Machine Learning — A Guide

Machine learning from first principles, told plainly and a little like a poem. Seven chapters, built to make the ideas click.

📘 Machine Learning, from the ground up

Seven chapters that build the whole field from a single idea — a function with dials you turn until it stops being wrong. No prior math assumed, plenty of pictures, and a few things you can drag and play with. Written plainly, and now and then like a poem.

The chapters. Read in order, or wander. Each stands on its own.

Chapter 1 — Neural Networks

From one neuron to a mind that learns: functions with dials, the slope of being wrong, and the quiet arithmetic of getting better.

10 min

Chapter 2 — Training & Optimization

The long walk downhill: batches and epochs, the temperamental learning rate, momentum and Adam, and the small tricks that keep a deep network from falling over on its way to the valley.

8 min

Chapter 3 — Evaluation, Overfitting & Regularization

The discipline of not fooling yourself: why a model that aces its training is often useless, how to measure honestly, and the gentle pressures that keep a model from memorising the world instead of understanding it.

7 min

Chapter 4 — Convolutional Networks

How a machine learns to see: the sliding filter, the feature hierarchy from edges to faces, pooling, and why a few small kernels beat a wall of dense connections for images.

6 min

Chapter 5 — Sequences & Memory

Giving a network a sense of before and after: recurrent networks, the vanishing-memory problem, and the gates of the LSTM that let a model remember what mattered and forget what didn’t.

6 min

Chapter 6 — Transformers & Attention

Learning to look at what matters, all at once: the attention mechanism, queries-keys-values, why parallel beat recurrent, and the architecture behind every modern language model.

6 min

Chapter 7 — Generative Models

From recognising the world to making new ones: autoencoders, the forger-and-detective duel of GANs, diffusion’s slow sculpting of noise, and how a next-word predictor became a writer.

7 min