2
$\begingroup$

I have a matrix which is "almost" like an upper triangular just that the last row has non zero elements. And I want to perform the QR decomposition on that matrix.

Does anyone know the "name" of such a matrix? And also since I don't want to use the built in matlab QR function for such a matrix, can anyone suggest a better algorithm for me to use?

| cite | improve this question | | | | |
$\endgroup$
  • $\begingroup$ Have you considered the Cholesky decomposition, please? Or LU decomposition? I use R rather than Matlab, but I believe that those may be effective for your problem. Would you happen to have a small numeric example, by any chance? $\endgroup$ – user1544953 Sep 28 '15 at 3:06
  • $\begingroup$ The main idea is to observe the sensitivity of the diagonal elements. The original matrix was a full matrix where I used the built in matlab function QR to decompose it. But now I have purturbed (R+E) the R upper triangular matrix of the decomposition with E( E is an NxN matrix with its elements all zeros except at the last row where it has non zero elements) The shape of (R+E) is almost like the Upper triangular matrix R since it has zeros below its diagonal except it's last row. $\endgroup$ – user1144656 Sep 28 '15 at 10:58
  • $\begingroup$ I am expected to take the QR decomposition of this new matrix (R+E) in other to observe how sensitive the diagonal elements of R with respect to (E) is. I no longer want to use the matlab QR built function since this new matrix is (almost upper triangular except at the last row where you have non zero elements). I do have numeric examples but since I'm new here, I don't know how to type in matrices. $\endgroup$ – user1144656 Sep 28 '15 at 10:58
  • $\begingroup$ This does not provide an answer to the question. To critique or request clarification from an author, leave a comment below their post - you can always comment on your own posts, and once you have sufficient reputation you will be able to comment on any post. $\endgroup$ – nicoguaro♦ Sep 28 '15 at 19:59
  • 1
    $\begingroup$ @user1544953 Cholesky only works for SPD matrices. I doubt that the matrix here is symmetric. $\endgroup$ – Federico Poloni Nov 20 '15 at 2:56
4
$\begingroup$

Look at section 5 of http://web.stanford.edu/group/SOL/papers/ggms74.pdf to see how to update QR decomposition when a row is added to the matrix you wish to factorize. You may find it not so easy to fully digest.

I think algorithm 6.16 of Chapter 6 of Scientific Computing - An Introduction using Maple and MATLAB will do the trick for you, allowing you to update in MATLAB an existing QR factorization when you add a row to the matrix you wish to factorize. Here are the MATLAB m files http://www.cs.cornell.edu/cv/GVL4/M-Files/Chapter%206/Chap6.htm .

| cite | improve this answer | | | | |
$\endgroup$
2
$\begingroup$

If the matrix is almost triangular, you can just update the QR to get rid of the unwanted non-zeros. You can do that by using Givens rotations (see Golub and Van-Loan, "Matrix Computations" 4th edition, Chapter 6.5).

You can also use MATLAB's "qrinsert" and add the additional row to the QR factorization.

| cite | improve this answer | | | | |
$\endgroup$

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.