edX: Applied Machine Learning Review

Applied Machine Learning is part of a 9 course Professional Program Certificate in Data Science offered by Microsoft on edX. It is one of 3 electives available for the 9th course of the program. You can take it without doing the lead-up courses in the program, but you should at least know the basics of R and Python beforehand. That said, the lead-up courses will make make things easier, since you'll already have a Microsoft Azure account set up which is used for the homework and you'll have the required background knowledge.


Each module of Applied Machine Learning covers a different type of data and machine learning techniques that apply to that data type. The 4 modules cover time series data, spacial data, text data and image data. Organizing the course by data type was a great decision, as it puts the focus on deriving practical insight using data-appropriate models. There are a couple hours of lecture in each module, giving adequate time to each topic. The lectures alternate between introducing new concepts and demonstrating how to apply machine learning techniques in code or Azure. Lecture video and instruction quality is generally good, although it sometimes shows that the instructors are data scientists first and teachers second.

The course grade depends entirely on 4 labs and a final exam. The labs each involve following along with code notebooks and lab instructions to reinforce lecture topics and implement a data experiment on Microsoft Azure. You don't really have to do any programming yourself: you just use the provided code and experiment skeletons, running code snippets on new inputs and filling out the Azure experiments as instructed. The final exam is worth half of your grade and only has 9 questions, each of which only give you one attempt. The questions aren't hard, but a few of them are subjective and some are multiple checkbox questions, so it would be quite easy to get some wrong and lose a significant amount of your overall grade.

Applied Machine Learning is a great course for learning about machine learning methods in the context of specific data types. The section on time series data was especially nice to see, since it is a topic that is rarely covered in other machine learning MOOCs. Parts of the labs feel overly scripted and the final exam is an ill-conceived afterthought, but that's not adequate reason to avoid this course.

I give Applied Machine Learning 4 out of 5 stars: Very Good.