P2 · Mathematics for ML

linear algebra, calculus, optimization

The mathematics models are made of — numbers as computers store them, functions, vectors and matrices, probability, gradients — introduced visually and only as deep as the models need.

The mathematics here is a working vocabulary, not a course in proofs. Models are built from a surprisingly short list of ideas — numbers as a machine stores them, functions you can draw, vectors and the matrices that move them, probabilities, and the gradients that let a model improve — and each one is shown as a picture before it is ever a formula. When an equation does appear, it arrives after you have already watched the idea move on screen.

Two tracks share the pillar. Mathematics Foundations climbs the stack below one interactive lab at a time, stopping exactly where models stop needing more. AI/ML Foundations then spends that vocabulary on a real story: how counting which words appear together turns into vectors that measure meaning — the runway to embeddings and, later, the Transformer itself.

Each layer leans on the one beneath it; the labs climb bottom-up.