Struct Run 

pub struct Run { /* private fields */ }

A run in progress: its distribution, the current generation’s candidates, and the best point so far.

Implementations

impl Run

pub fn parameters(&self) -> &Parameters

The strategy’s parameters.

pub fn candidates(&self) -> &[Vec<f64>]

The current generation’s candidates, λ points, each one value per variable, as the model is to be given them: an integer variable’s value is encoded, a whole number within its bounds. Empty once the run has stopped.

pub fn generation(&self) -> usize

The generations finished so far.

pub fn mean(&self) -> &[f64]

The distribution’s mean. An integer variable’s is a real number, not encoded, and may lie outside its bounds.

pub fn sigma(&self) -> f64

The step size σ.

pub fn covariance(&self) -> &[f64]

The covariance C, row-major, of the variables divided by their steps: the distribution of draws about the mean has covariance σ² S A C A S, with S the steps and A Run::margin_scale on diagonals.

pub fn margin_scale(&self) -> &[f64]

The margin’s diagonal A, one entry per variable: an integer variable’s draws spread A times as far as σ and C alone would spread them (Variable::integer). 1 for every continuous variable.

pub fn integer_margin(&self) -> f64

The margin α = 1/(n λ): the least chance, each generation, that an integer variable’s draw falls on a value other than the one its mean is nearest (at least α/2 on each side when the mean’s value has neighbours on both).

pub fn stopped(&self) -> Option<Stop>

Why the run stopped, or None while it runs.

pub fn tell(&mut self, values: &[f64]) -> Result<Option<Optimum>, AnalysisError>

Takes the model’s values at Run::candidates, in their order, and moves the distribution. Returns what the run found once it stops, and None while it goes on, with the next generation’s candidates drawn.

A value may be +∞, for a candidate the model can’t evaluate (a flight that fails, say): it ranks below every finite one.

Errors

AnalysisError::Length for a number of values other than the candidates’, and AnalysisError::Output for a NaN or −∞, with the index of its evaluation in the run. Telling a stopped run is AnalysisError::Length too, as it has no candidates.

pub fn tell_constrained( &mut self, evaluations: &[Evaluation], ) -> Result<Option<Optimum>, AnalysisError>

Run::tell with each candidate’s constraint violation: candidates are ranked by Deb’s feasibility rules (Evaluation::rank), the best point is the best by the same rules, and a target counts only for a point that keeps every constraint. The stop on values (Stop::TolFun) looks only at generations whose best keeps them, and at the values of this generation’s candidates that keep them.

Told only violations of zero, a run is bit for bit Run::tell’s.

Errors

Run::tell’s, and AnalysisError::Domain for a violation that is negative or NaN (+∞ is a failed candidate’s, Evaluation::failed).

Trait Implementations

impl Clone for Run

fn clone(&self) -> Run

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Run

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

Formats the value using the given formatter.