I am dealing with a theorem which relates to circularly symmetric complex Gaussian random matrices (CSGRM).

It seems quite tempting to assume that the theorem also extends to real-valued Gaussian random matrices, since CSGRMs have their real and imaginary parts completely uncorrelated, I can essentially "assume" the imaginary part to be zero.

In particular I am dealing with the following multi variate Gaussian model:

\begin{equation} y = Hx + n, \end{equation}

where $H$ is $r\times t$ complex valued, $x$ is $t$ complex valued, $y$ is $r$ complex valued. Theorem 1 of the paper "Capacity of Multi‐antenna Gaussian Channels" (available: http://web.mit.edu/18.325/www/telatar_capacity.pdf ) has Theorem 1 as:

\begin{equation} \mathbb{E}[\log \text{det} (I_r +(P/t)HH^{\dagger})] \end{equation}

which the author claims is true assuming $x$ is a CSGR vector.


Can I readily just assume it also naturally extends to real Gaussian random matrices? Based on the few random matrix properties Theorem 1 is based on, it seems like it should ... but I dont have the confidence to say this completely!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.