Response to 2014/15 survey feedback.
The MLPR course is running in 2015/16.
We have taken note of the feedback provided through the Student Online Survey. We have the following responses/clarifications:
This is a challenging course. The level of maths and technical detail is necessary for those wanting to research and develop new machine learning methods in future. Those that have the right background and work hard on it usually find it rewarding. However, unfortunately every year people take this course without the right background, and a larger fraction fail than many Informatics courses. The DRPS contains a brief list of the mathematics background assumed for this course. The course website gives more information, some reading and a self-test problem sheet.
Volume of material: This course moves fast. A majority of survey responses thought the course was managable, and we don't want to make the course a less useful preparation for research and development of machine learning methods. However, not every student in the School will have to the background to cope with the pace of this course, and they should choose other options. We intend to mark a little more of the material as non-examinable this year. However, the course will not be substantially different from last year.
MSc students: while this is a “core course” for the machine learning specialism, we would like to remind you that the specialisms are advisory. If you haven’t seen any machine learning before, and don’t have the strong maths background we assume, you would be better taking only IAML and other more applied courses in Semester 1.
Iain Murray and Chris Williams, September 2015.