Posted on

NUS Math Module Review: MA4272 Mathematical Tools in Data Science

Concepts Taught:

Dr Ji Hui: Discrete Fourier Transform, Convolution, Basic Signal Processing Techniques, Conjugate Gradient Method

Dr Zhao Gongyun: Modelling for Optimization, Proximal Gradient Algorithm, LP Relaxation

Dr Zhang Louxin: Machine Learning, Principal Component Analysis, Data Visualization


Dr Ji Hui: He does not have any notes of any kind, and you are required to be diligent in writing down your own notes from the whiteboard. He is a fairly decent teacher, not the best, not the worst either.

Dr Zhao Gongyun: Dr Zhao always gives excellent notes, regardless of the module he is teaching. One fine example is Game Theory. However, his teaching fails to convey the message at times. Feels like he is beating around the bush too often and not getting to the point.

Dr Zhang Louxin: His notes are decent, but his lectures are a snoozefest due to the way he speaks. Not a great teacher, but thankfully, the concepts he teaches are fairly easy.

In general, this was quite a poorly taught module, so you will need to be independent.

Tutorials (5% Participation)

9 tutorials, roughly 4-5 questions per tutorial. Questions range from simple to difficult. As with every module, please practice your tutorials well for your exams.

Assignment (45%, 15% per lecturer)

3 Assignments, set by each of the lecturers on the topics they have taught. Dr Ji Hui’s was the hardest, you had to use MATLAB to restore a signal using DFT. Dr Zhao Gongyun’s was fairly easy, 1 question on modelling and solving with Gurobi in MATLAB, and another question on proximal gradient method. Dr Zhang Louxin did not set a programming assignment, but instead, a very tedious set of 5 questions from machine learning, which was annoying and had no learning value.

Exams (50%. Open Book)

Be prepared for a difficult exam. Each lecturer sets 2 questions each, so 6 questions in all. Dr Ji Hui tend to set the same questions as he does in MA4268 Visual Data Processing, so get the finals paper from V.D.P and try some of them. Dr Zhao Gongyun set a very difficult modelling problem (I swear, it was nothing I have seen before), and 1 question on proximal gradient algorithm. Dr Zhang Louxin set an essay question on the benefits of machine learning (LOL wut) and also another question on clustering (which takes you 30 minutes to do, because of all the manual work involved).

Head over to our Shop for more module content!