AMSC 661 / CMSC 661 Scientific Computing II (Section 0101)

Information for Spring 2005

oleary@cs.umd.edu

When and Where: TuTh......9:30am-10:45am (CSI 3118) (CSI is the Computer Science Classroom building, attached to A.V. Williams and behind the Wind Tunnel.)

New! End of semester announcements:

Homework 4 grades will be emailed after all submissions are received.
Submissions of Homework 4 after Noon, May 17 will receive no credit.
The late penalties for the project are the same as for homework, but no credit will be given after noon on May 20.
Updated grades will be posted on May 13 or 14. Final grades will be posted when I finish grading projects.
The last office hours of the semester will be Thursday, May 19. I will also keep office hours on Tuesday May 31.
I will be out-of-town, attending the Householder Meeting on Numerical Linear Algebra, May 22-27.
I will hold unclaimed papers through the fall semester, so you can arrange to pick them up. I will officially be on sabbatical, though, so my schedule on campus will be irregular.
Best wishes in the remainder of your academic career and beyond!

Office Hours: Tuesday 8-9:15, Thursday 11-12, and by appointment, in AVW 3271.

Topics: Numerical solution of partial differential equations using finite element and finite difference methods. Solution of sparse systems of linear equations. Wavelet and fast transform methods.

Textbook: Stig Larsson and Vidar Thomee, Partial Differential Equations with Numerical Methods, Springer 2003: ISBN 3-540-01772-0. The remaining third of the course material (sparse matrix computation and FFT/wavelet) will be taken from other reference books.

Known typos in the textbook

Programming language: We will use Matlab for the programming assignments.

Grading:

New! Final Grades

Quizzes: 140 points. Quizzes will be 20 minutes long, scheduled every third class, beginning with Class 4. The lowest two scores will be dropped.

Assignments: approx. 160 points. This will involve a mix of written work, programming, and analysis of results. At least 2 weeks will be given for each assignment.

Term project: 100 points. Each student will be given a manuscript. The project will be to write a review of the manuscript and test the numerical method that it discusses.

Exams: none.

Questions? Please contact me.

Basic Information:

Course Information and Syllabus (pdf)

UMCP Code of Academic Integrity

Information about the Unix cluster and your account on it

Make sure you change your password when you get your new account. "Use of the passwd, chsh and chfn commands require your session be fully authenticated. This means you must login to the cluster in some way that requires you to supply your password (such as by the telnet command). Note that a secure login such as ssh (or slogin) does not give you full cluster authentication. If you login to the cluster by a method that does not give you full cluster authentication use the dce_login command to authenticate yourself. You will be prompted "Enter Principal Name:" which you answer with your cluster login id and "Enter Password:" which you answer with your cluster password."

Survival Guide for Scientific Computing

Guidelines on Writing Scientific Computing Software

Documentation for Matlab's PDE Toolbox

Lecture Notes (pdf)

Introduction

Solution of ordinary differential equations, boundary value problems

Notes Part 1: Some theory

Notes Part 2: Computational methods

Solution of elliptic partial differential equations

Notes Part 1: Some theory

Notes Part 2: Computational methods (Reposted 03/07)

slit.m Adaptive mesh example from class

Notes Part 3: Eigenvalue problems

Solution of sparse linear systems of equations

Notes Part 1: Direct methods

spar5.m Demo from class, reordering the 5-point operator using various algorithms. (The documentation is not good.)

Notes Part 2: Iterative methods

Notes Supplement: Convergence of SIMs

Notes Part 3: Convergence of Krylov methods and multigrid

Solution of parabolic differential equations

Notes on initial value problems for ODEs

Notes on theory for parabolic problems (reposted 04-19)

Notes on numerical methods for parabolic problems (reposted 04-21)

Solution of hyperbolic differential equations

Notes on theory for hyperbolic problems

Notes on numerical methods for hyperbolic problems

Fourier transforms, wavelets, and fast multipole algorithms

Notes on Fast Poisson Solvers (reposted to change "j=" to "k=" on the last page)

Notes on Transforms and Wavelets

Multipole article by Sun and Pitsianis (This link should work from the umd.edu domain.)

Notes on the Fast Multipole Method

Scribed notes
Extra credit: For each of the last 4 lectures in the course, I would like to post lecture notes to supplement the slides. These would include the examples from the board and extra comments made in class. The set up:

The schedule:
May 3: Vincent Chan, Brianne Omatick, Fei Xue
May 5: Nitin Madnani, Mian Li, Joanna Pressley
May 10: Steve Clark, Tom Salter, Hyunyoung Song
May 12: Charles Martin, Yiming Zhai, Ryan Harvey

Notes will be due, either by email (pdf) or by hardcopy in my office, 24 hours after the lecture (11am). Put your name on the notes.

Notes can be handwritten as long as they are legible. I will scan handwritten notes into a pdf file, or you can email a scanned pdf version to me.

If you are scribe and something from class was not clear, it is fine to ask me or a fellow student for help, but credit any student who helped you.

Extra credit will be up to 8 points per lecture, depending on completeness, correctness, and punctuality.

If you have already scribed one lecture, you won't be chosen for a second one unless there are not enough other volunteers.

For each lecture, I'll post one or more sets of scribed notes. The notes will serve as a resource to help students prepare for quizzes and homework, but there will be no extra credit points for finding errors in the scribed notes.

Homework

Homework 1: Due February 22.

The assignment

finitediff1.m

Frequently Asked Questions about the homework

Solution: (Average grade: 42.22)

a.m

c.m

f.m

trueu.m

Homework 2: Due March 15.

Tutorial information: To make up for some of the time lost to snow, I'll let you explore the tutorial information on your own, rather than doing an in-class demo. The tutorial comes in two parts:

Part 1: Recall the problem we discussed a bit on day 1 of class: torsion in an elasto-plastic bar. Here is

A sample assignment based on it. (Ignore the last 5 pages of this pdf document, the "Fitting Exponentials" piece.)

Solution code for it. (Looking at problem1.m is probably enough.)

A partial solution for it. (Ignore the first 4 pages of this pdf document, the "Yaw, Pitch, and Roll" part.)

Read this material, and get comfortable with the use of the relevant piece of Matlab's PDE Toolbox by playing with the solution code.

Part 2: The link to the entire set of documentation for the PDE Toolbox is given above, under "Basic Information". It is a good idea to play with the GUI.

getpetuc.m This program exports geometry and function information from pdetool to Matlab's usual workspace. You can use it, for example, if you save a ".m" file from one of your GUI sessions and want to run it again and compute the error norm.

The assignment

Frequently Asked Questions and submission instructions

Solution: (Average grade: 25.81)

Notes on the solution.

Homework 3

Frequently Asked Questions

Solution: (Average grade: 36.22)