Tight Bounds for Sketching the Operator Norm, Schatten Norms, and Subspace Embeddings
Not on the map. Why?
Abstract (arXiv)
We consider the following oblivious sketching problem: given ε ∈ (0,1/3) and n ≥ d/ε2, design a distribution D over ℝk × nd and a function f: ℝk × ℝnd rightarrow ℝ, so that for any n × d matrix A, PrS ~ D [(1-ε) |A|op ≤ f(S(A),S) ≤ (1+ε)|A|op] ≥ 2/3, where |A|op is the operator norm of A and S(A) denotes S · A, interpreting A as a vector in ℝnd. We show a tight lower bound of k = Omega(d2/ε2) for this problem. Our result considerably strengthens the result of Nelson and Nguyen (ICALP, 2014), as it (1) applies only to estimating the operator norm, which can be estimated given any OSE, and (2) applies to distributions over general linear operators S which treat A as a vector and compute S(A), rather than the restricted class of linear operators corresponding to matrix multiplication. Our technique also implies the first tight bounds for approximating the Schatten p-norm for even integers p via general linear sketches, improving the previous lower bound from k = Omega(n2-6/p) [Regev, 2014] to k = Omega(n2-4/p). Importantly, for sketching the operator norm up to a factor of α, where α - 1 = Omega(1), we obtain a tight k = Omega(n2/α4) bound, matching the upper bound of Andoni and Nguyen (SODA, 2013), and improving the previous k = Omega(n2/α6) lower bound. Finally, we also obtain the first lower bounds for approximating Ky Fan norms.
Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). First posted on arXiv 20 Feb 2022.
Builds on
No reference list yet.
Built on by
Citing papers per year: 2023: 1; 2024: 2; 2025: 1.
4 papers.
- 2023Krylov Methods are (nearly) Optimal for Low-Rank Approximationcited by 2
- 2024Spectral Guarantees for Adversarial Streaming PCAcited by 2
- 2024Optimal Sketching for Residual Error Estimation for Matrix and Vector Normscited by 1
- 2025k-PCA for (non-squared) Euclidean Distances: Polynomial Time Approximationcited by 0
- 2025k-PCA for (non-squared) Euclidean Distances: Polynomial Time Approximation19 Jul 2025
- 2024Spectral Guarantees for Adversarial Streaming PCA19 Aug 2024
- 2024Optimal Sketching for Residual Error Estimation for Matrix and Vector Norms16 Aug 2024
- 2023Krylov Methods are (nearly) Optimal for Low-Rank Approximation6 Apr 2023