Strong Law for the singular values and eigenvectors of random matrices III. Inequalities for the spectral radius of large random matrices (Q4490322)

[For parts I and II see ibid. 5, No.~1, 81-104 (1997; Zbl 0874.60024) and ibid. 6, No.~3, 291-310 (1998; Zbl 0926.60039), resp.]. The paper deals with the asymptotic analysis of distributions of singular eigenvalues of random matrices. One of the presented results is the following. Let the entries \(\xi^{(n)}_{ij}\), \(i=1,\ldots,m\), \(j=1,\ldots,n\), \(m\leq n\), of the random matrix \(\Xi\) be independent for every \(n\) and be defined on the common probability space, \(E\Xi=A=(a^{(n)}_{ij})\), \(E[\xi^{(n)}_{ij}- a^{(n)}_{ij}]^2=n^{-1}\), for certain \(\delta>0\), \(\sup _{n}\max _{i,j}E|(\xi^{(n)}_{ij}- a^{(n)}_{ij})\sqrt{n} |^{4+\delta}<\infty\), \(\alpha_{m}(AA^{\text{T}})< c<\infty\), where \(\alpha_{1}(AA^{\text{T}})\leq \ldots\leq\alpha_{m}(AA^{\text{T}})\) are the singular values of the matrix \(A\), \(\lim _{n\to\infty}\sup mn^{-1}<1\). Then in probability \(\lim _{n\to\infty} (\lambda_1(\Xi\Xi^{\text{T}})-\beta_1)=0\), \(\lim _{n\to\infty}(\lambda_{m}(\Xi\Xi^{\text{T}})-\beta_{2})=0\), where \(\lambda_{1}(\Xi\Xi^{\text{T}})\leq\ldots \leq\lambda_{m}(\Xi\Xi^{\text{T}})\) are the eigenvalues of the matrix \(\Xi\Xi^{\text{T}}\), and \(\beta_1, \beta_2\) are some numbers.