News 

Nov 21 A5 solutions are online Out of several excellent submissions, i picked two P2 solutions for you to look at.

Nov 8 There were some confusing typos in the original A5 handout. Also, I clarified some things in P2. So re-download from CMS the corrected version of the original document!

Nov 6 A4 Solutions online

Oct 24 A4 Corrections/Comments. For P1, try to make your function efficient by minimizing the number of times that Matlab is asked to "expand" the sparse array that you are using to track fill-in. In P2, there is a typo in Step 4 in the seminormal equation framework. Instead of Re = y is should be e = P'*z. Also, it is OK to set up the 4-by-5 block matrix in sparse format and let Matlab do all the work.. P3. The matrix A1 is dense but data-sparse. You will want to supply eigs a function that does A1-times-a-vector.

Oct. 19 A3 solutions are online . Regrade requests must be received via email by Tuesday Oct 4.

Oct 4. Use the "\" operator to solve triangular systems. The triangular systems that come up in the assignment are sparse and Matlab knows enough to exploit the structure.

Oct 3   A3 clarifications. In P1, the numbers you report should be reliable through one decimal place. That is to say, if you re-run your script a second time then the corresponding entries in the two tables should not differ by more than 0.1. P2. The U and V in the first line should be Y and Z. P3.The values used for s in the test script will be modest, e.g., s = 5.

Sept 30. A2 Solutions online . Regrade requests must be received via email by Tuesday Oct 4.

Sept 20 Typo in P3.    inv(Sn) = (2/(n+1))*Sn       not      inv(Sn ) = (2/n)*Sn

Sept 14. Typo in the write-up of A2. In P2, there is a missing sqrt(-1) in the specification of Twiddle.

Sept 14 A1  Solutions/test scripts are now online If you need an elaboration on how I graded your submission, send an email, but please, run the relevant test scripts first so that you can measure your solution against mine.

Aug 25. You must be registered in the course to log into CMS, the system that will be used for the submission and grading of assignments and the distribution of handouts. (Click on the CMS link above.)  If you add the course, then send me an email and I will enter you into the system. Right now, Chapter 1 of GVL4 is available through CMS. In about a week after the enrollment settles down, hard copies of GVL4 (Chapters1-9) will be distributed at nominal cost. Chapters 10-13 will be distributed as the semester progresses.

Aug 24. Worried about your linear algebra and/or Matlab background?

Some Linear Algebra References:  Matrix Analysis and Applied Linear Algebra (C. Meyer), Linear Algebra and its Applications (G. Strang)

Some Matlab References: Insight Through Computing: A Matlab Introduction to Computational Science and Engineering (Van Loan and Fan), Getting Started with Matlab 7  (Pratap), Matlab: An Introduction with Applications (Gilat), Mastering Matlab7 ( Hanselman & Littlefield)

Description:  An m-by-n matrix is data-sparse if it can be represented with << mn parameters. Sparse matrices are data-sparse, but so are Toeplitz matrices, semiseparable matrices, sparse-plus-low-rank matrices, and a host of “fast transform” matrices. We will survey methods for linear equation, least square, and eigenvalue problems when the matrices involved are big and data-sparse. The LU, Cholesky, and QR factorizations will have a central role to play as will as the SVD and various eigenvalue-related decompositions. Basic, dense-matrix algorithms for these reductions will be covered insofar as they are needed for the big-n, data-sparse problems that are the focus of the course.

Prerequisites: A course in scientific computing (or equivalent) , solid background in linear algebra, and working knowledge of Matlab (or equivalent).

Meeting Time and Place: MW 2:55-4:10,  Upson 315.

Grading: Seven Matlab assignments (70%) and a Final Project/presentation (30%).

Syllabus

 Top

Assignments and Solutions

 Top