Comp 205:

Syllabus, Spring 2000

Goal:

Teach error analysis and efficiency analysis via a driving problem that requires geometric algorithms. COMP 205 will address this goal by touching on aspects of computational geometry (convex hulls, triangulations, Voronoi diagrams), numerical analysis (SVD, Gaussian Elimination, forward and backward error analysis, interpolation), and mathematics (linear algebra, differential equations). A detailed syllabus for the whole course will be distributed on Tuesday, January 18, 2000, once I've seen the survey responses.

Methodology:

This course has traditionally been taught with simulation of Newtonian mechanics as the driving problem that motivates each topic covered. I will relax this slightly to include some puzzle problems drawn from other applications as well that all have a common thread of ``data in motion.''

Textbook:

Because this is a hybrid course, there is no single text that covers the material. The chosen textbook is a Dover reprint of R.W. Hamming's Numerical Methods for Scientists and Engineers. I really like its motto: ``The purpose of computing is insight, not numbers.'' And it's under $20. I will supplement this with other material, including text from a book that I started writing with Goodrich and Guibas. You are also expected to have a more modern textbook from whatever numerical analysis course you have taken as the prerequisite for this course. A recent one that has good coverage is C. Posrikidis' Numerical Computation in Science and Engineering, Oxford, 1998.

Problems:

Assigned problems are due on Tuesdays, either in class or under my door. (Tuesday extends until midnight. Once grading has begun, the grader will generally not consider late assignments. )

Programming:

Assignments that involve programming are designed to be done in MATLAB if they are primarily numerical, or Java/C/C++ if they involve data structures. You are welcome to use any programming language or tool that still involves you in the learning process. (I.e., if a problem asks for the Jordan normal form of a matrix, and your tool can compute JordanNormalForm(A) without you knowing what it is, then you better do it the hard way instead. Ask me if you are in doubt.)

Collaboration:

Collaboration on the puzzle problems is encouraged. Good scholarship requires that all collaboration must be acknowledged. Thus, if you collaborate on a solution, I expect that you name your collaborators at the top of the page.

Evaluation:

The midterm will be in-class and the final a take-home exam. The course evaluation will be based on the following:
Assignments (10) 40%
Midterm 25%
Final 35%

Fun:

One of the rewards of research is the ``Aha'' of gaining a new insight. Even though COMP 205 is possibly the class that students least look forward to taking, I hope that in this term we will have the fun of new insight. (So please give me feedback on whether the course is moving too fast or too slow.) As in research, one must invest time to gain familiarity with one's tools (computational and mathematical) to be able to play with them creatively and effectively.

Puzzle #1

Concept: IEEE 754 floating point

Do by lecture on Jan 18
Last updated: January 18, 2000