Abstract
Given two sets $x_1^{(1)},\ldots,x_{n_1}^{(1)}$ and$x_1^{(2)},\ldots,x_{n_2}^{(2)}\in\mathbb{R}^p$ (or $\mathbb{C}^p$) of randomvectors with zero mean and positive definite covariance matrices $C_1$ and$C_2\in\mathbb{R}^{p\times p}$ (or $\mathbb{C}^{p\times p}$), respectively,this article provides novel estimators for a wide range of distances between$C_1$ and $C_2$ (along with divergences between some zero mean and covariance$C_1$ or $C_2$ probability measures) of the form $\frac1p\sum_{i=1}^nf(\lambda_i(C_1^{-1}C_2))$ (with $\lambda_i(X)$ the eigenvalues of matrix $X$).These estimators are derived using recent advances in the field of randommatrix theory and are asymptotically consistent as $n_1,n_2,p\to\infty$ withnon trivial ratios $p/n_1<1$ and $p/n_2<1$ (the case $p/n_2>1$ is alsodiscussed). A first "generic" estimator, valid for a large set of $f$functions, is provided under the form of a complex integral. Then, for aselected set of $f$'s of practical interest (namely, $f(t)=t$, $f(t)=\log(t)$,$f(t)=\log(1+st)$ and $f(t)=\log^2(t)$), a closed-form expression is provided.Beside theoretical findings, simulation results suggest an outstandingperformance advantage for the proposed estimators when compared to theclassical "plug-in" estimator $\frac1p\sum_{i=1}^n f(\lambda_i(\hatC_1^{-1}\hat C_2))$ (with $\hatC_a=\frac1{n_a}\sum_{i=1}^{n_a}x_i^{(a)}x_i^{(a){\sf T}}$), and this even forvery small values of $n_1,n_2,p$.