Coursera: Applied Machine Learning in Python Review

Applied Machine Learning in Python is the third course in the Applied Data Science with Python Specialization specialization offered by the University of Michigan on Coursera. The 4-week course covers machine learning in Python, focusing on the scikit-learn package. You need to know the basics of Python programming to complete the course; the first course in the specialization provides the necessary background for numpy and pandas. Grading is based on weekly quizzes and programming assignments. Unlike the second course in the specialization, quizzes and assignments are not locked behind paywalls so you can complete the course as a freeware student.

Course content follows the same formula as the previous courses in the specialization, with each week offering roughly 6 to 12 lecture videos that are 5 to 20 minutes in length accompanied by programming notebooks that help you follow along with the material. Programming assignments are also administered and submitted via in-browser notebooks. Video and instruction quality are good, but the pacing could be better. The first week provides an introduction to machine learning and the skikit learn package, using K-nearest neighbors classification as a motivating example. Week 2 expands on KNN and then covers linear, ridge, lasso and logistic regression, support vector machines, cross validation and decision trees. Week 3 offers a reprieve from the breakneck pace set in week 2 with an excursion into model evaluation, but week 4 picks up where week 2 left off, covering Naive Bayes, random forests, gradient boosted trees, neural nets, deep learning, dimensionality reduction and clustering. Cramming so much material into weeks 2 and 4 will likely make it difficult for students to gain a deep understanding of any one machine learning method in particular. It seems that Coursera wants all its new courses to run on recurring 4-week schedules, forcing content providers to work around this constraint by packing weeks too full or making certain lectures optional.

Applied Machine Learning in Python is fine introduction to machine learning for students who are already comfortable with numpy and pandas. It covers a lot of ground in 4 weeks and would probably be better if the content was spread out over 6 to 12 week course that took more time to explore one topic before moving onto the next. The programming assignments give you practice working with the methods and functions discussed in class, but open-ended instructions and finicky data input/output formats mean you may spend more time trying to figure out what you're supposed to do and how to shape your data to do it than actually running and analyzing the models themselves. Still, the course introduces many of other most commonly used machine learning methods used in practice with examples in Python, which is valuable no matter how it is packaged.

I give Applied Machine Learning in Python 3.5 out of 5 stars: Good.