demonstrate the utility of the unbiased risk estimation for SVT-based denoising of real clinical cardiac MRI series data. We also give new results concerning the differentiability of certain matrix-valued functions.

Original languageEnglish (US)
Article number6545395
Pages (from-to)4643-4657
Number of pages15
JournalIEEE Transactions on Signal Processing
Volume61
Issue number19
DOIs
StatePublished - 2013

Fingerprint

Magnetic resonance imaging

Keywords

  • Differentiability of eigenvalues and eigenvectors
  • magnetic resonance cardiac imaging
  • singular value thresholding
  • Stein's unbiased risk estimate (SURE)

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Signal Processing

Cite this

Unbiased risk estimates for singular value thresholding and spectral estimators. / Candes, Emmanuel J.; Sing-Long, Carlos A.; Trazasko, Joshua D.

In: IEEE Transactions on Signal Processing, Vol. 61, No. 19, 6545395, 2013, p. 4643-4657.

Research output: Contribution to journalArticle

Candes, Emmanuel J. ; Sing-Long, Carlos A. ; Trazasko, Joshua D. / Unbiased risk estimates for singular value thresholding and spectral estimators. In: IEEE Transactions on Signal Processing. 2013 ; Vol. 61, No. 19. pp. 4643-4657.
@article{0f43b6bccc5448f2a8ffc03071b2b368,
title = "Unbiased risk estimates for singular value thresholding and spectral estimators",
abstract = "In an increasing number of applications, it is of interest to recover an approximately low-rank data matrix from noisy observations. This paper develops an unbiased risk estimate - holding in a Gaussian model - for any spectral estimator obeying some mild regularity assumptions. In particular, we give an unbiased risk estimate formula for singular value thresholding (SVT), a popular estimation strategy that applies a soft-thresholding rule to the singular values of the noisy observations. Among other things, our formulas offer a principled and automated way of selecting regularization parameters in a variety of problems. In particular, we demonstrate the utility of the unbiased risk estimation for SVT-based denoising of real clinical cardiac MRI series data. We also give new results concerning the differentiability of certain matrix-valued functions.",
keywords = "Differentiability of eigenvalues and eigenvectors, magnetic resonance cardiac imaging, singular value thresholding, Stein's unbiased risk estimate (SURE)",
author = "Candes, {Emmanuel J.} and Sing-Long, {Carlos A.} and Trazasko, {Joshua D}",
year = "2013",
doi = "10.1109/TSP.2013.2270464",
language = "English (US)",
volume = "61",
pages = "4643--4657",
journal = "IEEE Transactions on Signal Processing",
issn = "1053-587X",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "19",

}

TY - JOUR

T1 - Unbiased risk estimates for singular value thresholding and spectral estimators

AU - Candes, Emmanuel J.

AU - Sing-Long, Carlos A.

AU - Trazasko, Joshua D

PY - 2013

Y1 - 2013

N2 - In an increasing number of applications, it is of interest to recover an approximately low-rank data matrix from noisy observations. This paper develops an unbiased risk estimate - holding in a Gaussian model - for any spectral estimator obeying some mild regularity assumptions. In particular, we give an unbiased risk estimate formula for singular value thresholding (SVT), a popular estimation strategy that applies a soft-thresholding rule to the singular values of the noisy observations. Among other things, our formulas offer a principled and automated way of selecting regularization parameters in a variety of problems. In particular, we demonstrate the utility of the unbiased risk estimation for SVT-based denoising of real clinical cardiac MRI series data. We also give new results concerning the differentiability of certain matrix-valued functions.

AB - In an increasing number of applications, it is of interest to recover an approximately low-rank data matrix from noisy observations. This paper develops an unbiased risk estimate - holding in a Gaussian model - for any spectral estimator obeying some mild regularity assumptions. In particular, we give an unbiased risk estimate formula for singular value thresholding (SVT), a popular estimation strategy that applies a soft-thresholding rule to the singular values of the noisy observations. Among other things, our formulas offer a principled and automated way of selecting regularization parameters in a variety of problems. In particular, we demonstrate the utility of the unbiased risk estimation for SVT-based denoising of real clinical cardiac MRI series data. We also give new results concerning the differentiability of certain matrix-valued functions.

KW - Differentiability of eigenvalues and eigenvectors

KW - magnetic resonance cardiac imaging

KW - singular value thresholding

KW - Stein's unbiased risk estimate (SURE)

UR - http://www.scopus.com/inward/record.url?scp=84883334291&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84883334291&partnerID=8YFLogxK

U2 - 10.1109/TSP.2013.2270464

DO - 10.1109/TSP.2013.2270464

M3 - Article

AN - SCOPUS:84883334291

VL - 61

SP - 4643

EP - 4657

JO - IEEE Transactions on Signal Processing

JF - IEEE Transactions on Signal Processing

SN - 1053-587X

IS - 19

M1 - 6545395

ER -