# On the condition number of the shifted real Ginibre ensemble

Giorgio Cipolloni, László Erdős, Dominik Schröder

SIAM J. Matrix Anal. Appl.Vol. 43 (2022)

## Summary

We show that complex shifts of real matrices have a stronger regularizing effect than real shifts. As a consequence we obtain improved bounds on the practically relevant condition number and demonstrate that our results give sharp asymptotics for the running time of the CG algorithm.## Abstract

We derive an accurate lower tail estimate on the lowest singular value $\sigma_1(X-z)$ of a real Gaussian (Ginibre) random matrix $X$ shifted by a complex parameter $z$. Such shift effectively changes the upper tail behaviour of the condition number $\kappa(X-z)$ from the slower $\mathbf{P}(\kappa(X-z)\ge t)\lesssim 1/t$ decay typical for real Ginibre matrices to the faster $1/t^2$ decay seen for complex Ginibre matrices as long as $z$ is away from the real axis. This sharpens and resolves a recent conjecture in [arXiv:2005.08930] on the regularizing effect of the real Ginibre ensemble with a genuinely complex shift. As a consequence we obtain an improved upper bound on the eigenvalue condition numbers (known also as the eigenvector overlaps) for real Ginibre matrices. The main technical tool is a rigorous supersymmetric analysis from our earlier work [arXiv:1908.01653].