The Johnson-Lindenstrauss lemma is optimal for linear dimensionality reduction

TL;DR

This paper proves that the Johnson-Lindenstrauss lemma's bounds for linear dimensionality reduction are optimal, establishing matching lower bounds that confirm the best possible dimension reduction in terms of distortion and size.

Contribution

It provides a new lower bound matching the Johnson-Lindenstrauss upper bounds, confirming the optimality of the lemma for linear maps in dimensionality reduction.

Findings

Lower bound matches Johnson-Lindenstrauss upper bounds

Optimality of linear dimensionality reduction bounds

Improves previous bounds by a logarithmic factor

Abstract

For any $n>1$ and $0<\varepsilon<1/2$, we show the existence of an $n^{O(1)}$-point subset $X$ of $\mathbb{R}^n$ such that any linear map from $(X,\ell_2)$ to $\ell_2^m$ with distortion at most $1+\varepsilon$ must have $m = \Omega(\min\{n, \varepsilon^{-2}\log n\})$. Our lower bound matches the upper bounds provided by the identity matrix and the Johnson-Lindenstrauss lemma, improving the previous lower bound of Alon by a $\log(1/\varepsilon)$ factor.