1
$\begingroup$

4I am not in major in math. But currently I am working with a couple of matrices in the form like this: \begin{equation} \left[\begin{array}{rrrrrrrrrrrr} 1 & 0 & 0 & 0 & 0& 0& 0& 0& 0& 0& 0& 0\\ 0 & 1 &0.5 &0.5 & 0& 0& 0& 0& 0& 0& 0& 0\\ 0 & 0 &0.5 &0.5 & 1& 0.5& 0.5& 0& 0& 0& 0& 0\\ 0 & 0 & 0 & 0 & 0& 0.5& 0.5& 1& 0.5& 0.5& 0& 0\\ 0 & 0 & 0 & 0 & 0& 0& 0& 0& 0.5& 0.5& 1& 0\\ 0 & 0 & 0 & 0 & 0& 0& 0& 0& 0& 0& 0& 1 \end{array}\right] \end{equation} If we consider only nonzero elements, 1-3 columns is invertible, 4-6 columns is invertible, and so on. Is there any efficient way to compute Moore-Penrose pseudoinverse for this type of matrices?

$\endgroup$
3
  • $\begingroup$ How big are the matrices you are working with? $\endgroup$ – K. Miller Oct 10 '15 at 14:06
  • $\begingroup$ From tens of n by n matrices to hundreds of n by n matrices arranged in this pattern. $\endgroup$ – Di Miao Oct 10 '15 at 14:16
  • $\begingroup$ If the matrices are on the order of 10 or 100, then they are not very large and using the svd to compute the Moore-Penrose pseudoinverse should be fast enough. $\endgroup$ – K. Miller Oct 10 '15 at 14:59
1
$\begingroup$

Construction of a block pseudoinverse is an unanswered challenge in my research area. It's easy to build block inverses of Toeplitz matrices which are nonsingular.

Here is your data with symmetry axes: $$ \mathbf{A} = \left( \begin{array}{cccccc|cccccc} 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 1 & \frac{1}{2} & \frac{1}{2} & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & \frac{1}{2} & \frac{1}{2} & 1 & \frac{1}{2} & \frac{1}{2} & 0 & 0 & 0 & 0 & 0 \\\hline 0 & 0 & 0 & 0 & 0 & \frac{1}{2} & \frac{1}{2} & 1 & \frac{1}{2} & \frac{1}{2} & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & \frac{1}{2} & \frac{1}{2} & 1 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 \\ \end{array} \right), $$

$$ \mathbf{A}^{\dagger} = \left( \begin{array}{rrr|rrr} 56 & 0 & 0 & 0 & 0 & 0 \\ 0 & 41 & -11 & 3 & -1 & 0 \\ 0 & 15 & 11 & -3 & 1 & 0 \\ 0 & 15 & 11 & -3 & 1 & 0 \\ 0 & -11 & 33 & -9 & 3 & 0 \\ 0 & -4 & 12 & 12 & -4 & 0 \\\hline 0 & -4 & 12 & 12 & -4 & 0 \\ 0 & 3 & -9 & 33 & -11 & 0 \\ 0 & 1 & -3 & 11 & 15 & 0 \\ 0 & 1 & -3 & 11 & 15 & 0 \\ 0 & -1 & 3 & -11 & 41 & 0 \\ 0 & 0 & 0 & 0 & 0 & 56 \\ \end{array} \right). $$

Below are gray scale renderings; $\mathbf{A}$ on the left, $|\mathbf{A}^{\dagger}|$ on the right.

A Apinv

This is an open problem.

$\endgroup$

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.