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. 1408.4643
40
90

Asymptotics and Concentration Bounds for Bilinear Forms of Spectral Projectors of Sample Covariance

20 August 2014
V. Koltchinskii
Karim Lounici
ArXivPDFHTML
Abstract

Let X,X1,…,XnX,X_1,\dots, X_nX,X1​,…,Xn​ be i.i.d. Gaussian random variables with zero mean and covariance operator Σ=E(X⊗X)\Sigma={\mathbb E}(X\otimes X)Σ=E(X⊗X) taking values in a separable Hilbert space H.{\mathbb H}.H. Let {\bf r}(\Sigma):=\frac{{\rm tr}(\Sigma)}{\|\Sigma\|_{\infty}} be the effective rank of Σ,\Sigma,Σ, tr(Σ){\rm tr}(\Sigma)tr(Σ) being the trace of Σ\SigmaΣ and ∥Σ∥∞\|\Sigma\|_{\infty}∥Σ∥∞​ being its operator norm. Let \hat \Sigma_n:=n^{-1}\sum_{j=1}^n (X_j\otimes X_j) be the sample (empirical) covariance operator based on (X1,…,Xn).(X_1,\dots, X_n).(X1​,…,Xn​). The paper deals with a problem of estimation of spectral projectors of the covariance operator Σ\SigmaΣ by their empirical counterparts, the spectral projectors of Σ^n\hat \Sigma_nΣ^n​ (empirical spectral projectors). The focus is on the problems where both the sample size nnn and the effective rank r(Σ){\bf r}(\Sigma)r(Σ) are large. This framework includes and generalizes well known high-dimensional spiked covariance models. Given a spectral projector PrP_rPr​ corresponding to an eigenvalue μr\mu_rμr​ of covariance operator Σ\SigmaΣ and its empirical counterpart P^r,\hat P_r,P^r​, we derive sharp concentration bounds for bilinear forms of empirical spectral projector P^r\hat P_rP^r​ in terms of sample size nnn and effective dimension r(Σ).{\bf r}(\Sigma).r(Σ). Building upon these concentration bounds, we prove the asymptotic normality of bilinear forms of random operators P^r−EP^r\hat P_r -{\mathbb E}\hat P_rP^r​−EP^r​ under the assumptions that n→∞n\to \inftyn→∞ and r(Σ)=o(n).{\bf r}(\Sigma)=o(n).r(Σ)=o(n). In a special case of eigenvalues of multiplicity one, these results are rephrased as concentration bounds and asymptotic normality for linear forms of empirical eigenvectors. Other results include bounds on the bias EP^r−Pr{\mathbb E}\hat P_r-P_rEP^r​−Pr​ and a method of bias reduction as well as a discussion of possible applications to statistical inference in high-dimensional principal component analysis.

View on arXiv
Comments on this paper