Skip to menu Skip to content Skip to footer
Course profile

Numerical Linear Algebra & Optimisation (MATH3204)

Study period
Sem 2 2024
Location
St Lucia
Attendance mode
In Person

Course overview

Study period
Semester 2, 2024 (22/07/2024 - 18/11/2024)
Study level
Undergraduate
Location
St Lucia
Attendance mode
In Person
Units
2
Administrative campus
St Lucia
Coordinating unit
Mathematics & Physics School

At the heart of most modern data scientific methods in general, and machine learning in particular, lie computational techniques involving matrices as well as numerical linear algebra and optimisation algorithms. In this course, students will learn about the theory and practical aspects of many fundamental tools from matrix computations, numerical linear algebra and optimisation. In addition to classical applications, most examples will particularly focus on modern large-scale machine learning problems. Implementations will be done using MATLAB/Python. The students will also be exposed to cutting-edge developments including randomised variants of many classical deterministic methods. Students will be taught a range of analytical and algorithmic tools that are employed in research and industry, such as various matrix types, their properties and factorisations, iterative algorithms for matrix computations such as Krylov subspace methods, various eigen-solvers, elements of convex and non-convex analysis, derivative free as well as first and second-order optimisation methods, constrained and unconstrained optimisation algorithms, and introduction to non-smooth and stochastic optimisation.

Numerical Linear Algebra and Optimisation are fundamentalᅠtoᅠall areas of science, engineering and data analysis that involve computational techniques. In this light, MATH3204 will provide an opportunity to develop the basic and fundamental skills needed to tackle modern computational and data analysis problems.

MATH3204 will cover the following fundamental concepts and techniques:

  • Various matrix types and their properties
  • Various matrix factorisations
  • Iterative algorithms for matrix computations such as Krylov subspace methods
  • Elements of convex and non-convex analysis
  • First and second-order optimisation methods
  • Constrained and unconstrained optimisation algorithms
  • Applications in Machine Learning and Scientific Computing

Course requirements

Assumed background

Introductory courses such as MATH2000, MATH2001,ᅠMATH7000 or MATH7502 will cover parts of the prerequisite background. The computational techniques learnt as part of COSC2500 or MATH3201, though not mandatory, are considered helpful background knowledge. Students should have basic computer skills and some level of programming experience in MATLAB/Python is required.

Prerequisites

You'll need to complete the following courses before enrolling in this one:

MATH2000 or MATH2001 or MATH2901 or MATH7000 or MATH7502

Recommended prerequisites

We recommend completing the following courses before enrolling in this one:

COSC2500 or COSC7500

Incompatible

You can't enrol in this course if you've already completed the following:

MATH7234 (co-taught).

Course contact

Tutor

Mr Alexander Lim

Tutor

Mr Jacob Westerhout

Tutor

Mr Joseph Wilson

Course staff

Timetable

The timetable for this course is available on the UQ Public Timetable.

Additional timetable information

All classes will be conducted on campus. Consult your personal timetable for times and locations. Students are expected to attend these sessions inᅠperson unless they have a valid reason for being unable to attend (such as illness).ᅠ


Tutorials start in week 2. If a tutorial session falls on a public holiday, there will be no make-up session. Instead, students are welcome to attend any other session that fits their schedule.