Module gp 

Gaussian Process surrogate model for Bayesian optimization.

Provides a self-contained GP regression implementation with:

• Multiple kernel functions: RBF, Matern (1/2, 3/2, 5/2), Rational Quadratic
• Composite kernels: Sum and Product
• Efficient Cholesky-based prediction for mean and variance
• Hyperparameter optimization via type-II maximum likelihood (marginal likelihood)

The GP is specifically designed as a surrogate model for Bayesian optimization, prioritising numerically robust prediction of both mean and uncertainty.

Structs

GpSurrogate

Gaussian Process surrogate model for Bayesian optimization.

GpSurrogateConfig

Configuration for the GP surrogate.

MaternKernel

Matern kernel with selectable smoothness parameter nu.

ProductKernel

Product of two kernels: k(x,x’) = k1(x,x’) * k2(x,x’)

RationalQuadraticKernel

Rational Quadratic kernel.

RbfKernel

Squared Exponential / RBF kernel.

SumKernel

Sum of two kernels: k(x,x’) = k1(x,x’) + k2(x,x’)

Enums

MaternVariant

Variant of the Matern kernel.

Traits

SurrogateKernel

Trait for covariance (kernel) functions used by the GP surrogate.