Quantifying the estimation error of principal component vectors

DOI10.1093/IMAIAI/IAZ014MaRDI QIDQ5006534zbMATH OpenOpenAlexFDO

Authors Raphael A. Hauser, J. Lember, Raul Kangro, Heinrich Matzinger

Publication date 16 August 2021

Published in Information and Inference: A Journal of the IMA (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1710.10124, https://ora.ox.ac.uk/objects/uuid:3679a50c-07ab-4d5e-ae1e-325e0901a4e9

zbMATH Keywords

eigenvector approximationprincipal component analysissample complexity

Mathematics Subject Classification ID

Inequalities involving eigenvalues and eigenvectors (15A42) Inequalities; stochastic orderings (60E15) Gaussian processes (60G15) Estimation in multivariate analysis (62H12) Factor analysis and principal components; correspondence analysis (62H25)

Abstract: Principal component analysis is an important pattern recognition and dimensionality reduction tool in many applications. Principal components are computed as eigenvectors of a maximum likelihood covariance $widehatSigma$ that approximates a population covariance $Sigma$, and these eigenvectors are often used to extract structural information about the variables (or attributes) of the studied population. Since PCA is based on the eigendecomposition of the proxy covariance $widehatSigma$ rather than the ground-truth $Sigma$, it is important to understand the approximation error in each individual eigenvector as a function of the number of available samples. The recent results of Kolchinskii and Lounici yield such bounds. In the present paper we sharpen these bounds and show that eigenvectors can often be reconstructed to a required accuracy from a sample of strictly smaller size order.

Recommendations

• Finite sample approximation results for principal component analysis: A matrix perturbation approach
• Nonasymptotic upper bounds for the reconstruction error of PCA
• On consistency and sparsity for principal components analysis in high dimensions
• High-probability bounds for the reconstruction error of PCA
• High-dimensional principal projections

Cited in

(8)

• Finite sample approximation results for principal component analysis: A matrix perturbation approach
• A correction of approximations used in sensitivity study of principal component analysis
• Nonasymptotic upper bounds for the reconstruction error of PCA
• Convergence rate of Krasulina estimator
• High-probability bounds for the reconstruction error of PCA
• Principal component analysis with drop rank covariance matrix
• A New Measure of Relative Error for Vectors
• Error bounds for suboptimal solutions to kernel principal component analysis

This page was built for publication: Quantifying the estimation error of principal component vectors

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5006534)