Maths of Learning by Building

OBJECTIVE: understand the maths behind machine learning by building it in code — no hand derivations needed for applied learners. WHAT WE COVER: • A model and its loss; the derivative as a slope; gradient descent (and Momentum/Adam glimpse) • Sigmoid, log loss (cross-entropy), softmax and the confusion matrix • PCA: covariance and main directions • Statistics for decisions: bootstrap, significance tests and p-values • Capstone: a loan-default EDA and a model built from scratch AFTER THIS MODULE YOU CAN: • Explain how a model learns from its errors • Read probabilities, losses and test results correctly • Build a small model end to end from scratch