English

The smallest singular value of deformed random rectangular matrices

Probability 2018-10-17 v2

Abstract

We prove an estimate on the smallest singular value of a multiplicatively and additively deformed random rectangular matrix. Suppose nNMΛNn\le N \le M \le \Lambda N for some constant Λ1\Lambda \ge 1. Let XX be an M×nM\times n random matrix with independent and identically distributed entries, which have zero mean, unit variance and arbitrarily high moments. Let TT be an N×MN\times M deterministic matrix with comparable singular values csN(T)s1(T)c1c\le s_{N}(T) \le s_{1}(T) \le c^{-1} for some constant c>0c>0. Let AA be an N×nN\times n deterministic matrix with A=O(N)\|A\|=O(\sqrt{N}). Then we prove that for any ϵ>0\epsilon>0, the smallest singular value of TXATX-A is larger than Nϵ(Nn1)N^{-\epsilon}(\sqrt{N}-\sqrt{n-1}) with high probability. If we assume further the entries of XX have subgaussian decay, then the smallest singular value of TXATX-A is at least of the order Nn1\sqrt{N}-\sqrt{n-1} with high probability, which is an essentially optimal estimate.

Keywords

Cite

@article{arxiv.1702.04050,
  title  = {The smallest singular value of deformed random rectangular matrices},
  author = {Fan Yang},
  journal= {arXiv preprint arXiv:1702.04050},
  year   = {2018}
}

Comments

23 pages

R2 v1 2026-06-22T18:17:35.205Z