Szarek-type singular-value bound for random matrices satisfying the general assumption

From papers

Let Mn\boldsymbol{M}_n be a real random matrix satisfying the paper's Assumption with parameter K>0K>0, possibly with some moment assumptions on its entries. For a deterministic matrix ARn×nA\in\mathbb{R}^{n\times n}, let σnk+1(A+Mn)\sigma_{n-k+1}(A+\boldsymbol{M}_n) denote the (nk+1)(n-k+1)-st singular value. Generalized Szarek conjecture. There is a universal constant CC such that, for every such AA and every relevant kk and ε\varepsilon,

P[σnk+1(A+Mn)kεn](CKε)k2.\mathbb{P}\left[\sigma_{n-k+1}(A+\boldsymbol{M}_n)\leq \frac{k\varepsilon}{n}\right]\leq (CK\varepsilon)^{k^2}.

The claim would extend Szarek's singular-value bound for real Ginibre matrices to the more general class specified by the assumption and would improve the stated general theorem in its dependence on kk and nn. The supplied status evidence says that the claim was proven by Tikhomirov under weaker independence assumptions.

Progress summary

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Sources & referencesView supporting material

Primary source

Jess Banks, Jorge Garza Vargas, Archit Kulkarni and Nikhil Srivastava, “Overlaps, Eigenvalue Gaps, and Pseudospectrum under real Ginibre and Absolutely Continuous Perturbations”, arXiv:2005.08930 (2020).

Solutions 0

No solutions have been posted yet.