2000 character limit reached
Singular Vector Perturbation under Gaussian Noise
Published 9 Aug 2012 in math.NA, math.ST, and stat.TH | (1208.1811v3)
Abstract: We perform a non-asymptotic analysis on the singular vector distribution under Gaussian noise. In particular, we provide sufficient conditions on a matrix for its first few singular vectors to have near normal distribution. Our result can be used to facilitate the error analysis in linear dimension reduction.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.