Coursera: Mathematics for Machine Learning -- Linear Algebra Review

Mathematics for Machine Learning: Linear Algebra is the first course in a 3-part Mathematics for Machine Learning Specialization offered by the Imperial College of London on the Coursera MOOC platform. The 5-week course aims to teach the foundations of linear algebra and develop your intuition about how it is used in machine learning. The course does not require any particular background besides high school math, although knowledge of Python will be helpful if you go for the certificate. Grading is based on periodic quizzes and a handful of programming assignments. A few of the quizzes are available to audit students but the programming assignments are reserved for subscribers.


Weekly content in Mathematics for Machine Learning: Linear Algebra consists of several 3-10 minute lecture videos interspersed with comprehension quizzes and occasional programming assignments. The first two weeks cover vectors, weeks 3 and 4 cover matrices and the final week discusses eigenproblems and Google's Page Rank algorithm. The video and lecture quality are very good: the notation written on screen is crisp and the instructors are passionate and easy to understand. That said, the concepts themselves may not always be easy to follow, especially if you don't have much prior exposure to linear algebra. The course tries to introduce a lot of different concepts and results from linear algebra relatively quickly so you won't gain a deep understanding of what's going on and it might not always be clear to you why the results shown are important. The quizzes are well done and do a good job helping you develop a geometric understanding of the material, but since some assignments are locked, non-paying students will miss out on what is likely some of the most instructive content in the course.


Mathematics for Machine Learning: Linear Algebra is a fine course to get a whirlwind introduction to linear algebra, but it lacks the depth necessary to gain more than a basic grasp of the concepts presented. Apart from a couple allusions to using linear algebra to rotate faces and a description of Page Rank, there aren't many connections drawn between the material and machine learning. The course spends a too much time delving into fiddly calculations that computers can do in one function call and not enough time giving concrete examples of how linear algebra is applied in machine learning in practice.


I give Mathematics for Machine Learning: Linear Algebra 3 out of 5 stars: Okay.