Do not do this course if you are interested in programming and incorporating that knowledge with machine learning - although the course gives the option to use Python, next to nothing is taught in Python and almost no examples are given in it. I took the course with the intention to use Python but quickly realised using Python is essentially shooting yourself in the foot and making it far harder than using Matlab which they give all examples in and teach with. The lectures were a nightmare for me, Marcus is the most confusing lecturer I have ever come across and I struggled to get anything out the 2hrs a week he spends umming and ahhhing as he scrolls through a textbook.
If you plan to do well in the course and actually learn anything you will have to teach yourself or constantly ask questions throughout the prac sessions as well as read through a lot scattered textbooks - which require a strong maths background to understand.
The assessment is in the form of weekly homeworks which vary between comically easy to incredibly difficult - made even harder by the fact you cannot ask questions about them during the pracs or online, meaning if there is a concept you do not understand your only option is to read through textbooks as the lectures rarely cover the topics in enough detail to be useful.
The 3 prac assessment aren't too bad and should be easy marks if you do enough prep work.
Overall, I would not recommend this course unless you have a genuine interest in maths and are willing to spend a lot of time self-teaching, if you are expecting a 4th year software course where you will learn real-world application of machine learning as you might in the workforce / a non-academic scenario then avoid at all costs. I don't think I wrote a function with more than 20 lines in it the entire course and you will be using some ancient/irrelevant software.
Semester 1 - 2017
On 1.5 speed
CONTENT: I am an electrical student and took this elective in my second year out of interest. The content of the course is very diverse: lots of different topics are covered. This is a problem in the sense that they mostly aren't covered really in-depth, but is also good in the sense that it gives you a very good overview of the discipline of machine learning.
Textbook reading is required (online textbook resources are supplied for free), and the content in them is quite technical. You will need to have a decent knowledge in statistics and it would be good to have done a stats course, such as STAT2202 before. Calculus is essential.
The content covers regression, classification, density estimation, dimensionality reduction, non-parametric methods, clustering methods, neural networks, support vector machines, Bayesian networks, Bayesian learning and Gaussian processes, along with overarching themes and topics common to all areas of machine learning like overfitting, underfitting etc. Neural networks are probably the topic that is covered the most in depth, which most students would probably appreciate.
LECTURES: The lectures weren't very good. Marcus doesn't give you the impression of really knowing that much about the topics and generally just stumbles his way through the textbook, some notes, or papers on the screen. He does seem unprepared quite often. Sometimes he reads the textbook and then gets confused about it. He should be making sure his thoughts are all ordered before the lectures. At the same time, it's probably not a good idea to skip the lectures. The lectures come with pre-reading - I would make sure to do this.
PRACTICALS: The practicals were generally good. They may be hard for some to complete in the two hours, but this isn't a problem. You can complete them at home by using the student license for MATLAB. Prac questions are chosen randomly as assignment questions which must be handed in every few weeks. I would not leave these questions to the last minute as some of the questions might require time for you to understand and work through, especially if you haven't been paying attention to the lectures.
FINAL EXAM: The final exam was fairly easy. Time was not an issue for me, I finished it 40 minutes early. You get to choose some of the questions. As long as you have been paying attention in the course, doing the practicals, reading the textbook, and revised previous exams, you should be fine.
SUMMARY: The course is not very difficult, but could have probably been structured a bit more nicely, especially if the lecturer didn't rely so much on the textbooks. A downside is that you don't really come out of the course feeling like you know as much as you would have learnt in a 2 unit course. This is probably because of the variety of topics that are covered, a lot of which are very intuitive.
NOTE: I took this in 2018, not 2017 - but there's no option for 2018!
Semester 1 - 2017
Not compulsory, but you should attend anyway
Yes, but they provide online versions
Okay course with content covering machine learning basics but if you want to get into deep learning and possibly be part of the AI boom this course is not going to get you there. You will be spending most of your time doing stats in Matlab and making trivial predictions and models, which might let you down if you expect to be taught how to build AI that magically works.
The lecturer doesn’t really explain much in lecture so you need to do a lot of self learning to understand hard concepts. Most of what he talks about in lecture is super easy to miss unless you are on top of the content he’s talking about, which is a little ironic. I would suggest watching tutorial videos online to help you understand the content better.
Assignments are fairly doable but the requirements can be vague due to the lack of a marking criteria. It’s hard to know what to write about for short response questions and you often find yourself losing marks because you can’t guess exactly what needs to be included in your answer.
Semester 1 - 2017
This course was really interesting and fun. Although there is a fair bit of statistics at the start, the programming exercises in the practicals give you a really good insight into the different methods of machine learning. I found the assessment not to be too time consuming however still challenging. I did the summer course CSSE3080 just before this which covered some of the material in this course so the first few weeks were pretty easy. Marcus is approachable and very reasonable with assessment. The assignments can mostly be done in the practicals. I found the reading assignments really interesting and were reasonably easy marks. The exam was challenging however similar to previous years. Leading into this course the programming subjects i had done were CSSE1001, CSSE2010, CSSE3080. As long as you are comfortable with MATLAB and linear algebra though you wouldn't need that much programming experience. Awesome course, definitely recommend it!!
Semester 1 - 2016
BE Electrical Engineering
Yes if you want a good mark