Struct ContinuousUMDA 

pub struct ContinuousUMDA<F, Fit, Term>
where
    F: FitnessValue,
{ /* private fields */ }

UMDA for continuous optimization

Implementations

impl<F, Fit, Term> ContinuousUMDA<F, Fit, Term>

where F: FitnessValue + Send, Fit: Fitness<Genome = RealVector, Value = F> + Sync, Term: TerminationCriterion<RealVector, F>,

pub fn builder() -> UMDABuilder<RealVector, F, (), ()>

Create a builder for continuous UMDA

pub fn run<R: Rng>( &self, rng: &mut R, ) -> Result<EvolutionResult<RealVector, F>, EvolutionError>

Run the UMDA algorithm.

pub fn run_with_callback<R: Rng, Cb: FnMut(usize, f64) -> bool>( &self, rng: &mut R, on_generation: Cb, ) -> Result<EvolutionResult<RealVector, F>, EvolutionError>

Run UMDA, invoking on_generation(generation, best_fitness) once per generation before that generation is sampled/evaluated.

Returning false from the callback cancels the run early and returns the best-so-far result (AUDIT EV-34: lets the WASM layer report per-generation progress and support cancellation without a separate step API).

pub fn run_with_model<R: Rng>( &self, rng: &mut R, ) -> Result<(EvolutionResult<RealVector, F>, ContinuousUnivariateModel), EvolutionError>

Run the UMDA algorithm and also return the final learned univariate model (EV-10), so callers/tests can inspect whether the distribution actually converged toward the optimum rather than merely tracking a best-so-far.