Least singular value and condition number of a square random matrix with i.i.d. rows (Q2244514)

The authors investigate the general probability distributions on \(\mathbb R^n\) and, using the notion of a mould, they generalize some results about the least singular value \(\sigma_{\min}\) and the condition number \(\kappa\) of random matrices with independent and identically distributed rows. The two main results are as follows. \begin{itemize} \item[1.] If \(\tilde X\) is a square random matrix with i.i.d. rows, then \[ \liminf_{\varepsilon\to 0^+}\frac{\mathbb P\left(\sigma_{\min}(\tilde X)<\varepsilon\right)}{\varepsilon}>0. \] Moreover, if \(\tilde X\) is invertible almost surely, then \[ \mathbb E\left[\frac 1{|\sigma_{\min}(\tilde X)|}\right]=\mathbb E\left[||\tilde X^{-1}||\right]=\infty. \] \item[2.] For any square random matrix \(\tilde X\) with i.i.d. rows and for any matrix norm \[ \mathbb E\left[\kappa(\tilde X)\right]=\infty. \] \end{itemize}