Effective recovery of Fourier spectra and spectral approximation by finite groups

TL;DR

This paper establishes high-probability methods for recovering subgroups of finite nonabelian groups from noisy data, extending Fourier recovery techniques from Euclidean spaces to nonabelian and nilpotent Lie groups.

Contribution

It introduces a novel approach for approximate subgroup recovery in nonabelian groups, generalizing Fourier recovery results to polynomial growth groups and nilpotent Lie groups.

Findings

High-probability subgroup recovery from random perturbations

Extension of Fourier recovery to nonabelian and nilpotent groups

Effective convergence rates for spectral approximation

Abstract

We prove a result on approximate recovery, with high probability, of subgroups of a finite nonabelian group $\Gamma$ from their random perturbations. We use this for ad-hoc sequences of $\Gamma_n$ while passing to the continuum limit, in order to obtain asymptotic almost sure recovery for rational lcsc nilpotent Lie groups. By comparison to limit theorems for groups of polynomial growth, it turns out that this setting is the natural general setting for recovery results, under polynomial growth assumptions on the $\Gamma_n$. This approach makes effective the convergence rate in previous Fourier recovery theorems in Euclidean space, and extends them to the nonabelian setting. A series of interesting further directions are highlighted by this approach.