An interactive handbook
Machine learning,
step by step.
Twenty-two chapters covering the foundations of machine learning, built around visualisations you can scrub through, code you can run in the browser, and problems that check your work. Designed for readers approaching the subject for the first time.
I — Foundations
- 01
What is machine learning?soon
Framing, vocabulary, and the supervised / unsupervised divide.
- 02
The ML workflowsoon
Data, train and test, generalisation, and the spectre of overfitting.
- 03
A mathematical toolkitsoon
Linear algebra, calculus, and probability — only what you need, exactly when you need it.
- 04
Your first model: k-nearest neighbourssoon
Building intuition for prediction without writing a single equation.
II — Linear models
- 05
Linear regressionsoon
The mother of all models. Closed form, gradient descent, and what they reveal.
- 06
Gradient descent
The optimisation engine behind nearly every modern model.
- 07
Logistic regression
From regression to classification with one elegant change of perspective.
- 08
Multi-class classification
Softmax, one-vs-rest, and choosing between them.
- 09
Regularisation
The geometric story of L1 and L2 and why they tame overfitting.
III — Evaluating models
IV — Trees and ensembles
V — More classical models
- 15
Support vector machinessoon
Maximum margins and the kernel trick, geometrically.
- 16
Naive Bayessoon
Probability done plainly. A surprisingly strong baseline.
- 17
Feature engineeringsoon
The unglamorous work that decides whether anything else matters.
VI — Unsupervised learning
- 18
k-means clustering
Finding structure without labels. Step through every iteration.
- 19
Hierarchical and density-based clusteringsoon
When clusters are not blobs.
- 20
PCA and dimensionality reductionsoon
Seeing high-dimensional data as it really is.
VII — Neural networks
- 21
The perceptron and multilayer perceptronssoon
From a single neuron to a network.
- 22
Backpropagation and trainingsoon
How a network learns, derived from first principles.