Struct BayesianAdaptiveGA 

pub struct BayesianAdaptiveGA<P, F>
where
    P: GenomePrior,
    F: Fitness<Genome = P::Genome, Value = f64> + Clone + Send + Sync + 'static,
{ /* private fields */ }

A single-level Bayesian adaptive genetic algorithm.

See the module docs for the model. Operator step sizes are selected by Thompson sampling over per-operator Beta success posteriors, which are updated by conjugate Bayesian updates from observed improvement events.

What the prior is used for

This is a genetic algorithm, not a posterior sampler: the GenomePrior draws the initial population and defines the feasible region — a child whose encoding falls outside the prior’s support (a UniformBoxPrior’s box, a grammar’s depth limit) is discarded before evaluation and its parent keeps the slot, counted as a failed trial. The prior’s density does not otherwise enter selection; the Bayesian content is the conjugate operator-selection model. For the Boltzmann posterior itself use EvolutionChain or EvolutionSMC.

Implementations

impl<P, F> BayesianAdaptiveGA<P, F>

where P: GenomePrior, F: Fitness<Genome = P::Genome, Value = f64> + Clone + Send + Sync + 'static,

pub fn new( prior: P, fitness: F, population_size: usize, generations: usize, ) -> Self

Create a new adaptive GA with a default set of mutation step sizes.

pub fn with_step_sizes(self, sigmas: Vec<f64>) -> Self

Replace the operator step sizes (each gets a fresh Beta(1, 1) prior).

pub fn with_mutation_rate(self, rate: f64) -> Self

Set the coordinate mutation probability.

pub fn with_tournament_size(self, size: usize) -> Self

Set the tournament size for parent selection.

pub fn run<R: Rng>( &mut self, rng: &mut R, ) -> BayesianAdaptiveGAResult<P::Genome>

Run the adaptive GA.