Old ML course by Andrew Ng vs the new ML Specialization

Hi everyone.
I have been taking the old ML course by Andrew Ng on Coursera for a while and I had advanced to week 4, but I have decided to switch to the new course. It seems to me that the first course of the specialization covers almost the same topics as the first 3 weeks of the old course. My question is, should I advance to the second course of the specialization, or should I start over with the first course?

And how much Python knowledge is required for taking the first course? I mean, is there any Python taught during the course? if so, what topics are covered? I have some prior knowledge of coding with Python but I am not familiar with NumPy and Pandas. Should I learn them before starting the course?

Hey @Sama_san,
Welcome to the community. I haven’t done the original ML course myself, so, I am not really familiar with the extent of over-lapping between the original course and the new specialization.

But if you feel that they have over-lapping content, then I think that you can easily complete the first course in a matter of days. This will also help you to revise your foundations, and if I am not wrong, Course 1 is filled with optional labs, something which was missing from the original course. Additionally, completing course 1 will help you to get up to speed with Python, since the original course was not served with Python.

For this, check out 03 of this thread.

As to this, it really depends on you how you want to proceed. If you want, you can go through some of the tutorials of these packages before you start with MLS, and it is most certainly going to help you. You can also start with MLS and check out these tutorials as and when you need them, and this strategy is perfectly fine as well. You will find that the MLS labs provide links to the documentation whenever a new function is used from these packages.

I hope this helps.


1 Like

I’m also a mentor for the original ML course.

This course adds a few new ML methods (decision trees and reinforcement learning, for example), but does not include some material from the original (no SVM or PCA).

One benefit of this course is that the videos were recorded recently, and Andrew has had lots of practice at presenting and improving this material (since the original course was recorded some 10 years ago).

And of course the programming environments are entirely different.

So it doesn’t hurt to complete both courses.

Both courses have the same philosophy about your programming expertise - you’re expected to be moderately fluent in programming in the appropriate language.

Personally I think Octave/MATLAB from the original course was much easier for novices to use. Python is no picnic if this is your first programming rodeo.

1 Like