Introduction
All of probabilistic machine learning rests on two rules — the sum rule and the product rule — and their famous consequence, Bayes' theorem. With them you can get any marginal or conditional probability from a joint distribution, and you can update beliefs when data arrive.