Struct ActiveLearningSearch 

pub struct ActiveLearningSearch<S> { /* private fields */ }

Active Learning search optimizer.

Wraps any base search strategy and adds uncertainty-based stopping. Implements the “Pull System” from Lean manufacturing - only generates samples while uncertainty is high (Settles, 2009).

Muda Elimination (Waste Reduction)

Traditional batch generation (“Push System”) produces many redundant samples. Active Learning stops when confidence saturates, eliminating overproduction.

Example

use aprender::automl::{ActiveLearningSearch, RandomSearch, SearchSpace, SearchStrategy};
use aprender::automl::params::RandomForestParam as RF;

let space = SearchSpace::new()
    .add_continuous(RF::NEstimators, 10.0, 500.0);

let base = RandomSearch::new(1000).with_seed(42);
let mut search = ActiveLearningSearch::new(base)
    .with_uncertainty_threshold(0.1)  // Stop when uncertainty < 0.1
    .with_min_samples(10);            // Need at least 10 samples

// Pull system: generate until confident
while !search.should_stop() {
    let trials = search.suggest(&space, 5);
    if trials.is_empty() { break; }
    // ... evaluate trials ...
    // search.update(&results);
}

Implementations

impl<S> ActiveLearningSearch<S>

pub fn new(base: S) -> Self

Create active learning wrapper around base strategy.

pub fn with_uncertainty_threshold(self, threshold: f64) -> Self

Set uncertainty threshold for stopping.

When estimated uncertainty drops below this threshold, should_stop() returns true.

pub fn with_min_samples(self, min: usize) -> Self

Set minimum samples before stopping is considered.

pub fn should_stop(&self) -> bool

Check if optimization should stop due to low uncertainty.

Returns true when:

1. At least min_samples have been evaluated
2. Uncertainty is below uncertainty_threshold

pub fn uncertainty(&self) -> f64

Get current uncertainty estimate.

Uses coefficient of variation (std_dev / mean) as uncertainty metric. Returns infinity if not enough samples.

pub fn sample_count(&self) -> usize

Get number of samples collected.

Trait Implementations

impl<S: Clone> Clone for ActiveLearningSearch<S>

fn clone(&self) -> ActiveLearningSearch<S>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<S: Debug> Debug for ActiveLearningSearch<S>

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

Formats the value using the given formatter.

impl<S, P> SearchStrategy<P> for ActiveLearningSearch<S>

where S: SearchStrategy<P>, P: ParamKey,

fn suggest(&mut self, space: &SearchSpace<P>, n: usize) -> Vec<Trial<P>>

Generate candidate configurations to evaluate.

fn update(&mut self, results: &[TrialResult<P>])

Update strategy with evaluation results (for adaptive methods).