Testing noisy linear functions for sparsity

TL;DR

This paper investigates the problem of efficiently testing whether a high-dimensional vector is sparse or far from sparse using a small, dimension-independent number of samples, introducing a noise-tolerant algorithm for non-Gaussian distributions.

Contribution

The authors present a novel, noise-tolerant property testing algorithm that determines sparsity with a number of samples independent of the dimension for non-Gaussian distributions.

Findings

Efficient sparsity testing is possible for non-Gaussian distributions with dimension-independent samples.

The proposed algorithm is noise tolerant and approximates the distance to the closest sparse vector.

It is shown that fewer than logarithmic samples are insufficient for the problem under certain conditions.

Abstract

We consider the following basic inference problem: there is an unknown high-dimensional vector $w \in \mathbb{R}^n$, and an algorithm is given access to labeled pairs $(x,y)$ where $x \in \mathbb{R}^n$ is a measurement and $y = w \cdot x + \mathrm{noise}$. What is the complexity of deciding whether the target vector $w$ is (approximately) $k$-sparse? The recovery analogue of this problem --- given the promise that $w$ is sparse, find or approximate the vector $w$ --- is the famous sparse recovery problem, with a rich body of work in signal processing, statistics, and computer science. We study the decision version of this problem (i.e.~deciding whether the unknown $w$ is $k$-sparse) from the vantage point of property testing. Our focus is on answering the following high-level question: when is it possible to efficiently test whether the unknown target vector $w$ is sparse versus far-from-sparse using a number of samples which is completely independent of the dimension $n$? We consider the natural setting in which $x$ is drawn from a i.i.d.~product distribution $\mathcal{D}$ over $\mathbb{R}^n$ and the $\mathrm{noise}$ process is independent of the input $x$. As our main result, we give a general algorithm which solves the above-described testing problem using a number of samples which is completely independent of the ambient dimension $n$, as long as $\mathcal{D}$ is not a Gaussian. In fact, our algorithm is fully noise tolerant, in the sense that for an arbitrary $w$, it approximately computes the distance of $w$ to the closest $k$-sparse vector. To complement this algorithmic result, we show that weakening any of our condition makes it information-theoretically impossible for any algorithm to solve the testing problem with fewer than essentially $\log n$ samples.