Enum SelectionStrategy 

pub enum SelectionStrategy {
    Sequential,
    UncertaintySampling {
        uncertainty_weight: f64,
    },
    ExpectedInformationGain {
        temperature: f64,
    },
    CoverageAware {
        min_evaluations: usize,
        exploration_bonus: f64,
    },
}

Active learning selection strategy

Variants

Sequential

Sequential selection (current default behavior)

Selects candidates in order of their index, cycling through the population. Simple but not optimal for learning.

UncertaintySampling

Uncertainty sampling

Prioritizes candidates with the highest uncertainty (variance) in their fitness estimates. This helps reduce overall uncertainty in the ranking.

Fields

uncertainty_weight: f64

Weight for uncertainty vs coverage balance (default: 1.0) Higher values prioritize uncertain candidates more strongly

ExpectedInformationGain

Expected information gain (for pairwise comparisons)

Selects pairs of candidates where the comparison result is most uncertain (probability close to 0.5). This maximizes the expected reduction in entropy.

Fields

temperature: f64

Temperature for softmax selection (default: 1.0) Higher values make selection more random

CoverageAware

Coverage-aware selection

Ensures minimum coverage before exploring uncertain candidates. Good for balancing exploration with ensuring all candidates are evaluated at least some minimum number of times.

Fields

min_evaluations: usize

Minimum evaluations before considering a candidate “covered”

exploration_bonus: f64

Bonus weight for under-evaluated candidates

Implementations

impl SelectionStrategy

pub fn uncertainty_sampling(uncertainty_weight: f64) -> Self

Create uncertainty sampling strategy

pub fn information_gain(temperature: f64) -> Self

Create expected information gain strategy

pub fn coverage_aware(min_evaluations: usize, exploration_bonus: f64) -> Self

Create coverage-aware strategy

pub fn select_batch<G, R>( &self, candidates: &[Candidate<G>], aggregator: &FitnessAggregator, batch_size: usize, rng: &mut R, ) -> Vec<usize>

where G: EvolutionaryGenome, R: Rng,

Select a batch of candidates for evaluation

Arguments

• candidates - All candidates in the population
• aggregator - Fitness aggregator with current estimates
• batch_size - Number of candidates to select
• rng - Random number generator

Returns

Indices of selected candidates (into the candidates slice)

pub fn select_pair<G, R>( &self, candidates: &[Candidate<G>], aggregator: &FitnessAggregator, rng: &mut R, ) -> Option<(usize, usize)>

where G: EvolutionaryGenome, R: Rng,

Select a pair for pairwise comparison

Arguments

• candidates - All candidates in the population
• aggregator - Fitness aggregator with current estimates
• rng - Random number generator

Returns

Tuple of indices for the two candidates to compare

Trait Implementations

impl Clone for SelectionStrategy

fn clone(&self) -> SelectionStrategy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SelectionStrategy

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

Formats the value using the given formatter.

impl Default for SelectionStrategy

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for SelectionStrategy

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for SelectionStrategy

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.