Rohan Kapur profile image

Rohan Kapur


  • 1 review
  • 0 completed
This course is probably one of the best data mining courses on Coursera, but it has a few flaws. Firstly, it's difficult. This is NOT a flaw, but something to note. It's a Stanford graduate level course that, per the instructors, "has not been watered down at all". This results in many people complaining in the discussion forums that the course has grown "legs", and can be quite stressful. Advice to ANYONE taking the course: make sure you have a lot of prior Linear Algebra knowledge. It's much more rigorous than the Linear Algebra used in something like Andrew Ng's Machine Learning course. However, there's also a great book that's published with the course. It's a wonderful reference when you're stuck on something. With that being said, the quizzes are not hard. A lot of the quizzes are normally very short, and do not require much insight into the material. A lot of them are just simple mathematical calculations that are somewhat in the smaller processes of the algorithms in the lectures. On top of this, there's no programming exercises, which are usually helpful for gaining insight on these topics. I think, however, there are a few set of lectures which are extremely difficult to understand. Week 2 is renown on the discussion forums to be the hardest portion of the course, most specifically because Jeffrey Ullman (who is an amazing Computer Scientist!) is a very monotone, rather boring and un-motivational lecturer (he directly reads off a transcript). He doesn't give many examples and his lectures are VERY long (30 minutes at one time), which leaves things rather abstruse. Sometimes, his train of thought can be near impossible to follow. This is where the MMDS book comes in handy. Luckily, Jure and Anand are great lecturers who make things concise, easy, and fun to understand. The content is GREAT. That's definitely an important thing to note. The topics they explore are rather interesting, and relevant. But the course moves very fast. There's a LOT of content to go through, usually two chapters of the MMDS book per week. The lectures can be quite long, and it makes things rather tiring. I think the biggest problem with this course is the lack of structure. I don't know why I'm learning what I'm learning. I don't know how it even fits back into Data Mining as a whole, or what is Data Mining? I don't know how single sets of lectures relate to the next set (it seems as if they do not at all). Weeks jump from one to another without any coherence. Why or when I would I use this in my startup or project? How do these topics relate together? There's no coherence or underlying motif to the course. The videos within the weeks lack coherence too. There are times when material from past lectures are randomly brought back up, and it seems as if random videos introducing new topics that aren't even explained into depth until WEEKS later are inserted at random points in the course. I've been able to retain very little of this course. I find that, each week, I almost certainly forget the intricacies and sometimes even the bigger pictures of the past week. This reflects on the sheer volume of content and the lack of effective structure of the course. MMDS has been mentally stimulating, but definitely needs an overhaul.