This class has destroyed any possible interest I have in machine learning. I initially came into the class eager to learn. Then, shit hit the fan.

Lectures: He's a pretty decent lecturer, but when he started getting into very detailed math, I was completely lost. A few weeks in, he began giving a quick demo of cool modern day ML applications each week, which was fun to see the bigger picture of what we were learning. Still, so much of the content flew over my head.

Homeworks: Overall, there were many typos and the lectures did not help at all. The homework, which contained coding portions and free-response, was clearly copied from other sources. They took a long time and did not seem to help prepare for the exams, especially because the exams do not test any coding whatsoever. Also towards the end of the quarter, the TAs and instructor selectively answered questions on Campuswire, leaving 20+ questions about the HW/class unanswered.

Exams: The midterm had so much math that I had absolutely no idea how to do. Sure, the class said we needed linear algebra as a prerequisite, but the entire class seemed way too math heavy for a computer science class. The final was a disaster and I felt completely unprepared.

The only plus side is after failing the exams, the curve was thicc so bless Grover for saving my grade. I have no idea how he curved the class but I am surprised I passed the class, let alone got the grade I got. My recommendation is to not take this class with Grover, or better yet, not take this class at all. I'm sure there are better electives taught by better professors.

Prof Grover is a slightly below average professor. His lectures are generally uninteresting. He practically reads directly off of the slides - which is good, because you can't hear him in a large lecture hall, so if he said anything else it wouldn't be helpful anyways. At least the slides are usually posted on Bruinlearn so you can follow along, but lectures don't really add a whole lot.

The homework is absolutely terrible. The written problems are fairly math-heavy, but nothing too bad if you've taken calculus and linear algebra. The problem is with the coding sections. The instructions on the handout and in the Jupyter notebooks often differ substantially. It seems clear that at least the coding portion of the homework was stolen directly from some other instution - probably CMU or Stanford if I had to hazard a guess - with significant portions cut out (but the instructions not updated). That's extremely discouraging, and feels like it borders on academic dishonesty. If I submitted the answer keys from the homework from wherever it was taken from, I would be referred for plagiarism. Ultimately, this results in homework that takes several hours, not because of the difficulty, but because it takes hours to understand what the instructions are actually trying to say.

The tests were probably a little more difficult than the average CS class, but not deathly hard if you actually put in some time to study for them. The math in them wasn't difficult, just basic calculus (Lagrange optimization) and linear algebra (matrices, orthonormal, matrix composition). There wasn't really any probability knowledge required for the course either.

The course content was mostly focused on motivating the machine learning algorithms discussed, and deriving them semi-rigorously. That's important knowledge to have if you want to go deeper in machine learning, especially at the graduate level. If you just want a cursory exposure to machine learning, then despite this course's name, a better course would be CS M148, which doesn't go into the same mathematical depth. If you like M148 (or if you already have an interest in and knowledge of machine learning), then take CS M146.

Perhaps the only saving grace for Grover, however, was his very generous grading policies. He allowed 6 late days across 4 homeworks, and the final grades were generously curved based on the final performance. However, he didn't release any grading scheme during the quarter, which was quite unhelpful.

I really enjoyed this class. Please give me token Hotseat.

The concepts were all interesting and quite insightful if you're into ML. Coding was involved but it was mostly writing a few lines here and there for each assignment.

Simple/Good introduction to ML; has become less proof based than the past (most of the rigorous mathematics is taken as fact) but covers the important concepts well.

Only 4 homeworks which progressively get easier - if you start early they won't be a problem.

The midterm and final are fairly straightforward; prof. gives extremely useful practice tests which simulate the actual exams almost.

Would recommend; the notes are accessible and lectures are also recorded. Very student friendly.