On Sketching the $q$ to $p$ norms

TL;DR

This paper studies the problem of data dimensionality reduction for the $q o p$ norms of matrices, providing bounds on the sketching dimension needed for approximation, with applications in data mining and routing.

Contribution

It introduces the first bounds on sketching dimensions for $q o p$ norms, including tight bounds in many cases, expanding understanding of norm approximation in data sketching.

Findings

Established upper and lower bounds on sketching dimension $k$ for all $p,q o ext{infinity}$.

Provided tight bounds for specific $p,q$ norm cases.

Analyzed the impact of approximation factor $eta$ and low-rank matrices.

Abstract

We initiate the study of data dimensionality reduction, or sketching, for the $q\to p$ norms. Given an $n \times d$ matrix $A$, the $q\to p$ norm, denoted $\|A\|_{q \to p} = \sup_{x \in \mathbb{R}^d \backslash \vec{0}} \frac{\|Ax\|_p}{\|x\|_q}$, is a natural generalization of several matrix and vector norms studied in the data stream and sketching models, with applications to datamining, hardness of approximation, and oblivious routing. We say a distribution $S$ on random matrices $L \in \mathbb{R}^{nd} \rightarrow \mathbb{R}^k$ is a $(k,\alpha)$-sketching family if from $L(A)$, one can approximate $\|A\|_{q \to p}$ up to a factor $\alpha$ with constant probability. We provide upper and lower bounds on the sketching dimension $k$ for every $p, q \in [1, \infty]$, and in a number of cases our bounds are tight. While we mostly focus on constant $\alpha$, we also consider large approximation factors $\alpha$, as well as other variants of the problem such as when $A$ has low rank.