Say I have a frequency response of a $2 \times 2$ system at a fixed frequency: $G(i \omega$). It can be anything, for example $\begin{bmatrix} 1 & 3 \\ 4 & 2 \end{bmatrix}$. If I compute the SVD, I get the following: $U = \begin{bmatrix} -0.53 & -0.85 \\ -0.85 & 0.53 \end{bmatrix}$, $V = \begin{bmatrix} -0.77 & 0.64 \\ 0.64 & -0.77 \end{bmatrix}$ and $\Sigma = \begin{bmatrix} 5.12 & 0 \\ 0 & 1.95 \end{bmatrix}$. Usually $\bar{u}$ and $\bar{v}$ are defined as the first columns of $U$ and $V$, respectively. These two are related to "strongest" singular value, which is obvious from this example. However, I don't know how to interpret this computation: \begin{equation} \bar{v} \cdot \bar{u}^T\end{equation} \begin{equation} \begin{bmatrix} -0.77 \\ -0.64 \end{bmatrix} \cdot \begin{bmatrix} -0.53 \\ -0.85 \end{bmatrix}^T = \begin{bmatrix} 0.40 & 0.65 \\ 0.34 & 0.55 \end{bmatrix}\end{equation} If I do this, what am I exactly computing and what is the interpretation? The norm of the outcome is always equal to 1, I guess because U and V are unitary.


If $A=U\Sigma V^*$ is the SVD, $\Sigma$ is the diagonal matrix of singular values and $u_k, v_k$ are the columns of $U,V$ respectively then we have $A = \sum_k \sigma_k u_k v^*_k$.

Then if $\sigma_1 $ is much bigger than the other $\sigma_k$s then $\sigma_1 u_1 v^*_1$ is the 'dominant' part of $A$.

Note that you probably need the Hermitian transpose not just the transpose if you are dealing with frequency response. (And what is this $i$ in $G(i \omega)$?)

  • $\begingroup$ Regarding the parenthetical, $G(s)$ is usually used to refer to the transfer function (Laplace transoform) so that plugging in $s = i \omega$ (where $i^2 = -1$) gives us the Fourier transform, i.e. frequency response. $\endgroup$ Jun 21 '20 at 9:32
  • 1
    $\begingroup$ @Omnomnomnom: Sorry, that was tongue-in-cheek; I am an electrical engineer, usually it is written $G(j \omega)$, and is typically complex, hence my eyebrow raise :-). $\endgroup$
    – copper.hat
    Jun 21 '20 at 15:01
  • 1
    $\begingroup$ Ah, I guess I was a bit slow on the uptake there :) $\endgroup$ Jun 21 '20 at 15:24

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.