Coursera: Mathematics for Machine Learning: Multivariate Calculus Review

Mathematics for Machine Learning: Multivariate Calculus is the second course in the new
Mathematics for Machine Learning Specialization offered by the Imperial College of London on Coursera. The six week course covers the various aspects of calculus necessary understand important machine learning concepts like optimization with gradient descent and backpropagation. The course does not require a background in calculus, but it assumes you've taken the first course in the specialization and you'll need some basic programming skills to complete all of the assignments. Most of the quizzes and all the programming assignments are reserved for subscribers, so while you can audit the course, you'll get more out of it with a subscription.

Similar to the first course in the specialization on linear algebra, this course attempts to build from the basics of calculus to applications relevant to data science in a single, relatively short course. As a result, weekly content jumps quickly from one topic to another: week 1 covers the basics of derivatives, weeks 2 and 3 jump right multivariate calculus culminating in a programming assignment on backpropagation and the final 3 weeks cover a wide range of additional topics including Taylor series, gradient descent and linear regression. The lecture videos are high quality and the lecturers are enthusiastic and easy to understand. Despite this, the pace at which new concepts are introduced is probably going to leave students who are truly new to calculus left scratching their heads and either giving up or struggling through re-watching videos several times. I audited the course so I cannot review the programming assignments themselves, but the course did include an interesting interactive visualization game called "The Sandpit" that helped illustrate concepts related to gradients and hill climbing.

Mathematics for Machine Learning: Multivariate Calculus is a well-made whirlwind introduction to the key parts of calculus that have important applications in machine learning. Unlike the first course in the specialization, this course does a good job connecting the concepts presented to common uses in machine learning so even if you don't understand all the math, you will gain an appreciation for how calculus fits into in machine learning. Unfortunately, several of the most important topics like backpropagation are only touched on briefly outside the programming assignments so you will have to pay for the certificate track to get the most out of this course.

I give Mathematics for Machine Learning: Multivariate Calculus 3.75 out of 5 stars: Good.