Four courses that take you through machine learning, deep learning, natural language processing and computer vision. Each lesson starts with a picture, works a small example by hand, then checks it in Python.
Free, no account needed. Machine learning is open now, with 48 lessons and a quiz in every one. The other three courses have their syllabus ready, 107 lessons planned in total.
Start with machine learning: it's open now, and deep learning, language and vision all build on it. The rest are being written, and you can read their full syllabus today.
Foundations, open now
Start from zero. Every lesson begins with a small real example you can work out by hand, then shows the same idea in scikit-learn, with a narrated video for the trickiest step.
How neural networks learn, why they work so well on images, sound and sequences, and how to train them without losing your mind.
6 parts, 20 lessons plannedRead the syllabus
Everything between raw text and a model that understands it, from cleaning and tokenising to embeddings, attention and fine-tuning transformers.
5 parts, 22 lessons plannedRead the syllabus
How images become numbers, how convolutional networks learn to recognise what is in them, and how to find, outline and describe objects in photos, and even generate new images.
7 parts, 17 lessons plannedRead the syllabus
This is the idea behind linear regression, the first model in the machine learning course. Five students, the hours they studied and the marks they got. Move the line until the misses are as small as you can make them.
In the lesson you'll find the best line by hand, then let gradient descent find it for you. The same idea, a computer nudging numbers to shrink an error, is how deep networks and language models learn too.
Open the linear regression lessonFive Students, One Line
Drag the blue handles. Can you get the squared error down to 240?
Red lines are the misses. Squaring and adding them gives the error.
Every lesson in every course follows the same five steps, so you always know what comes next.
A real situation first, like guessing a student's marks from the hours they studied.
A short narrated video shows what the maths is doing, with subtitles word by word.
A few rows of data, small enough for a notebook, so every number has a reason.
The same numbers again with real libraries, with the output shown.
A short quiz and a few edits to try, plus the mistakes beginners make most.