Spectral alignment of correlated Gaussian random matrices - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Journal Articles Advances in Applied Probability Year : 2022

Spectral alignment of correlated Gaussian random matrices


In this paper we analyze a simple method ($EIG1$) for the problem of matrix alignment, consisting in aligning their leading eigenvectors: given $A$ and $B$, we compute $v_1$ and $v'_1$ two leading eigenvectors of $A$ and $B$. The algorithm returns a permutation $\hat{\Pi}$ such that the rank of the coordinate $\hat{\Pi}(i)$ in $v_1$ is the rank of the coordinate $i$ in $v'_1$ (up to the sign of $v'_1$). We consider a model where $A$ belongs to the Gaussian Orthogonal Ensemble (GOE), and $B= \Pi^T (A+\sigma H) \Pi $, where $\Pi$ is a permutation matrix and $H$ is an independent copy of $A$. We show the following 0-1 law: under the condition $\sigma N^{7/6+\epsilon} \to 0$, the $EIG1$ method recovers all but a vanishing part of the underlying permutation $\Pi$. When $\sigma N^{7/6-\epsilon} \to \infty$, this algorithm cannot recover more than $o(N)$ correct matches. This result gives an understanding of the simplest and fastest spectral method for matrix alignment (or complete weighted graph alignment), and involves proof methods and techniques which could be of independent interest.
Fichier principal
Vignette du fichier
Spectral_alignment_of_correlated_Gaussian_matrices.pdf (562.12 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-02941069 , version 1 (16-09-2020)



Luca Ganassali, Marc Lelarge, Laurent Massoulié. Spectral alignment of correlated Gaussian random matrices. Advances in Applied Probability, 2022, 54, Issue 1, 54 (Issue 1), pp.279 - 310. ⟨hal-02941069⟩
62 View
89 Download



Gmail Facebook Twitter LinkedIn More