On the overlaps between eigenvectors of correlated random matrices

Abstract

We obtain general, exact formulas for the overlaps between the eigenvectors of large correlated random matrices, with additive or multiplicative noise. These results have potential applications in many different contexts, from quantum thermalisation to high dimensional statistics. We find that the overlaps only depend on measurable quantities, and do not require the knowledge of the underlying "true" (noiseless) matrices. We apply our results to the case of empirical correlation matrices, that allow us to estimate reliably the width of the spectrum of the true correlation matrix, even when the latter is very close to the identity. We illustrate our results on the example of stock returns correlations, that clearly reveal a non trivial structure for the bulk eigenvalues. We also apply our results to the problem of matrix denoising in high dimension.

12 pages, 5 figures

References in corpus (7)

• Many body localization and thermalization in quantum statistical mechanics
• Cleaning large correlation matrices: tools from random matrix theory
• Rotational invariant estimator for general noisy matrices
• Eigenvector dynamics: general theory and some applications
• Estimating the principal components of correlation matrices from all their empirical eigenvectors
• Optimal design of experiments by combining coarse and fine measurements
• Thermalisation of a quantum system from first principles

Cited by in corpus (9)

• Eigenvectors distribution and quantum unique ergodicity for deformed Wigner matrices
• Spectra of large time-lagged correlation matrices from Random Matrix Theory
• Top Eigenpair Statistics for Weighted Sparse Graphs
• Two-step estimators of high dimensional correlation matrices
• Interlacing Eigenvectors of Large Gaussian Matrices
• Functional Renormalization Group Approach for Signal Detection
• The High-Dimensional Asymptotics of Principal Component Regression
• Overlaps between eigenvectors of spiked, correlated random matrices: from matrix PCA to random Gaussian landscapes
• A Consistently Oriented Basis for Eigenanalysis: Improved Directional Statistics