I enjoyed this subject overall. The lecturer for the first half of the course, Chris, is just amazing, she just knows how to teach. The lectures covered the high level theory for different ML models, with 4 lectures dedicated to neural networks and large language models at the end. Tutorials were super useful and engaging (would recommend Jey) and the two assignments were not too heavy and very good practice. Overall I walked away feeling like I gained lot, from the theoretical properties of different models to the actual practice of implementing a ML model. I feel I can go on to use the skills I learnt for personal projects now (exactly what I'd want out of a subject as a data science major). The workload was also very manageable with 2 lectures a week.

As someone who has little statistics background, this course had a steep learning curve. That being said, the challenge was really enjoyable. Material is presented in an interesting and engaging way, both lecturers are great (though Ling moves really fast). Tutorials were genuinely helpful and interesting which is uncommon in CS i find. Best part of this subject is that all the content is taught at a high level. Algorithms are demonstrated (and examined) but this really just serves to deep your understanding of all the models and concepts. The final exam is mostly short and long answer questions. Assignments are code-based but no serious programming skill is necessary, again, they test more high level understanding. Very cool.

Overall content was interesting, but I wished more focus was put on the later topics. The subject started out a bit too slowly and reiterated some concepts from Elements of Data Processing in the first few weeks. The later weeks involved more interesting topics like hidden Markov models, mixture models and deep learning. Part of the second assignment involved a Kaggle in-class competition, which was quite a fun experience. I would recommend this subject as a starting point for those interested in machine learning or data science.

Subject content was alright overall with a strong focus on theory rather than maths. The lecturer Kris is great, absolutely loved her. The assignments weren’t too exciting, where the first one was building a Naive Bayes model from scratch and the second one was writing a report for a sentiment analysis task. If you really want to learn machine learning there are way better courses online (eg Andrew Ng). I wanted to give this subject a try but now that I’ve completed it I wish I took another subject instead.

Subject has no maths pre-req, but they did try to attempt to cover "Linear Statistical Models" in 1 lecture. The usual lecturer is great, given that the maths is not a pre-req (at least tries to make it interesting). However, he was sick for a vast majority of the time we took the subject so my experience may be a bit more biased to the worse side. A lot of content was attempted to be covered and felt a bit too ambitious given that the maths was not pre-req. First assignment is a joke if you have done maths, but the second is a lot more interesting. Quoting my tutor: "this subject is a money grabber because no one would take this subject if Probability was an actual pre-req". Next year, it will be run under a different lecturer who has taken a lot of the feedback into account. I can now say that I will highly recommend this subject next year !

Content's pretty interesting with some enjoyable projects. Jeremy who lectured before becoming ill mid semester was really good, and showed genuine enthusiasm in the content. From what I've seen with EoDP and now this subject, tutes and labs are hit or miss. I was lucky to get a good lab demonstrator who seemed interested in teaching though my tutor literally did the tutorial for us like it was high school or something. The focus of the subject is a bit obscured at times, as they included these large number of slides with mathematical formulas yet were never properly assessed as this is not a maths subject, which begs the question for their inclusion in the first place. The lecturer told us that they're like a necessary evil, where you need to know some basics of maths to understand some of the content which is fine, but including multiple slides explaining hyperplane optimisation with grandiose formulas only for us to never use it is a bit weird. Overall I enjoyed this subject even though the exam was VERY crammed for 2 hours. I look forward to using the skills I've learnt to work on personal data projects as really that's what you should be getting out of these subjects.