Struct Belief 

pub struct Belief { /* private fields */ }

A Gaussian process fitted to a run’s completed trials.

Implementations

impl Belief

pub fn new(controller: &Controller) -> Option<Self>

Fits the model to controller’s completed trials, or returns None with fewer than MIN_POINTS of them or a fit that fails.

The training data is GP-EI’s: every completed trial with a finite value and both parameters, in trial-number order, mapped into the unit square through the declared space, capped at the newest [MAX_TRAIN]. For CARBS’s two-objective study the value is the first objective.

pub fn new_with_training_data( inputs: &[Vec<f64>], targets: &[f64], ) -> Option<Self>

Fits GP-EI’s collected training pairs without accessing a controller. Returns None for fewer than MIN_POINTS, mismatched lengths, or a failed fit. Inputs must be unit-square points.

pub fn points(&self) -> usize

How many trials the model was fitted to.

pub const fn lengthscale(&self) -> f64

The kernel lengthscale, in unit-square units.

pub const fn best(&self) -> f64

The lowest value the model was fitted to.

pub fn predict(&self, unit: [f64; 2]) -> (f64, f64)

The posterior mean and standard deviation at a unit-square point, in the function’s units.

pub fn log_ei(&self, unit: [f64; 2]) -> f64

The natural log of the expected improvement over the lowest fitted value at a unit-square point (standardized units, as GP-EI scores it).

pub fn sample(&self, map: ModelMap, unit: [f64; 2]) -> f64

The value map shows at a unit-square point.

Trait Implementations

impl Debug for Belief

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.