Extreme value theory for singular subspace estimation in the matrix denoising model
Junhyung Chang, Joshua Cape · 2025
Read Extreme value theory for singular subspace estimation in the matrix denoising model by Junhyung Chang, Joshua Cape (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-19978.txt.
Introduction
Junhyung Chang and Joshua Cape study how statisticians can recover structure from a noisy table of numbers. The setting is matrix denoising: a clean signal table is buried under random Gaussian noise, and the goal is to recover its leading directions. Most earlier work measured error in average ways. This paper asks about the worst row instead, using a two-to-infinity norm that captures the largest row-wise miss.
They prove that, after centering and scaling, that error follows a Gumbel curve, the law of extremes, as the table grows. From the limit they build a hypothesis test for low-rank structure, along with corrected estimates of the signal strengths it needs. In trials, the new test beat the average-based distance at spotting alternatives that differ in only a few entries or rows. It held up when the noise was not Gaussian.
Catalog record
| CREATOR | Junhyung Chang, Joshua Cape |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 218 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-19978