On bilinear forms based on the resolvent of large random matrices (Q1943319): Difference between revisions
From MaRDI portal
Created a new Item |
Added link to MaRDI item. |
||
links / mardi / name | links / mardi / name | ||
Revision as of 16:54, 1 February 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | On bilinear forms based on the resolvent of large random matrices |
scientific article |
Statements
On bilinear forms based on the resolvent of large random matrices (English)
0 references
19 March 2013
0 references
Consider an \(N\times n\) random matrix \(\Sigma_n=\frac{D_n^{1/2}X_n\tilde{D}_n^{1/2}}{\sqrt{n}}+A_n=Y_n+A_n\) satisfying { Assumption A-1:} The entries of matrices \((X_n)\), i.e. \((X_{ij}^n; i,j,n)\), are complex, independent and identically distributed with \(\mathbb{E}X_{ij}^n=0\) and \(\mathbb{E}|X_{ij}^n|=1\). { Assumption A-2:} The family of deterministic \(N\times n\) matrices \((A_n)\) is bounded for the spectral norm as \(N,n\rightarrow \infty\). { Assumtpion A-3:} The family of real deterministic nonnegative diagonal \(N\times N\) and \(n\times n\) matrices \((D_n)\) and \((\tilde{D}_n)\) are bounded for the spectral norm and \(\inf_N \frac{1}N\text{Tr} D_n>0\) and \(\inf_n \frac{1}n\text{Tr} \tilde{D}_n>0\). Denote by \(Q_n(z)\) the resolvent associated to \(\Sigma_n\Sigma^*_n\), i.e. \(Q_n(z)=(\Sigma_n\Sigma_n^*-zI_n)^{-1}\) and given two sequences of deterministic vectors \((u_n)\) and \((v_n)\), the authors obtain the main result: { Theorem 1.1:} Assume that there exists an integer \(p\geq 1\) such that \(\sup_n \mathbb{E}|X_{ij}^n|^{8p}< \infty\), then for every \(z\in\mathbb{C}-\mathbb{R}^+\), \[ \mathbb{E}|u_n^*(Q_n(z)-T_n(z))v_n|^{2p}\leq \frac{1}{n^p} f_p(z) \|u_n\|^{2p}\|v_n\|^{2p}, \] where \(T_n(z)\) is a special approximation of \(Q_n(z)\) related to the Stieltjes transformation and \(f_p\) is a function depending on \(p\) but not on \(u_n\) and \(v_n\). Two applications of the main result to wireless communication system and signal processing are stated in Section 2. The idea of the proof is to write \[ u^*(Q-T)v=u^*(Q-\mathbb{E}Q)v+u^*(\mathbb{E}Q-R)v+u^*(R-T)v \] and study each part separately.
0 references
bilinear forms
0 references
random matrix
0 references
empirical distribution of the eigenvalues
0 references
Stieltjes transform
0 references
spectral norm
0 references
wireless communication system
0 references
signal processing
0 references