Struct SurrogateFitness 

pub struct SurrogateFitness {
    pub posterior: Arc<TastePosterior>,
    pub standardizer: Arc<Standardizer>,
    pub phrase: PhraseSpec,
    pub memo: RenderMemo,
}

Expected posterior utility as a scalar fitness over patch terms.

Fields

posterior: Arc<TastePosterior>

The fitted taste posterior.

standardizer: Arc<Standardizer>

The standardizer the posterior’s observations were made under.

phrase: PhraseSpec

The audition stimulus (must match the one used for observations).

memo: RenderMemo

The engine’s featurization memo.

Not an optimization detail — it is what makes the MH walk affordable. adaptive_single_site_mh executes the model twice per step: once to re-score the current trace, which is bit-identically the tree the previous step accepted and therefore already featurized, and once for the proposal. Without a memo, one render in two is a recomputation of a number the walk already has.

Trait Implementations

impl Clone for SurrogateFitness

fn clone(&self) -> SurrogateFitness

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SurrogateFitness

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

Formats the value using the given formatter.

impl Fitness for SurrogateFitness

type Genome = PatchTree

The genome type being evaluated

type Value = f64

The fitness value type

fn evaluate(&self, genome: &PatchTree) -> f64

Evaluate fitness (higher = better by convention)

fn as_log_likelihood(&self, genome: &Self::Genome, temperature: f64) -> f64

Convert fitness to log-likelihood for probabilistic selection

fn gradient(&self, _genome: &Self::Genome) -> Option<Vec<f64>>

Optional: Provide gradient for gradient-assisted mutation