Tackling the Choice Paralysis for Beginners — Machine Learning by Coursera VS Machine Learning…

A brief comparison you might say.

As a beginner, there is a common problem almost everyone faces — choice paralysis. Specially, in the field of Computer Science, there are just too many resources. This is of course a highly positive thing, but it is also true that too many choices confuse people. And for a guy who has just started to explore things, this confusion is a significant barrier.

This post is particularly for people who are trying to get their first grip on Machine Learning. I know many of you are confused where to start. In this post, I will make a solid differentiation between the two most popular ML resources: Machine Learning by Professor Andrew Ng on Coursera and Google Machine Learning Crash Course.

I took the ML on Coursera more than a half year ago. After Google released their ML resource, I went to check them out and found the ML crash course. It looked quite interesting. So I went through the Crash Course and tried to get an idea what was going on right there.

Let me point out couple of the features I noticed in both the courses:

• The theoretical contents of the courses are not totally same. As far as I have looked into them, Coursera’s ML looks heavier than Google’s crash course.

• 
  

  The ML crash course is well organized with videos, reading materials and programming exercises. Conveniently, you won’t need to install Python libraries or TensorFlow on your own machine to try the exercises as they are hosted online in Google Colab. This is relieving for beginners. On the other hand, Professor Ng’s course needs MATLAB/Octave to be installed on your own machine. But of course this course is pretty well organized too.

• I should point out that as a programming language for Machine Learning, MATLAB is not anywhere near Python in popularity and community support. Python is kind of everywhere in this field of technology. In fact, in Deep Learning Specialization by Professor Ng, programming part is done in Python. And also TensorFlow is used as a framework. But I would also point out that it won’t be an inconvenience for you to do the coding in MATLAB. It is really easy to pick up the syntax.

• 
  

  Professor Ng’s course is more like a formal course. You have quizzes on each week’s topics along with programming assignments. You have to finish those exercises within deadlines and submit them online. Then there is this grading system which can help you to keep motivated and focused. This is not the case for ML crash course.In fact, I personally felt an urge to finish the course of Coursera because of the deadlines and the grading system. This helps to keep yourself in check.

• 
  

  Professor Ng teaches all the related topics literally from scratch. He teaches the ML concepts in good detail. For example, while teaching Neural Networks, he tries to describe the intuition behind all those complex things going on. He also talks about how you should do the coding and provides assignments based on the lectures. It won’t even matter if you don’t know how to code in MATLAB because you will be taught on the fly. The assignments are designed in such a way that you won’t have to write much code. Only core few lines of code are enough to finish the assignments. And they are easy. Contrarily, ML crash course expects some kind of familiarity with Python. It uses a higher level API of TensorFlow called tf.estimator. So you will do things using a higher level API and much of them will probably happen under the hood as the API will take care of lots of things for you.

• The above point leads us to another important point: if you are a beginner, it is more likely that you will feel quite uncomfortable with TensorFlow along with all the extensive usages of Python libraries. In this case, Google’s crash course will not suit you.

• 
  

  Mathematical knowledge (like Linear Algebra or Calculus) is not quite important in any of these courses. Since Google uses TensorFlow to do the exercises, mathematics will mostly go under the hood. In fact, they have made the Calculus of gradient descent optional to check out in one of their reading materials as far as I have noticed.Similarly, Professor Ng assumes that the students do not know any Calculus or Linear Algebra. He avoids complex mathematical proofs and derivations so that the materials are accessible to everyone completely.

• Finally, Professor Ng is, without any doubt, a great teacher. He talks about lots of other aspects of Machine Learning apart from the core topics in his course. These are worth listening to.

Bottom-line, if someone asks me which one he should take, I will definitely suggest the course by Professor Ng. This course has initiated the journey for many ML engineers. It will surely help you.

But it is also true that if you don’t take the Coursera’s course and decide to check out the other one, remember that there is definitely no harm in it. The contents are rich. You will indeed be benefited.

Best of luck for your journey towards the exciting world of Machine Learning!