Module Review: EE3731C Signal Analytics

Introduction 

This was taken as an evening module during the pandemic. Other than zoom lessons, the format is largely the same. It's divided into two parts, classical signals AND probability/pattern recognition. It's math heavy. If you have a general dislike for maths; i.e. matrix, fourier transforms, integration, probability and other exciting mathematical techniques, please avoid this module!!

 

Beyond the mathematical rigour, the second part is algorithmic in nature, just like in year 1 linear algebra. You need to perform calculations step by step with great care as one mistake will screw up the whole question. But I must qualify here; my batch allows the use of Matlab (because of e-exams), significantly easing matrix manipulations and the calculation of convolution. Of course if you cannot use Matlab, your scientific Casio calculator should allow matrix manipulations as well. 

Lectures/Tutorial

Package lectures are held once a week. Prof Thomas will go through difficult tutorial questions and questions from students before starting the lecture proper. I must confess I didn't attend live lessons as it's recorded but he is a good lecturer, no worries here. He is quite funny too, joking about (poor) lecture attendance or anything under the sun. The certain part of the content can be quite abstract and you may be confused at times. 

One advice for this module is to ignore those super technical concepts and just whack the questions, even if you don't know what you are doing. This probably not the right way to learn but errrrrm, but who cares right? As long you can do the questions?

 

For example there's this thing called MAP (maximum a posteriori) , which is explained in very complicated terms in the lecture. But these explanations are just a facade. Just take the literal meaning of it a.k.a. maximum probability; and chose cases with highest term and you will magically arrive at the same answer. Other stuff includes the Metropolis Algorithm or the Transitional Matrix or PCA Analysis (topics from the second half), even if you don't fully get the purpose of them, just forcefully practice it as the steps are always the same and guaranteed to appear in exams. Very algorithmic indeed. 😅

But of course on the flip side, some questions like random walk process, LTI system properties and part 1 stuff requires some understanding in the concepts to tackle, but these questions are in the minority in final paper. 

A little snippet to scare off mathematically challenged farrrr away!!

Programming Assignment

The programming assignment (20%) can be done within a day. It's basically Matlab programming to decipher encrypted text using substitution decoder. A written report is expected. You need to crack your brain a little but it's not difficult. If you suck at programming, you probably can discuss with peers. Tbh... There's no room for error. I think most students get more than 95% for it. So don't fall behind. 

Midterm/Final exams

Midterm tested part 1 of the module; classical signals. There are quite a number of questions to explain this and that and if you don't know what's going on, that's not very good... 

The justification for a difficult midterm exam was because, it's an open book paper due to the pandemic measures (normally, just a single A4 cheatsheet). IMO that's a false pretence because open book assessments will result in students wasting time searching for content instead of summarising it on a cheat sheet. This is ultimately true because everyone died and lecturer said sorry!!, and promised to "re-calibrate" the difficulty for the finals (which he did!!). The median was 30/50 and I got slightly above it. One big negative is past midterms are not provided, so we are going in blind, not too sure what to expect. 

One interesting fact mentioned a few times is that in certain years, the final paper was set to be very difficult because the department bureaucracy complained that the previous year's exam was too easy. And what I observed is if the preceding year was difficult, next year's paper will be way easier. It seems like there's a predictable cycle going on.. Hmmm... I'm kinda lucky as I'm part of the "easy year" (but I had to contend with a difficult midterm). Because of the "re-calibration" from the killer midterms, our batch got lucky for finals! 😅😅😅

The questions are doable if you attempted past year papers. Very doable. 3 out of 4 questions are from the second part of the module, so just practice, have good sleep and it shouldn't be an issue. 

Conclusion

This module is interesting but can be difficult to understand. The math can be insane at times and certain questions will require deep higher order thinking that I lack 😂 (i.e. random step process). The key bell weather to see if you're suitable for this module is your EE2023 grade. Another consideration is this is held 3 hour-once a week (and recorded)... So if you're doing internship or just do not like attending lessons, this will be perfect. 

If you read my past reviews, you probably can infer I place greater emphasis (and do better) on spamming practice papers and rote learning rather than fully understanding what's going on. If you're someone like me, this module is a great fit. If you like nitty gritty things like BJT, MOSFET and all those exciting stuff, this module may not be suitable. Consider something like EE3431C (which I have no plans on taking 😉). Prof Thomas is a good lecturer in both teaching and joking around. I have no regrets taking this module. 

Sorry for the long wall of text because I got very bored. 

My rating: Difficulty: 3/5 Workload: 2.5/5 Teaching staff: 4/5 Overall: 4/5

Graded components: Programming project: 20% Midterm test: 20% Final examination: 20%

Expected grade: B+ Final grade: B+