Lecture 11 - 11 The QR Algorithm 11.1 QR Algorithm without...

• biscuit1993
• 8

This preview shows pages 1–3. Sign up to view the full content.

11 The QR Algorithm 11.1 QR Algorithm without Shifts In the previous chapter (in the Maple worksheet 473 Hessenberg.mws ) we investigated two different attempts to tackling the eigenvalue problem. In the first attempt (which we discarded) the matrix A was multiplied from the left and right by a unitary House- holder matrix Q . We now consider the following Algorithm (“Pure” QR) Initialize A (0) = A for k = 1 , 2 , . . . Compute the QR factors Q ( k ) R ( k ) = A ( k - 1) Reassemble the factors A ( k ) = R ( k ) Q ( k ) end Note that R ( k ) = Q ( k ) T A ( k - 1) and so A ( k ) = Q ( k ) T A ( k - 1) Q ( k ) , the similarity transform from “approach 1” mentioned above. We recall that this algorithm will not produce a triangular (or, in our special real symmetric case, a diagonal) matrix in one step. However, it will succeed in an iterative setting. In fact, we will see below that the “pure” QR algorithm is equivalent to the orthogonal simultaneous iteration algorithm presented earlier. Remark In the next section we will discuss a “practical” QR algorithm that will use shifts and converge cubically like the Rayleigh quotient iteration. We now show the equivalence of the “pure” QR algorithm and orthogonal simulta- neous iteration. Recall the orthogonal simultaneous iteration algorithm: Initialize ˆ Q (0) with an arbitrary m × n matrix with orthonormal columns for k = 1 , 2 , . . . Z = A ˆ Q ( k - 1) Compute the QR factorization ˆ Q ( k ) ˆ R ( k ) = Z end Now consider the special case of square (real symmetric) matrices, which is what we will be working with in eigenvalue problems. We start the orthogonal simultaneous iteration with Q (0) = I R m × m , (S1) 88

This preview has intentionally blurred sections. Sign up to view the full version.

where we no longer need the hat (since the matrices are square), i.e., we will be perform- ing full QR factorizations. The underline is used to distinguish between the Q -matrices appearing in this algorithm and those in the “pure” QR below. Next, we compute Z = AQ ( k - 1) (S2) Q ( k ) R ( k ) = Z (S3) so that A ( k ) = Q ( k ) T AQ ( k ) . (S4) For the “pure” QR algorithm, on the other hand, we have A (0) = A (Q1) Q ( k ) R ( k ) = A ( k - 1) (Q2) A ( k ) = R ( k ) Q ( k ) (Q3) For this algorithm we also define Q ( k ) = Q (1) Q (2) . . . Q ( k ) , (Q4) and for both algorithms R ( k ) = R ( k ) R ( k - 1) . . . R (1) . (R5) Here we make the convention that empty products yield the identity. Theorem 11.1 Orthogonal simultaneous iteration (S1) (S4) and the “pure” QR algo- rithm (Q1) (Q4) are equivalent, i.e., both algorithms produce identical sequences R ( k ) , Q ( k ) and A ( k ) . In fact, the k -th power of A is given by A k = Q ( k ) R ( k ) , (28) and the k -th iterate by A ( k ) = Q ( k ) T AQ ( k ) . (29) Proof The proof uses induction on k .
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

What students are saying

• As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

Kiran Temple University Fox School of Business ‘17, Course Hero Intern

• I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

Dana University of Pennsylvania ‘17, Course Hero Intern

• The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

Jill Tulane University ‘16, Course Hero Intern