the course

Six modules, eighteen units. Each unit is three short lessons, a theory test and a practical you solve in Python right here. Each module ends with a guided project on real data. Everything is open; the order is the recommended path.

01

The maths of learning

Vectors, gradients and probability: the three tools every model is built from.

…, about 85 min of reading
  1. unit 1.1Vectors and matrices 3 lessons · test · practical
  2. unit 1.2Derivatives and gradients 3 lessons · test · practical
  3. unit 1.3Probability and statistics 3 lessons · test · practical
  4. projectFit a line by gradient descentopens when the units are solved
02

Supervised learning

Learning a mapping from labelled examples: regression, classification, neighbours, trees and ensembles.

…, about 85 min of reading
  1. unit 2.1Linear regression 3 lessons · test · practical
  2. unit 2.2Classification 3 lessons · test · practical
  3. unit 2.3Neighbours, trees and ensembles 3 lessons · test · practical
  4. projectA flower classifier from scratchopens when the units are solved
03

Models that generalise

Overfitting, regularisation, honest evaluation, and learning from data without labels.

…, about 90 min of reading
  1. unit 3.1Overfitting and regularisation 3 lessons · test · practical
  2. unit 3.2Measuring models honestly 3 lessons · test · practical
  3. unit 3.3Learning without labels 3 lessons · test · practical
  4. projectAn honest diagnosis modelopens when the units are solved
04

Neural networks

From a single neuron to backpropagation, and the tricks that make deep networks train.

…, about 90 min of reading
  1. unit 4.1Neurons and networks 3 lessons · test · practical
  2. unit 4.2Backpropagation 3 lessons · test · practical
  3. unit 4.3Training deep networks 3 lessons · test · practical
  4. projectA digit recogniser from scratchopens when the units are solved
05

Deep architectures

Convolutions for images, recurrence for sequences, and attention, the idea behind transformers.

…, about 90 min of reading
  1. unit 5.1Convolutional networks 3 lessons · test · practical
  2. unit 5.2Sequence models 3 lessons · test · practical
  3. unit 5.3Attention and transformers 3 lessons · test · practical
  4. projectTrain a convolutional network from scratchopens when the units are solved
06

Language models and generative AI

How large language models and image generators work: tokens, next-token prediction, scaling, alignment and diffusion.

…, about 90 min of reading
  1. unit 6.1Language models 3 lessons · test · practical
  2. unit 6.2Large language models 3 lessons · test · practical
  3. unit 6.3Generating images 3 lessons · test · practical
  4. projectA tiny language modelopens when the units are solved