Johnson-Lindenstrauss Embeddings with Kronecker Structure
Abstract
We prove the Johnson-Lindenstrauss property for matrices where has the restricted isometry property and is a diagonal matrix containing the entries of a Kronecker product of independent Rademacher vectors. Such embeddings have been proposed in recent works for a number of applications concerning compression of tensor structured data, including the oblivious sketching procedure by Ahle et al. for approximate tensor computations. For preserving the norms of points simultaneously, our result requires to have the restricted isometry property for sparsity . In the case of subsampled Hadamard matrices, this can improve the dependence of the embedding dimension on to while the best previously known result required . That is, for the case of at the core of the oblivious sketching procedure by Ahle et al., the scaling improves from cubic to quadratic. We provide a counterexample to prove that the scaling established in our result is optimal under mild assumptions.
Keywords
Cite
@article{arxiv.2106.13349,
title = {Johnson-Lindenstrauss Embeddings with Kronecker Structure},
author = {Stefan Bamberger and Felix Krahmer and Rachel Ward},
journal= {arXiv preprint arXiv:2106.13349},
year = {2021}
}