MAT-51626 Matrix Algebra 2
Spring 2010

Lecturer: Robert Piché
Lectures Mondays 9-11 and Tuesdays 14-16 in Tb214
Textbook: James W. Demmel, Applied Numerical Linear Algebra, SIAM 1997, online, errata
Course requirement: homework problems or take-home exam.

1. Introduction, norms, floating point arithmetic, backward stability

Textbook chapter 1; video, slides
Homework (due 9.3 at 14:15): chapter 1 questions 1, 3, 7, 10, 11, 14 part 2 (i.e. equivalence of inf norm and 2-norm), 16 part 11.

2. Linear equations: gaussian elimination, SPD matrices

Textbook sections 2.3, 2.7.1; video, slides
Homework: chapter 2 questions 1, 7, 11, 13 parts 1-2, 17, 19 bullet 2, 20 parts a-b.

3. Linear equations: perturbation analysis and backward stability

Textbook sections 1.3.2, 1.6, 2.2, 2.4.2; video, slides
Homework: chapter 1 question 20; chapter 2 questions 9, 10 (n-by-n only), 12, 19 bullet 1 (using lemma 2.1) and the following simplified version of question 14:

gecp implements gaussian elimination with complete pivoting in Matlab. Compare this algorithm with Matlab's lu function (which uses gaussian elimination with partial pivoting) by testing single precision matrices A of dimensions up to 30. Test well-conditioned matrices as well as ill-conditioned matrices; generate them either by following the advice in Question 2.14 or by using Matlab's gallery('randsvd',...).  Report and discuss three indicators: the pivot growth factors g, the factorization backward error norm ||E||max=||P1*A*P2-L*U||max, and its estimate n*e*||abs(L)*abs(U)||max from section 2.4.2, where e=roundoff unit (=half of machine epsilon).

4. Linear equations: practical error bounds

Textbook sections 2.2, 2.4.3-4; video, slides
Homework: pdf

5. Linear equations: blocking algorithms, banded and sparse matrices

Textbook sections 2.6 (except Strassen), 2.7.2-4; video, slides
Homework: pdf

6. Least squares: normal equations, QR, backward stability

Textbook sections 3.1, 3.2 (except SVD), 3.4; video, slides
Homework: chapter 3 questions 1, 3 parts 1&3, 4, 5, 6, 15 (except SVD)

7. SVD, perturbation analysis of LS, pseudoinverse, regularisation

Textbook sections 3.2.3, 3.3, 3.5 (up to defn 3.2);  slides, demo1.m, demo2.m, demo3.m
Homework: chapter 3 questions 3 part 2, 9, 11, 12, 14 (square A only), and redo p.114's table for Matlab's gatlin.mat image (note erratum: sigma_{k+1}/sigma_k in the heading of column 2 should be sigma_{k+1}/sigma_1)

8. Eigenvalue problems, perturbation analysis

Textbook sections 4.2 & 4.3 (except Defn 4.4, Prop. 4.3, Thms 4.3, 4.6, 4.7); video, slides
Homework: chapter 4 questions 2, 3, 9 (replace min by inf), 13, 14

9. Jacobi's method

Textbook sections 6.3 and 6.5 (except Gauss-Seidel, SOR, Chebyshev, SSOR); video, slides
Homework: pdf , lap2d

10. Preconditioned Conjugate Gradient method

Textbook section 6.6; video, slides
Homework: chapter 6 problems 11, 12 (Lanczos, replace y1 by b), 13, 14, and redo question A3 of set 9 using the conjugate gradient method and without forming A.