A CLT for information-theoretic statistics of non-centered Gram random matrices

Random matrices (algebraic aspects) (15B52) Random matrices (probabilistic aspects) (60B20) Central limit and other weak theorems (60F05) Communication theory (94A05) Information theory (general) (94A15) Coding theorems (Shannon theory) (94A24)

Abstract: In this article, we study the fluctuations of the random variable:

{mathcal I}_n( ho) = frac 1N logdet(Sigma_n Sigma_n^* + ho I_N),quad ( ho>0)

where

S i g m a_{n} = n^{- 1 / 2} D_{n}^{1 / 2} X_{n} i l d e D_{n}^{1 / 2} + A_{n}

, as the dimensions of the matrices go to infinity at the same pace. Matrices

X_{n}

and

A_{n}

are respectively random and deterministic

N i m e s n

matrices; matrices

D_{n}

and

i l d e D_{n}

are deterministic and diagonal, with respective dimensions

N i m e s N

and

n i m e s n

; matrix

X_{n} = (X_{i j})

has centered, independent and identically distributed entries with unit variance, either real or complex. We prove that when centered and properly rescaled, the random variable

{m a t h c a l I}_{n} (h o)

satisfies a Central Limit Theorem and has a Gaussian limit. The variance of

{m a t h c a l I}_{n} (h o)

depends on the moment

E X_{i j}^{2}

of the variables

X_{i j}

and also on its fourth cumulant

k a p p a = E | X_{i j} |^{4} - 2 - | E X_{i j}^{2} |^{2}

. The main motivation comes from the field of wireless communications, where

{m a t h c a l I}_{n} (h o)

represents the mutual information of a multiple antenna radio channel. This article closely follows the companion article "A CLT for Information-theoretic statistics of Gram random matrices with a given variance profile", {em Ann. Appl. Probab. (2008)} by Hachem et al., however the study of the fluctuations associated to non-centered large random matrices raises specific issues, which are addressed here.

Review text

The authors consider an \(N\times n\) random matrix \(\Sigma_n\) given by \(\Sigma_n=n^{-1/2}D_n^{1/2}X_n\tilde D_n^{1/2}+A_n\), where \(X_n\) and \(A_n\) are, respectively, random and deterministic \(N\times n\) matrices; \(D_n\) and \(\tilde D_n\) are deterministic square diagonal matrices of appropriate dimensions; the matrix \(X_n=(X_{ij})\) has centered, independent and identically distributed entries with unit variance, either real or complex. The authors study the fluctuations of the random variable NEWLINE\[NEWLINE{\mathcal I}_n(\rho)=(1/N)\log\det(\Sigma_n\Sigma_n^*+\rho I_N),\;\rho>0,NEWLINE\]NEWLINE as the dimension \(n\) goes to infinity.NEWLINENEWLINEThey prove that when centered and properly rescaled, the random variable \({\mathcal I}_n(\rho)\) satisfies a central limit theorem (CLT) and has a Gaussian limit.NEWLINENEWLINEThe main motivation comes from the field of wireless communications, where \({\mathcal I}_n(\rho)\) represents the mutual information of a multiple antenna radio channel.

Recommendations

Cites work

Cited in

(10)

This page was built for publication: A CLT for information-theoretic statistics of non-centered Gram random matrices

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2884857)