Two sided Matrix Algorithms

# Acm transactions on mathematical software vol 28 no 4

This preview shows pages 7–10. Sign up to view the full content.

ACM Transactions on Mathematical Software, Vol. 28, No. 4, December 2002.

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

View Full Document
Recursive Blocked Algorithms—Part II 423 Since all matrices are square, we only have one way of doing a recursive splitting . We split A , E , and C by rows and columns: A 11 A 12 A 22 ‚ • X 11 X 12 X 21 X 22 A T 11 A T 12 A T 22 - E 11 E 12 E 22 ‚ • X 11 X 12 X 21 X 22 E T 11 E T 12 E T 22 = C 11 C 12 C 21 C 22 . Since X 21 = X T 12 , the recursive splitting results in three triangular generalized discrete-time Lyapunov equations: A 11 X 11 A T 11 - E 11 X 11 E T 11 = C 11 - A 12 X T 12 A T 11 - ( A 11 X 12 + A 12 X 22 ) A T 12 + E 12 X T 12 E T 11 + ( E 11 X 12 + E 12 X 22 ) E T 12 , A 11 X 12 A T 22 - E 11 X 12 E T 22 = C 12 - A 12 X 22 A T 22 + E 12 X 22 E T 22 , A 22 X 22 A T 22 - E 22 X 22 E T 22 = C 22 . We start by solving for X 22 in the third equation. After updating C 12 with respect to X 22 , we can solve for X 12 . Finally, after updating C 11 with respect to X 12 and X 22 , we solve for X 11 . Four of the two-sided matrix product updates of C 11 can be expressed as two SYR2K operations: C 11 = C 11 - ( A 11 X 12 ) A T 12 - A 12 ( A 11 X 12 ) T , and C 11 = C 11 + ( E 11 X 12 ) E T 12 + E 12 ( E 11 X 12 ) T , where A 11 X 12 and E 11 X 12 are TRMM operations. The GEMM-rich updates C 11 = C 11 - A 12 X 22 A T 12 and C 11 = C 11 + E 12 X 22 E T 12 are performed by the SLICOT MB01RD subroutine; see Section 4.1. In Algorithm 2, a Matlab-style function [ X ] = rtrglydt ( A , E , C , blks ) im- plementing our recursive blocked solver is presented. The function [ X ] = trglydt ( A , E , C ) implements an algorithm for solving triangular generalized discrete-time Lyapunov kernel problems. For solving the triangular general- ized Sylvester equations that appear, we make use of the recursive algorithm rtrgsyl , described in Section 3.1. 3.4 Recursive Triangular Discrete-Time Lyapunov Solvers The real discrete-time Lyapunov (LYDT) or Stein matrix equation is obtained by setting E = I N in (3). We get the equation AXA T - X = C , (4) where A is upper triangular or upper quasitriangular, that is, in real Schur form. This equation is also the result of setting A = B in (2). If C is symmetric, then X is symmetric as well, and the Stein matrix equation corresponds to the discrete-time standard algebraic Lyapunov equation. The LYDT equation (4) has a unique solution if and only if λ i λ j 6= 1 for all i and j , or equivalently ACM Transactions on Mathematical Software, Vol. 28, No. 4, December 2002.
424 I. Jonsson and B. K˚agstr¨om Algorithm 2: rtrglydt Input: ( A , E ) ( N × N ) in upper generalized Schur form. C ( N × N ) dense matrix. blks , block size that specifies when to switch to a standard algorithm for solving small-sized triangular generalized Lyapunov equations. Output: X ( N × N ), the solution of AXA T - EXE T = C . X is allowed to overwrite C , and is symmetric if C = C T on entry. function [ X ] = rtrglydt ( A , E , C , blks ) if 1 N blks then X = trglydt ( A , E , C ); elseif C is symmetric % Split ( A , E ), and C (all by rows and columns) X 22 = rtrglydt ( A 22 , E 22 , C 22 , blks ); C 12 = axb ( - A 12 , X 22 , A T 22 , C 12 ); C 12 = axb ( E 12 , X 22 , E T 22 , C 12 ); X 12 = rtrgsyl ( A 11 , A 22 , E 11 , E 22 , C 12 , blks ); X 21 = X T 12 ; C 11 = syr2k ( trmm ( A 11 , X 12 ), - A 12 , C 11 ); C 11 = syr2k ( trmm ( E 11 , X 12 ), E 12 , C 11 ); C 11 = axb ( - A 12 , X 22 , A T 12 , C 11 ); C 11 = axb ( E 12 , X

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

View Full Document
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