3
$\begingroup$

I would appreciate if one could help me figure out this problem.

I have a matrix $G$ (for simplicity assume square matrix $n\times n$). I know that if I multiply $G$ with a unitary matrix $U$ as ($A = GU$), the resultant matrix has the same distribution as $G$, i.e., the distribution of the singular values of $G$ and $A$ are equal/identical.

Now let's assume $k<n$ and $V_{n\times k}$ is a semi-unitary matrix ($V^H V = I_{k\times k}$). I am interested in the distribution of the singular values of $B=GV$. Clearly, $B$ has $k$ singular values. Is there anyway to make connection between the singular values of $B$ and $G$.

Any help/hint/reference is appreciated.

$\endgroup$
  • 2
    $\begingroup$ Isn't it $B=GV$ ? $\endgroup$ – Jean Marie Feb 22 '16 at 21:39
  • $\begingroup$ sorry for the mistake. you are right. I just edited the question. $\endgroup$ – M.X Feb 23 '16 at 13:42
0
$\begingroup$

Let $\sigma_1\geq\cdots\geq \sigma_n$ be the singular values of $G$ and let $x\in \mathbb{C}^k$ be a unitary vector. Then $x^*(B^*B)x=(Vx)^*(G^*G)(Vx)=y^*(G^*G)y$ where $y$ is a unitary vector. Then $x^*(B^*B)x\in [\sigma_n^2,\sigma_1^2]$ and the non-zero singular values of $B$ are in $[\sigma_n,\sigma_1]$.

$\endgroup$
  • $\begingroup$ I see that the assertion "Any help/hint/reference is appreciated" is not true. $\endgroup$ – loup blanc Apr 21 '16 at 9:40

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.