This course is an elective for the Data-centric Computing Stream.

This course is of interest to students in any discipline who are dealing with computer solutions of numerical linear algebra problems. Matrix computations may be found useful in disciplines such as engineering, statistics, physics, optimization, operations research, computational chemistry and signal processing.

Prerequisites:  COMP 3731

Availability: This course is usually offered once per year, in Fall or Winter.

Course Objectives

An introduction to the techniques of numerical linear algebra. Emphasis is placed upon developing the most recent and reliable algorithms. The stability of these algorithms as well as the sensitivity of the problems they solve will also be studied.

Representative Workload

• Assignments 50%
• In-class Exam 20%
• Final Exam 30%

There will be about nine assignments given throughout the semester. Programming assignments are chosen to illustrate topics discussed in the lecture material and can be written in any programming language unless otherwise specified. Non-programming problems are also assigned.

Representative Course Outline

• An introduction to necessary topics of linear algebra
• Systems of linear equations
• Scaling
• Iterative refinement
• Estimating the condition number of a matrix
• Introduction to pipelining and parallel matrix computations
• The linear least squares problem
• The symmetric and unsymmetric eigenproblems
• The singular value problem of a matrix

Page last updated May 24th 2021