Part 3 of 8

The Maths I: Gaussian Process Surrogates

Everything BO does flows through two numbers per candidate point — μ_n(x) and σ_n(x) — so this section builds the machine that produces them. Notation follows Frazier’s tutorial (and Garnett’s book): n observations X = (x₁, …, x_n) with values y = (y₁, …, y_n), maximization convention throughout.

1. A prior over functions
2. Conditioning: the posterior in closed form
3. Choosing hyperparameters: the marginal likelihood
4. Quiz: 5 questions