Stable recovery and the coordinate small-ball behaviour of random vectors

TL;DR

This paper introduces the coordinate small-ball condition to analyze stable recovery in data science, demonstrating that many random vectors exhibit stable point separation even under minimal assumptions, with applications to sub-sampled convolutions.

Contribution

The paper develops a new coordinate small-ball framework for understanding stable point separation in general sampling methods, extending beyond iid subgaussian vectors.

Findings

Many coordinates of the transformed vector are significantly large.

Stable point separation holds under minimal assumptions on the random vector.

Random sub-sampled convolutions satisfy stable point separation without restrictive conditions.

Abstract

Recovery procedures in various application in Data Science are based on \emph{stable point separation}. In its simplest form, stable point separation implies that if $f$ is "far away" from $0$, and one is given a random sample $(f(Z_i))_{i=1}^m$ where a proportional number of the sample points may be corrupted by noise, that information is still enough to exhibit that $f$ is far from $0$. Stable point separation is well understood in the context of iid sampling, and to explore it for general sampling methods we introduce a new notion---the \emph{coordinate small-ball} of a random vector $X$. Roughly put, this feature captures the number of "relatively large coordinates" of $(|<TX,u_i>|)_{i=1}^m$, where $T:\mathbb{R}^n \to \mathbb{R}^m$ is an arbitrary linear operator and $(u_i)_{i=1}^m$ is any fixed orthonormal basis of $\mathbb{R}^m$. We show that under the bare-minimum assumptions on $X$, and with high probability, many of the values $|<TX,u_i>|$ are at least of the order $\|T\|_{S_2}/\sqrt{m}$. As a result, the "coordinate structure" of $TX$ exhibits the typical Euclidean norm of $TX$ and does so in a stable way. One outcome of our analysis is that random sub-sampled convolutions satisfy stable point separation under minimal assumptions on the generating random vector---a fact that was known previously only in a highly restrictive setup, namely, for random vectors with iid subgaussian coordinates.