Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Distributions of singular values for some random matrices.

A M Sengupta1, P P Mitra

  • 1Bell Laboratories, Lucent Technologies, 600 Mountain Avenue, Murray Hill, New Jersey 07974, USA.

Physical Review. E, Statistical Physics, Plasmas, Fluids, and Related Interdisciplinary Topics
|April 24, 2002
PubMed
Summary

Singular value decomposition (SVD) analysis of noisy, low-rank matrices is extended to complex noise types. This research provides error bar estimates for reconstructing matrices using truncated SVD.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Laboratory evolution of adenylyl cyclase independent learning in Drosophila and missing heritability.

Genes, brain, and behavior·2014
Same author

High-resolution zero-dispersion wavelength mapping in single-mode fiber.

Optics letters·2007
Same author

Non-additivity in protein-DNA binding.

Bioinformatics (Oxford, England)·2005
Same author

Studying the song development process: rationale and methods.

Annals of the New York Academy of Sciences·2004
Same author

Song development: in search of the error-signal.

Annals of the New York Academy of Sciences·2004
Same author

Effect of chromatic dispersion on nonlinear phase noise.

Optics letters·2003

Area of Science:

  • Multivariate data analysis
  • Linear algebra
  • Signal processing

Background:

  • Singular Value Decomposition (SVD) is crucial for analyzing complex multivariate data, including space-time images in physical and biological sciences.
  • Previous research on SVD in noisy matrices focused on uniform, uncorrelated noise.
  • Understanding SVD behavior under diverse noise conditions is essential for accurate data interpretation.

Purpose of the Study:

  • To investigate the distribution of singular values in low-rank matrices affected by additive noise.
  • To extend existing SVD analysis beyond uniform noise to heterogeneous and correlated noise sources.
  • To develop methods for estimating errors in low-rank matrix reconstruction using truncated SVD.

Main Methods:

  • Application of diagrammatic techniques to analyze singular value distributions.

Related Experiment Videos

  • Utilizing saddle point integration for mathematical analysis of noise effects.
  • Developing perturbative estimation methods for error bars.

    • Main Results:

    • The study successfully extends the analysis of singular value distributions to heterogeneous and correlated noise.
    • New insights into the behavior of singular values under complex noise conditions are provided.
    • Perturbative estimates for error bars on reconstructed low-rank matrices are derived.

    Conclusions:

    • The findings enhance the robustness of SVD analysis for multivariate data corrupted by complex noise.
    • This work offers improved accuracy in reconstructing low-rank matrices from noisy data.
    • The developed methods have implications for various fields utilizing SVD, such as image processing and systems analysis.