Machine Learning

Course Website

Average Workload

21.7 hrs/wk

Average Difficulty

4.1 /5

Average Overall

3.2 /5
CS-7641
Machine Learning
Taken Spring 2024
Reviewed on 5/5/2024
Workload: 8 hr/wk
Difficulty: Hard
Overall: Liked

This class is certainly difficult, and this is coming from someone who has done stuff directly related to the material at their job for the past 5 years. The material/lectures/textbook is maybe a bit outdated (I didn't look at the textbook after about 2 modules), but I'd say it covers all of the core concepts of machine learning. The assignments are the real crux of the class, and the variance in student experience probably stems mostly from opinions about how fair the expectations/instructions/grading of the assignments are. My two cents: There is definitely a difference between the expectations for the assignments in this class vs others. They expect you to really dig into the techniques and discuss results in a way that directly links to the material in lectures/office hours. The TA's have made efforts to make what they want to see more explicit (they make an ed post for each assignment going into more detail), but there is still some ambiguity about when you have 'checked a box'. Make sure you include and meaningfully discuss everything they mention. If you spend like an hour just playing around with some hyperparameters and then briefly mentioned the results in the report (like I did a lot of the time) you are probably gonna get dinged. Use LaTeX (GaTech has free overleaf accounts) or a two column format, since it is sometimes near impossible to pack all the material into a standard google doc/word document. Use a ton of figures, every time there is an opportunity for a plot, include it.

Overall I think the grading is fair and matched my level of effort (I really tried to wing it, ~8 hrs/week), however it is somewhat ambiguous/opaque (you will pry that rubric from their cold dead hands). I do think this ambiguity encourages learning by digging into the algorithms and matching the results to the material, and the curve makes things feel more fair. The TA's are also very active in office hours (don't be like me, go to them) and the discussion boards.

The curve is harder than other classes, don't give up just because you got 50% on the first assignment. I was in a similar situation and ended up with an A after recovering a bit on other assignments and the final.