Eigenvector Under Random Perturbation: A Nonasymptotic Rayleigh-Schr\"{o}dinger Theory

arXiv1702.00139MaRDI QIDQ6282686FDOQ6282686

Publication date: 1 February 2017

Abstract: Rayleigh-Schr"{o}dinger perturbation theory is a well-known theory in quantum mechanics and it offers useful characterization of eigenvectors of a perturbed matrix. Suppose

A

and perturbation

E

are both Hermitian matrices,

A^{t} = A + t E

,

{l a m b d a_{j}}_{j = 1}^{n}

are eigenvalues of

A

in descending order, and

u_{1}, u_{1}^{t}

are leading eigenvectors of

A

and

A^{t}

. Rayleigh-Schr"{o}dinger theory shows asymptotically,

l a n g l e u_{1}^{t}, u_{j} a n g l e p r o p t o t / (l a m b d a_{1} - l a m b d a_{j})

where

t = o (1)

. However, the asymptotic theory does not apply to larger

t

; in particular, it fails when

t | E |_{2} > l a m b d a_{1} - l a m b d a_{2}

. In this paper, we present a nonasymptotic theory with

E

being a random matrix. We prove that, when

t = 1

and

E

has independent and centered subgaussian entries above its diagonal, with high probability, �egin{equation*} | langle u^1_1, u_j angle | = O(sqrt{log n} / (lambda_1 - lambda_j)), end{equation*} for all

j > 1

simultaneously, under a condition on eigenvalues of

A

that involves all gaps

l a m b d a_{1} - l a m b d a_{j}

. This bound is valid, even in cases where

| E |_{2} g g l a m b d a_{1} - l a m b d a_{2}

. The result is optimal, except for a log term. It also leads to an improvement of Davis-Kahan theorem.

This page was built for publication: Eigenvector Under Random Perturbation: A Nonasymptotic Rayleigh-Schr\"{o}dinger Theory

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6282686)