Posted on

NUS Math Module Review: MA6252 Topics in Applied Mathematics 2

Expected grade: B

Actual grade: B+

Assessment and workload:

– 40% Project 1 (20% report, 20% presentation)

– 60% Project 2 (30% report, 30% presentation)

– No Final examination


– This module is very difficult, especially for students with insufficient background in computer science and topics related to machine learning. The module begins with a crash refresher course on numerical linear algebra and probability. It then suddenly escalates to topics in deep learning. The main forms of deep learning were introduced, essentially neural networks, convolutional neural networks, random boltzmann machines, recurrent neural networks and approximation theory for neural networks. There is simply too much breadth.

– Prof Yang is obviously pretty knowledgeable in this area and he trivialises many things. Delivery of this module is pretty inadequate as he introduces many big ideas, but with insufficient elaboration and emphasis on the fundamentals. He also assumes a high level of mathematical background from us as he uses advanced concepts without elaboration. Prof Yang tries to encourage class participation but failed to do so, as he does not give clear directions on the kind of answer he wants. This causes the student to feel embarassed or appear dumb in front of the class, which causes a vicious cycle of people being unwilling to answer any questions he poses.

– There are no tutorials for postgraduate modules. This is a seminar style module.

– Project 1 involved us having to heavily modify a code on Github on the fundamental aspects of deep learning, and apply regularisation techniques taught. My group choose logistic regression and we wrote a report and presentation based on our experiments.

– Project 2 involved us having to pick an area of deep learning, and try to reproduce results from a research paper or come up with something revolutionary. My group choose Maxout networks and we wrote a report and presentation based on this.

– This year, Prof Yang decided to give us the power to judge our peers’ reports and presentations, and gave us the power to decide 50% of our peers’ grades. While this is a good idea because of the difference in benchmark between the professor and the student, this idea was poorly executed in my opinion. We had to give 2 groups A and 2 other groups B for the report. Next, each of us had to grade 12 other students based on their presentation skills as well as Q&A, and we were allocated 5A 5B and 2C. After all the presentations, I feel no student deserves to get C due to the effort everybody put in to prepare for the presentation. I felt so bad having to give 2 people C because it definitely will affect their final grade. In addition, Prof Yang’s Q&A is unfair because he asks questions of varying difficulties. The student who failed to answer a tough question may have gotten a bad grade from their peers because they were unable to answer to Prof Yang’s satisfaction.

Head over to our Shop for more module content!