pith. sign in

arxiv: 1702.00139 · v1 · pith:BRDY7HIOnew · submitted 2017-02-01 · 🧮 math.PR · math.NA

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

classification 🧮 math.PR math.NA
keywords lambdatheorydingerperturbationrayleigh-schreigenvalueseigenvectorsequation
0
0 comments X
read the original 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, $A^t = A + tE$, $\{\lambda_j\}_{j=1}^n$ are eigenvalues of $A$ in descending order, and $u_1, u^t_1$ are leading eigenvectors of $A$ and $A^t$. Rayleigh-Schr\"{o}dinger theory shows asymptotically, $\langle u^t_1, u_j \rangle \propto t / (\lambda_1 - \lambda_j)$ where $ t = o(1)$. However, the asymptotic theory does not apply to larger $t$; in particular, it fails when $ t \| E \|_2 > \lambda_1 - \lambda_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, \begin{equation*} | \langle u^1_1, u_j \rangle | = 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 $\lambda_1 - \lambda_j$. This bound is valid, even in cases where $\| E \|_2 \gg \lambda_1 - \lambda_2$. The result is optimal, except for a log term. It also leads to an improvement of Davis-Kahan theorem.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.