English

Extreme least singular values of Gaussian row submatrices and a phase retrieval stability problem

Probability 2026-07-07 v1

Abstract

Let F{R,C}\mathbb F\in\{\mathbb R,\mathbb C\} and dF=dimRFd_{\mathbb F}=\dim_{\mathbb R}\mathbb F. If AmFNm×mA_m\in\mathbb F^{N_m\times m} has independent standard Gaussian entries and Nm/mγ>1N_m/m\to\gamma>1, then minT[Nm]T=mσmin(Am,T)=(γγ(γ1)γ1)m/dF+oP(m). \min_{\substack{T\subset[N_m]\\ |T|=m}} \sigma_{\min}(A_{m,T}) = \left(\frac{\gamma^\gamma}{(\gamma-1)^{\gamma-1}}\right)^{-m/d_{\mathbb F}+o_P(m)} . If Nm=γm+O(1)N_m=\gamma m+O(1), the convergence of m1logMmFm^{-1}\log M_m^{\mathbb F} has probability error O(m1)O(m^{-1}). In particular, at the real phase-retrieval threshold N=2m1N=2m-1, ω(Am)=4m+oP(m), \omega(A_m)=4^{-m+o_P(m)}, so the Gaussian Balan--Wang critical exponential base is 1/41/4.

Keywords

Cite

@article{arxiv.2607.06249,
  title  = {Extreme least singular values of Gaussian row submatrices and a phase retrieval stability problem},
  author = {Yitzchak Shmalo},
  journal= {arXiv preprint arXiv:2607.06249},
  year   = {2026}
}