A unified framework for linear dimensionality reduction in L1
Abstract
For a family of interpolation norms on , we provide a distribution over random matrices parametrized by sparsity level such that for a fixed set of points in , if then with high probability, for all . Several existing results in the literature reduce to special cases of this result at different values of : for , and we recover that dimension reducing linear maps can preserve the -norm up to a distortion proportional to the dimension reduction factor, which is known to be the best possible such result. For , , and we recover an variant of the Johnson-Lindenstrauss Lemma for Gaussian random matrices. Finally, if is -sparse, then and we recover that -sparse vectors in embed into via sparse random matrix constructions.
Cite
@article{arxiv.1405.1332,
title = {A unified framework for linear dimensionality reduction in L1},
author = {Felix Krahmer and Rachel Ward},
journal= {arXiv preprint arXiv:1405.1332},
year = {2015}
}
Comments
18 pages, 1 figure