Dimension Expanders via Rank Condensers

TL;DR

This paper develops new algebraic objects called rank condensers and uses them to construct explicit dimension expanders and two-source condensers, advancing the understanding of linear-algebraic pseudorandomness.

Contribution

It introduces and constructs rank condensers and uses them to build explicit dimension expanders and two-source condensers, simplifying previous complex constructions.

Findings

Constructed near-optimal seeded rank condensers.

Built explicit constant-degree dimension expanders.

Developed near-optimal lossy two-source condensers.

Abstract

An emerging theory of "linear-algebraic pseudorandomness" aims to understand the linear-algebraic analogs of fundamental Boolean pseudorandom objects where the rank of subspaces plays the role of the size of subsets. In this work, we study and highlight the interrelationships between several such algebraic objects such as subspace designs, dimension expanders, seeded rank condensers, two-source rank condensers, and rank-metric codes. In particular, with the recent construction of near-optimal subspace designs by Guruswami and Kopparty as a starting point, we construct good (seeded) rank condensers (both lossless and lossy versions), which are a small collection of linear maps $\mathbb{F}^n \to \mathbb{F}^t$ for $t \ll n$ such that for every subset of $\mathbb{F}^n$ of small rank, its rank is preserved (up to a constant factor in the lossy case) by at least one of the maps. We then compose a tensoring operation with our lossy rank condenser to construct constant-degree dimension expanders over polynomially large fields. That is, we give $O(1)$ explicit linear maps $A_i:\mathbb{F}^n\to \mathbb{F}^n$ such that for any subspace $V \subseteq \mathbb{F}^n$ of dimension at most $n/2$, $\dim\bigl( \sum_i A_i(V)\bigr) \ge (1+\Omega(1)) \dim(V)$. Previous constructions of such constant-degree dimension expanders were based on Kazhdan's property $T$ (for the case when $\mathbb{F}$ has characteristic zero) or monotone expanders (for every field $\mathbb{F}$); in either case the construction was harder than that of usual vertex expanders. Our construction, on the other hand, is simpler. Via an equivalence to linear rank-metric codes, we then construct optimal lossless two-source condensers. We then use our seeded rank condensers to obtain near-optimal lossy two-source condensers for constant rank sources.