A Novel and Optimal Spectral Method for Permutation Synchronization

Authors Duc Manh Nguyen, Anderson Ye Zhang

Publication date 21 March 2023

Abstract: Permutation synchronization is an important problem in computer science that constitutes the key step of many computer vision tasks. The goal is to recover

n

latent permutations from their noisy and incomplete pairwise measurements. In recent years, spectral methods have gained increasing popularity thanks to their simplicity and computational efficiency. Spectral methods utilize the leading eigenspace

U

of the data matrix and its block submatrices

U_{1}, U_{2}, l d o t s, U_{n}

to recover the permutations. In this paper, we propose a novel and statistically optimal spectral algorithm. Unlike the existing methods which use

{U_{j} U_{1}^{o} p}_{j g e q 2}

, ours constructs an anchor matrix

M

by aggregating useful information from all the block submatrices and estimates the latent permutations through

{U_{j} M^{o} p}_{j g e q 1}

. This modification overcomes a crucial limitation of the existing methods caused by the repetitive use of

U_{1}

and leads to an improved numerical performance. To establish the optimality of the proposed method, we carry out a fine-grained spectral analysis and obtain a sharp exponential error bound that matches the minimax rate.

This page was built for publication: A Novel and Optimal Spectral Method for Permutation Synchronization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6430384)