Review of University of Washington’s Machine Learning Specialization on Coursera

A few weeks ago, I received a somewhat disappointing news from Coursera — courses 5 and 6 of the Machine Learning Specialization are not going to be launched. The specialization is now complete at 4 courses. Having gone through the first three, I thought it is time to review the Specialization and I hope that this review will be of help to others who are deciding among the growing number of MOOC offerings.*

The Specialization

  1. Machine Learning Foundations: A Case Study Approach
  2. Machine Learning: Regression
  3. Machine Learning: Classification
  4. Machine Learning: Clustering & Retrieval
  5. Machine Learning: Recommender Systems (CANCELLED)
  6. Capstone: Deep Learning (CANCELLED)

You can see more details about this specialization here.

My Review

The Verdict

This Specialization consists of a series of beginner-friendly yet practical courses, covering basic machine learning concepts as well as the most useful techniques. Python is used throughout the courses. A lot of MOOCs use Octave or Matlab but I think Python is preferable given it is the industry standard. Overall, I would give it a recommendation of 4.5 out of 5.


If you know how for loops and if-else statements work in Python, you are good to go. If not, learn the language from CodeAcademy or The Python Tutorial, then come back. Also, you will need the basics of Linear Algebra, but it is covered in the course with tutorials.

The Good

Despite feeling somewhat disappointed by the cancellation of the last two courses, I have to say that this Specialization developed by University of Washington and Coursera was, in my opinion, one of the few data science courses that peel back to the basics and really help you understand the behind-the-scene magic (read: math) in Data Science. The instructors are both Amazon professors of Machine Learning and very knowledgeable about what they are teaching. The math part is important to me because, although there are many open source libraries that offer sophisticated algorithms packed in a few lines of code (which is excellent), it is hard for beginners to understand what these algorithms are and when to apply them.

The first course of the Specialization is an overview, presenting each of the following courses in a case study using real-world examples. For example, regression is used in predicting housing prices; and clustering is used for categorizing documents. The following courses are presented in a way that guides you to write machine learning algorithms from scratch. This is the best part of the course. It might sound tedious and unnecessary at first, but the process really helped me understand, for instance, what gradient descent really does, as opposed to choosing to use gradient simply because it looks familiar. It also does a good job of contrasting different algorithms and informing learners about gotchas and pitfalls.

In terms of job prospects, any MOOC alone is not going to get you very far. So one probably would not feel job-ready after this course, but it is definitely a huge step in the right direction.

The Bad

The Specialization is developed by Emily Fox and Carlos Guestri who is the CEO of Dato, so a lot of the course content is centered around GraphLab Create, a machine learning library by Dato. So Sometimes it is a bit tricky to understand the programming quiz if you are not using GraphLab. But it is completely possible to complete the courses with open source libraries like Pandas, Numpy and Scikit-learn, as I have done in the first three courses of the Specialization.

In terms of support when you get stuck, Coursera is not the best in being responsive if you have a question about the course content. This is because most mentors are volunteers and have limited time to help learners, and many MOOCs have the same problem. But I consider it a drawback of the platform but not of this specialization.


Although some improvements can be made, I do feel that the Specialization is one of its own kind in that it offers the theoretical background in machine learning that no other courses seem to provide in detail. If I were to do it again I would choose to take the courses in this specialization in a heartbeat.

*Disclaimer: I have volunteered to be a mentor for Machine Learning Foundations: A Case Study Approach.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.