Professor Diggavi is definitely an expert in the field of ML, as he would often share with us stories behind the development of ML algorithms. Even though his lectures are dry and math-heavy, he is able to explain most of the abstract concepts clearly. Contrary to popular opinion, I actually appreciated the time and effort he put into going through the mathematical derivations behind the theorems and algorithms. Exams were on the tough end but fair - just make sure to include all the key concepts + proofs in your cheatsheet, and fully understand the practice exam. However, homework specs can be confusing at times, with a couple of mistakes here and there. Fortunately, the TAs (esp Sadik) were really responsive on Campuswire to clarify any doubts we had. Overall, I do think this is a well-structured course, especially if you are keen to learn more about the math behind ML, which complements well with the more applied ML courses like ECE C147 and CS 162/163.

I do think that taking 115A and 170S concurrently with this class helped me a lot. As mentioned in the previous comments, classes like PIC 16A, Math 115A, Math 170S and CS M148 are helpful pre-requisites.

I agree with the last comment. Professor Diggavi has to be one of the greats in machine learning and definitely cares about his students. The truth is, this class was all just very rigorous mathematical concepts and proofs (even on exams) and Professor Diggavi does not shy away from that. Unfortunately, I do not think my coursework adequately prepared me to take this class, but looking back, CS M148 now looks like a walk in the park. I don't think taking this course deters me at all from taking future data science coursework (especially at the graduate level), but it was definitely a wake-up call that the theory of ML is quite complex and is no easy task. Weirdly, now I want to get better at math to better understand this stuff. Relevant Prerequisite knowledge that I had going into course: 33A, PIC 10C, MATH 170E, MATH 170S, CS M148, STATS 115. I would probably recommend some more rigorous proof based math (e.g., MATH 115) to get more out of the class...this stuff is all linear algebra.

Professor Diggavi is clearly very passionate about the subject matter and is dedicated to ensuring students understand the material. His enthusiasm for Machine Learning is evident, and he genuinely cares about student learning. However, despite his efforts, this class falls short in terms of engagement and enjoyment.

The course is overwhelmingly math-heavy, which detracts from the overall appeal of Machine Learning. While a solid understanding of the mathematical foundations is undoubtedly important, the current approach makes the subject matter feel unnecessarily daunting and unapproachable. There are numerous resources and instructional methods available that present these complex concepts in a more engaging and digestible manner. Integrating some of these methods could make the class more enjoyable and less intimidating.

Another challenge in this course is the assumption that students have a firm grasp of advanced mathematical concepts, often beyond what was covered in the prerequisite courses. While Professor Diggavi does attempt to review these concepts, it is unrealistic to expect us to recall all the intricate details without sufficient review. This often leads to a significant gap in understanding, making the material difficult to grasp.

All in all, I did not enjoy this class and felt like I was just trying to get through it. I'm not sure if it would be any better with another instructor either but I believe that Professor Diggavi is perhaps one of few people who might be able to make this class amazing if he wishes to make it a bit more accessible.