245
$\begingroup$

Let

$$ H=\left(\begin{array}{cccc} 0 & 1/2 & 0 & 1/2 \\ 1/2 & 0 & 1/2 & 0 \\ 1/2 & 0 & 0 & 1/2\\ 0 & 1/2 & 1/2 & 0 \end{array}\right), $$

$K_1=\left(\begin{array}{c}1 \\ 0\end{array}\right)$ and consider the sequence of matrices defined by $$ K_L = \underset{2^{L}\times 2^{L}}{\underbrace{\left[H\otimes I_{2^{L-2}}\right]}}\underset{2^{L}\times 2^{L-1}}{\underbrace{\left[I_2 \otimes K_{L-1}\right]}}\in\mathbb{R}^{2^L\times 2^{L-1}}, $$ where $\otimes$ denotes the Kronecker product, and $I_n$ is the $n\times n$ identity matrix.

I am interested in the limiting behaviour of the singular values of $K_L$ as $L$ tends to infinity. Some calculations indicate that the $2^L\times 2^{L-1}$-matrix $K_L$ has $L$ non-zero singular values and that the empirical distribution of those nonzero singular values converges to some limit. Can this limit be described in terms of the matrix $H$?

I am wondering if it is possible to use some kind of fixed-point theorem to characterise the limit (in any sense) $\lim_{L\to\infty}K_L$ as an operator on some sequence space.


Edit: I did some more experiments and it seems that the limiting behaviour of the singular values of $K_L$ does not only depend on the matrix $H$, but also on the initial value $K_1$.

To illustrate this, let $K_1(\alpha)=\left(\begin{array}{c}1 \\ \alpha\end{array}\right)$ and consider the sequence $$ K_L(\alpha) = \left[H\otimes I_{2^{L-2}}\right]\left[I_2 \otimes K_{L-1}(\alpha)\right]. $$

The largest singular value of $K_{10}(\alpha)$ is depicted in the following figure. (The graph looks essentially the same for all $L\geq 4$ instead of $L=10$.)

The largest singular value of <span class=$K_{10}(\alpha)$">

The minimum is approximately $(-.2936,0.7696)$.

This makes it unlikely for fixed-point arguments to work in this setting. I, therefore, modify my question and ask if the limiting behaviour of the singular values of $K_L$ (or $K_L(\alpha)$) can be characterised directly in terms of $H$ and the initial value $K_1$ (or $K_1(\alpha)$).

Edit 2 (March 2015): As the question is still receiving attention, let me add that I came up with a conjecture for the asymptotic behaviour of the singular values of $K_L(\alpha)$, as detailed in this MO post.

$\endgroup$
  • 7
    $\begingroup$ Could one propose a formulation in terms of tensors of an increasing rank over 2-dimensional vector space? My intuition is jammed with this Kronecker product stuff, but I’ll possibly understand something when saw tensors. $\endgroup$ – Incnis Mrsi Aug 11 '14 at 16:46
  • 9
    $\begingroup$ I do not think that I could contribute, but do the brackets in the definition of $K_L$ denote something special? Or are they just for emphasizing that the Kronecker product is evaluated before the matrix multiplication? $\endgroup$ – flawr Nov 5 '14 at 20:22
  • 8
    $\begingroup$ @flawr It's only that Kronecker product is evaluated before the matrix multiplication $\endgroup$ – Paul Jan 25 '16 at 16:32
  • 5
    $\begingroup$ Does kind of resemble this other question involving Haar Wavelets, if multipling whole matrix by -1 before conversion : math.stackexchange.com/questions/2198335/… $\endgroup$ – mathreadler Apr 17 '17 at 18:18
  • 7
    $\begingroup$ What was the motivation for this problem? $\endgroup$ – Vítězslav Štembera May 1 '19 at 16:47

Browse other questions tagged or ask your own question.