Struct UpperConfidenceBound 

pub struct UpperConfidenceBound {
    pub kappa: f64,
}

Upper Confidence Bound acquisition function (adapted for minimization).

For minimization, we use the Lower Confidence Bound: LCB(x) = -(mu(x) - kappa * sigma(x))

We negate so that higher acquisition values are still better to sample.

The kappa parameter controls exploration:

• kappa = 0: pure exploitation (just the mean)
• kappa > 0: more exploration (wider confidence interval)

A common adaptive schedule is kappa = sqrt(2 * ln(t * d^2 * pi^2 / (6 * delta))) where t is the iteration, d is the dimension, delta is a confidence parameter.

Fields

kappa: f64

Exploration parameter (>= 0).

Implementations

impl UpperConfidenceBound

pub fn new(kappa: f64) -> Self

Trait Implementations

impl AcquisitionFn for UpperConfidenceBound

fn evaluate( &self, x: &ArrayView1<'_, f64>, surrogate: &GpSurrogate, ) -> OptimizeResult<f64>

Evaluate the acquisition function at a single point.

fn name(&self) -> &str

Name of the acquisition function.

fn evaluate_batch( &self, x: &Array2<f64>, surrogate: &GpSurrogate, ) -> OptimizeResult<Array1<f64>>

Evaluate the acquisition function at multiple points (batch).

impl Clone for UpperConfidenceBound

fn clone(&self) -> UpperConfidenceBound

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for UpperConfidenceBound

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

Formats the value using the given formatter.

impl Default for UpperConfidenceBound

fn default() -> Self

Returns the “default value” for a type.