← All papers
First page of Subspace embeddings with the rerandomized SRHT

Subspace embeddings with the rerandomized SRHT

Yuning Yang

math.NA Sep 12, 2026 · v1
A Lean 4 formalization of the prescribed-width subspace embedding theorem and supporting results is provided in a companion repository.
This work studies subspace embeddings obtained by two normalized real Walsh transforms, two independent sign diagonals, and uniform coordinate sampling without replacement. The main result shows that the prescribed sample size $k=\min\{n,\lceil Cr/\varepsilon^2\rceil\}$, for a universal constant $C$, suffices to preserve all squared norms on each fixed $r$-dimensional subspace within $1\pm\varepsilon$ with probability at least $0.99$. The result holds for every ambient Walsh dimension and all ranks, and answers Problem TR-01 in the Open Problems in Numerical Linear Algebra repository. The proof controls joint entry cumulants of the transformed projection through connected graph contractions and Walsh character identities. These estimates then bound the expected trace of even powers of a product of centered projections. A two-projection decomposition converts this estimate into control of both spectral edges. Bernoulli sampling at arbitrary densities and a deterministic upper bound near full sampling yield the prescribed number of coordinates.

Open Problem TR-01 in the Open Problems in Numerical Linear Algebra repository asks whether a rerandomized SRHT (two sign diagonals and two Walsh transforms) with uniform coordinate sampling gives subspace embeddings of dimension proportional to r/ε² at a prescribed sample size, avoiding the logarithmic obstruction of one-round constructions.

Two normalized real Walsh transforms, two independent Rademacher sign diagonals, and uniform coordinate sampling without replacement define the embedding. The proof controls joint entry cumulants of the transformed projection via connected graph contractions and Walsh character identities, bounding the expected trace of even powers of a product of centered projections. A two-projection decomposition converts trace estimates into control of both spectral edges, and coupling of Bernoulli densities transfers the result to a fixed sample size. A Lean 4 formalization of the prescribed-width theorem and supporting lemmas accompanies the manuscript.

The sample size k=min{n,⌈Cr/ε²⌉} for a universal constant C suffices to preserve all squared norms on each fixed r-dimensional subspace within 1±ε with probability at least 0.99, for every ambient Walsh dimension and all ranks, affirmatively resolving TR-01.