Module review
Weightage
Team-based component
- Project proposal: 10%
- F2F project updates: 15%
- Project presentation: 20%
- Project report: 30%
Individual component
- Project team member feedback: 5%
- Quiz: 10%
- Attendance: 5%
Lecture content
The module covers frequent itemsets, classification, regression, graph mining, data streams, clustering, and anomaly detection. The content is fairly standard and not too difficult (roughly the same level as an undergraduate course if you’ve already come across these concepts before). It mostly feels a bit touch-and-go at times, but the prof does try to focus on how these algorithms and techniques scale to the real world, where you can’t simply load all your data into memory etc. He covers how the algorithms still work using approximations and estimates, e.g. how K-nearest neighbours can be applied when you can’t fit all the data points in memory. I think this is the only major difference from how these topics are usually taught.
There are participation marks for attendance, but the prof has uploaded all the recordings and slides since week 1. So if you don’t mind losing the attendance marks, you can just watch the recordings instead.
Group project
The group project is a Kaggle competition on two temporal datasets. Deep learning is not allowed. Both datasets need to share the same pipeline, and your final marks come from your ranking on the public and private leaderboards. I found the project just okay, nothing particularly interesting. Many teams score well on the public test cases but do poorly on the private ones because of overfitting. Most people end up with roughly the same score, so the bell curve is quite steep.
Quiz
The quiz is multiple choice and very easy if you know the content well. Sometimes the wrong options are also very obviously wrong. Most people probably scored full marks.
Overall
I don’t think the module is very well managed. There doesn’t seem to be much effort put into facilitating discussion or improving the learning experience. Personally it felt like the prof is just there to get through his lectures. My group had to email the prof and TAs several times for clarifications or issues we ran into, but most of the time we got no reply (and when we did, the responses were carelessly written, with obvious spelling mistakes etc.)
I didn’t have a great learning experience with this module, but it is a module you can complete with relatively low effort.