Abstract
For fixed positive integers , given an n-dimensional random vector , consider its k-dimensional projection , where is an -dimensional matrix belonging to the Stiefel manifold of orthonormal k-frames in . For a class of sequences that includes uniform distributions on suitably scaled balls, , and product measures with sufficiently light tails, it is shown that the sequence of projected vectors satisfies a large deviation principle whenever the empirical measures of the rows of converge, as , to a probability measure on . In particular, this implies a (quenched) large deviation principle for the sequence for almost every realization of , where each is a random matrix, independent of , that is distributed according to the normalized Haar measure on . Moreover, a variational formula is obtained for the rate function of the large deviation principle for the annealed projections , in terms of a family of quenched rate functions and a modified entropy term. A key step in this analysis is a large deviation principle for the sequence of empirical measures of rows of the random matrices , , which may be of independent interest. The study of multidimensional random projections of high-dimensional measures is of interest in asymptotic functional analysis, convex geometry and statistics. Prior results on quenched large deviations for random projections of balls have been essentially restricted to the one-dimensional setting.
Funding Statement
This work reports results from Steven Soojin Kim’s PhD Thesis. Kavita Ramanan was supported in part by NSF-DMS Grants 1713032 and 1954351.
Dedication
This article is dedicated to the memory of Elizabeth Meckes
Citation
Steven Soojin Kim. Kavita Ramanan. "Large deviation principles induced by the Stiefel manifold, and random multidimensional projections." Electron. J. Probab. 28 1 - 23, 2023. https://doi.org/10.1214/23-EJP1023
Information