Struct std_dev::regression::spiral::Options

pub struct Options {
    pub exponent_coefficient: f64,
    pub angle_coefficient: f64,
    pub num_lockon: usize,
    pub samples_per_rotation: f64,
    pub range: Range<f64>,
    pub turns: f64,
}

Available on crate feature regression only.

Options for the spiral.

This also implements most estimator traits. These all use the manhattan distance as their fitness function. The estimators have O(n) runtime performance and O(1) size performance.

Polynomial estimator only supports degrees 1 & 2.

See module-level documentation for more info about concepts.

The Self::range limit samples on the spiral. This causes a upper limit for coefficients and a lower “precision” cutoff. You can increase Self::exponent_coefficient if you know your data will have large numbers. The algorithm will increase the size of the spiral of a line outside the spiral is found.

Use a graphing tool (e.g. Desmos) and plot r=ae^(kθ). a is Self::exponent_coefficient. k is Self::angle_coefficient.

To keep the same “size” of the spiral, you have to multiply both ends of Self::range with the factor of Self::angle_coefficient.

Performance

The methods are O(1)*O(fitness_function) where O(fitness_function) is the time complexity of your fitness_function. That’s often O(n) as you’d probably in some way sum up the points relative to the model.

The following options affect the performance as follows (roughly, no coefficients) O(num_lockon * samples_per_rotation * range.length).

Keep in mind you should probably not lower Self::angle_coefficient bellow 0.15 if you don’t increase the Self::range.

Fields

exponent_coefficient: f64

The initial scale of the spiral.

This gets adjusted when locking on.

angle_coefficient: f64

The “density” of the spiral.

num_lockon: usize

How many lockons we are allowed to do. This is a max value.

samples_per_rotation: f64

The count of samples per each rotation in the spiral.

range: Range<f64>

The range of angles to test.

turns: f64

The turns of the 3d spiral when using three_variable_optimization. Frequency of turns per sphere. Is unnecessary to turn up when Self::samples_per_rotation is low.

Implementations

impl Options

pub fn new(level: u8) -> Self

Create a new set of options with good defaults.

level is allowed to be in the range [1..=9]. Higher values are more “precise” - they take longer but are also way (especially level>4) more likely to return good results.

Expect a 2-4x increase in runtime per increment of level.

Panics

Panics if !(1..=9).contains(level).

Trait Implementations

impl Clone for Options

fn clone(&self) -> Options

Returns a copy of the value.

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

Performs copy-assignment from source.

impl CosecantEstimator for Options

fn model_cosecant( &self, predictors: &[f64], outcomes: &[f64], max_frequency: f64 ) -> CosecantCoefficients

Model the CosecantCoefficients from predictors and outcomes.

fn boxed_cosecant(self) -> Box<dyn CosecantEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl CosineEstimator for Options

fn model_cosine( &self, predictors: &[f64], outcomes: &[f64], max_frequency: f64 ) -> CosineCoefficients

Model the CosineCoefficients from predictors and outcomes.

fn boxed_cosine(self) -> Box<dyn CosineEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl CotangentEstimator for Options

fn model_cotangent( &self, predictors: &[f64], outcomes: &[f64], max_frequency: f64 ) -> CotangentCoefficients

Model the CotangentCoefficients from predictors and outcomes.

fn boxed_cotangent(self) -> Box<dyn CotangentEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl Debug for Options

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

Formats the value using the given formatter.

impl Default for Options

fn default() -> Self

Returns the “default value” for a type.

impl ExponentialEstimator for Options

fn model_exponential( &self, predictors: &[f64], outcomes: &[f64] ) -> ExponentialCoefficients

Model the ExponentialCoefficients from predictors and outcomes.

fn boxed_exponential(self) -> Box<dyn ExponentialEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl LinearEstimator for Options

fn model_linear( &self, predictors: &[f64], outcomes: &[f64] ) -> LinearCoefficients

Model the LinearCoefficients from predictors and outcomes.

fn boxed_linear(self) -> Box<dyn LinearEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl LogisticEstimator for Options

fn model_logistic( &self, predictors: &[f64], outcomes: &[f64] ) -> LogisticCoefficients

Model the LogisticCoefficients from predictors and outcomes.

fn boxed_logistic(self) -> Box<dyn LogisticEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl PartialEq<Options> for Options

fn eq(&self, other: &Options) -> bool

This method tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

This method tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl PolynomialEstimator for Options

fn model_polynomial( &self, predictors: &[f64], outcomes: &[f64], degree: usize ) -> PolynomialCoefficients

Model the PolynomialCoefficients from predictors and outcomes. Also takes a degree of the target polynomial. Some estimators may panic when degree is out of their range.

fn boxed_polynomial(self) -> Box<dyn PolynomialEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl PowerEstimator for Options

fn model_power(&self, predictors: &[f64], outcomes: &[f64]) -> PowerCoefficients

Model the PowerCoefficients from predictors and outcomes.

fn boxed_power(self) -> Box<dyn PowerEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl SecantEstimator for Options

fn model_secant( &self, predictors: &[f64], outcomes: &[f64], max_frequency: f64 ) -> SecantCoefficients

Model the SecantCoefficients from predictors and outcomes.

fn boxed_sesecant(self) -> Box<dyn SecantEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl SineEstimator for Options

fn model_sine( &self, predictors: &[f64], outcomes: &[f64], max_frequency: f64 ) -> SineCoefficients

Model the SineCoefficients from predictors and outcomes.

fn boxed_sine(self) -> Box<dyn SineEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl TangentEstimator for Options

fn model_tangent( &self, predictors: &[f64], outcomes: &[f64], max_frequency: f64 ) -> TangentCoefficients

Model the TangentCoefficients from predictors and outcomes.

fn boxed_tangent(self) -> Box<dyn TangentEstimator>where Self: Sized + 'static,

Put this estimator in a box. This is useful for conditionally choosing different estimators.

impl StructuralPartialEq for Options