Struct GpSurrogate 

pub struct GpSurrogate { /* private fields */ }

Gaussian Process surrogate model for Bayesian optimization.

Maintains training data and the fitted model (Cholesky factor + alpha vector) for efficient prediction.

Implementations

impl GpSurrogate

pub fn new(kernel: Box<dyn SurrogateKernel>, config: GpSurrogateConfig) -> Self

Create a new GP surrogate with the given kernel.

pub fn default_rbf() -> Self

Create a GP surrogate with an RBF kernel and default configuration.

pub fn fit(&mut self, x: &Array2<f64>, y: &Array1<f64>) -> OptimizeResult<()>

Fit the GP to training data.

Optionally optimises kernel hyperparameters via marginal likelihood maximisation.

pub fn update( &mut self, x_new: &Array2<f64>, y_new: &Array1<f64>, ) -> OptimizeResult<()>

Add new observations incrementally and refit the model.

pub fn predict_mean(&self, x_test: &Array2<f64>) -> OptimizeResult<Array1<f64>>

Predict the GP mean at test points.

pub fn predict_variance( &self, x_test: &Array2<f64>, ) -> OptimizeResult<Array1<f64>>

Predict the GP variance at test points.

pub fn predict( &self, x_test: &Array2<f64>, ) -> OptimizeResult<(Array1<f64>, Array1<f64>)>

Predict both mean and variance at test points.

pub fn predict_single( &self, x: &ArrayView1<'_, f64>, ) -> OptimizeResult<(f64, f64)>

Predict mean and standard deviation at a single point.

pub fn log_marginal_likelihood(&self) -> OptimizeResult<f64>

Compute the log marginal likelihood of the current model.

pub fn kernel(&self) -> &dyn SurrogateKernel

Return a reference to the current kernel.

pub fn kernel_mut(&mut self) -> &mut dyn SurrogateKernel

Return a mutable reference to the kernel.

pub fn n_train(&self) -> usize

Number of training samples.