Crate anofox_ml_gaussian_process 

Gaussian Process Regression.

Mirrors sklearn.gaussian_process.GaussianProcessRegressor with a fixed kernel composed from the kernel zoo below. Hyperparameter learning is not yet implemented — provide explicit kernel parameters.

Kernels supported

• Rbf — squared-exponential σ² exp(-||x-x'||² / (2ℓ²))
• Matern — Matern kernel for ν ∈ {0.5, 1.5, 2.5} (closed-form parameterisations)
• RationalQuadratic — (1 + ||x-x'||² / (2αℓ²))^(-α)
• White — σ² if x == x' else 0 (diagonal noise)
• Constant — σ² everywhere
• Sum / Product — composite kernels

Re-exports

pub use classifier::FittedGaussianProcessClassifier;

pub use classifier::FittedMulticlassGaussianProcessClassifier;

pub use classifier::GaussianProcessClassifier;

pub use classifier::MulticlassGaussianProcessClassifier;

Modules

classifier

Gaussian Process binary classifier with Laplace approximation.

Structs

FittedGaussianProcessRegressor

GaussianProcessRegressor

KernelOptimResult

Result of multi-parameter hyperparameter optimisation.

Enums

GpKernel

Composable kernel.

Functions

log_marginal_likelihood

Compute the log-marginal-likelihood log p(y | X, kernel, alpha) for a given kernel and noise level on (X, y).

optimize_kernel_lbfgs

Multivariate quasi-Newton (BFGS) optimisation of arbitrary kernel hyperparameters on the log-scale, maximising log-marginal likelihood.

optimize_rbf_length_scale

Find the length_scale (RBF kernel) that maximises log-marginal-likelihood via golden-section search over log(length_scale). Other kernel parameters are kept fixed at the provided values.