Back to index
UdemyAround 40 hours of video, fully self-paced·Usually $15 to $20 on one of Udemy's frequent sales, list price much higher

Machine Learning A-Z: AI, Python & R

4.1

A broad, friendly tour of classic machine learning that is great for getting moving fast and seeing the whole map, but light on the maths and increasingly dated in places. Brilliant as a first hands-on course, not enough on its own to make you a practitioner.

What We Liked

  • Enormous breadth, you touch nearly every classic algorithm in one place
  • Kirill and Hadelin are warm, energetic teachers who keep a long course watchable
  • The code templates mean you have working examples for almost everything from day one
  • On sale it is absurdly cheap for 40-plus hours of structured material

What Could Be Better

  • The maths is mostly skipped, so you learn which button to press more than why
  • Some sections lean on older libraries and conventions that have moved on since
  • Doing both Python and R doubles the runtime for skills most people will not both use
  • Copy-the-template format means you can finish without writing much yourself

Detailed review

This is the course that introduced a huge number of people to machine learning, and I understand why it sells the way it does. The breadth is the headline. In one package you move through data preprocessing, regression, classification, clustering, association rule learning, reinforcement learning, natural language processing and a chunk of deep learning, and seeing that whole landscape laid out in a single coherent path is genuinely useful when you are starting out and do not yet know what you do not know. Kirill Eremenko and Hadelin de Ponteves are the reason 40 hours does not feel like a slog.

They are upbeat, they explain clearly, and they have a knack for making an intimidating topic feel approachable, which for a beginner matters more than people admit. The ready-made code templates are a real practical strength too, because you finish each section with a working script you can adapt rather than a blank page. Where I have to be honest is the depth. This course is built around intuition and application, and it deliberately skips most of the maths, so you come away knowing how to apply an algorithm in scikit-learn without really understanding what it is doing or when it will fail.

For getting started that is a reasonable trade, but you will hit a ceiling, and you should plan for it rather than be surprised by it. A few sections also show their age, leaning on older library patterns that have since changed, so you occasionally have to translate to current versions yourself. And the both-Python-and-R structure, while nice in theory, means a lot of duplicated runtime for two ecosystems most learners will not actually use side by side, so I would pick one language and skim the other. My take after going through it is simple.

As a first, cheap, motivating hands-on tour it is excellent and worth every cent of the sale price. Just do not mistake finishing it for mastery. Use it to find the topics that grab you, then go deep on those somewhere more rigorous, and write code from scratch alongside the templates so the skill is yours and not just the instructor's.

[ final ]

The verdict.

Buy it on sale, treat it as your guided tour of the ML landscape, and use it to find out which areas you want to go deep on. Then back it up with something more rigorous like Andrew Ng's specialization, because this course shows you the what far better than the why.