A long overdue post on some Coursera courses that I completed.
Probabilistic Graphical Models is probably the most difficult course that I did. I had ventured into this course with a fair background of linear algebra and probability, but I still found this course extremely demanding.
Here are some reasons why:
● the course material is compressed and the instructor (Prof. Koller) moves through the material rather quickly.
● This course has a few prerequisites that aren’t mentioned as, well, prerequisites: basic machine learning, felicity with Octave/Matlab, the ability to convert mathematical notation to working Matlab code.
● I found the programming assignments to be quite tough, or put differently, more difficult than any other Coursera course. If not for the generous help of other students in the forum, (a view echoed by many other students), it would be impossible to get through the assignments. In my case, I had to sometimes grind through the assignment to finish it in time, sometimes missing the intuition behind a method that was used in the assignment.
● The time estimate of about 15 hours per week (spent on the course) is quite low in my opinion, it would be closer to 20/25 hours per week.
However, this is still a must-do course for students of ML. The material, as well as the instructor are excellent, and the programming exercises are beautifully designed, in that they give you a real world problem to solve in almost all the assignments which leads to much better insights. Having completed the course, I keep going back to the assignments and course material to refresh my memory every once in a few weeks.
Course difficulty (on a scale of 5): 5/5
Time required per week: 20 hours.
Prerequisites: Probability, Octave, Basic ML
Fun quotient: 4/5
One cannot ask for a better introduction to Machine learning. This course has a gentle pace, no prerequisites, and gives a 10000 ft view of the many aspects of Machine learning in general.
Prof Ng is a very gentle instructor, and he tries to break down the complicated math and explain it in a very understanding manner.
The assignments are probably too easy as they just ask you to fill up a few lines in Octave files, but that does not take away the usefulness of the course. It shouldn’t be hard to complete this course in parallel with a couple of others
Course difficulty (on a scale of 5): 2.5/5
Time required per week: 6 hours.
Fun quotient: 4/5
Prof. Hinton is one of the leading lights of Neural Networks, an area of ML research that had been relegated to the sidelines in the 80s and 90s but is now in the limelight thanks to recent advances in the field.
The basic ML course does dip its toes in the neural networks pool, but this course (but naturally) goes much deeper. The material as well as the instructor are excellent, and the course lectures are punctuated by Prof. Hinton’s deadpan humour. This course requires a little mathematical background (such as a calculating derivatives for backpropagation) but the math is explained quite nicely.
My only suggestion in this course would be to have more programming assignments (there were only 4 in the first edition), however, the quizzes do require some programming to answer questions.
Course difficulty (on a scale of 5): 3.5/5
Time required per week: 10 hours.
Prerequisites: Bits of calculus, octave,
Fun quotient: 5/5