ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 1404.0788
36
151

On the principal components of sample covariance matrices

3 April 2014
Alex Bloemendal
Antti Knowles
H. Yau
J. Yin
ArXivPDFHTML
Abstract

We introduce a class of M×MM \times MM×M sample covariance matrices Q\mathcal QQ which subsumes and generalizes several previous models. The associated population covariance matrix Σ=EQ\Sigma = \mathbb E \cal QΣ=EQ is assumed to differ from the identity by a matrix of bounded rank. All quantities except the rank of Σ−IM\Sigma - I_MΣ−IM​ may depend on MMM in an arbitrary fashion. We investigate the principal components, i.e.\ the top eigenvalues and eigenvectors, of Q\mathcal QQ. We derive precise large deviation estimates on the generalized components ⟨w,ξi⟩\langle \mathbf w, \boldsymbol \xi_i \rangle⟨w,ξi​⟩ of the outlier and non-outlier eigenvectors ξi\boldsymbol \xi_iξi​. Our results also hold near the so-called BBP transition, where outliers are created or annihilated, and for degenerate or near-degenerate outliers. We believe the obtained rates of convergence to be optimal. In addition, we derive the asymptotic distribution of the generalized components of the non-outlier eigenvectors. A novel observation arising from our results is that, unlike the eigenvalues, the eigenvectors of the principal components contain information about the \emph{subcritical} spikes of Σ\SigmaΣ. The proofs use several results on the eigenvalues and eigenvectors of the uncorrelated matrix Q\mathcal QQ, satisfying EQ=IM\mathbb E \mathcal Q = I_MEQ=IM​, as input: the isotropic local Marchenko-Pastur law established in [9], level repulsion, and quantum unique ergodicity of the eigenvectors. The latter is a special case of a new universality result for the joint eigenvalue-eigenvector distribution.

View on arXiv
Comments on this paper