Struct changeforest::Control

pub struct Control {
    pub minimal_relative_segment_length: f64,
    pub minimal_gain_to_split: Option<f64>,
    pub model_selection_alpha: f64,
    pub model_selection_n_permutations: usize,
    pub number_of_wild_segments: usize,
    pub seeded_segments_alpha: f64,
    pub seed: u64,
    pub random_forest_parameters: RandomForestParameters,
    pub forbidden_segments: Option<Vec<(usize, usize)>>,
}

Storage container for hyperparameters.

Fields

minimal_relative_segment_length: f64

Segments with length smaller than 2 * n * minimal_relative_segment_length will not be split.

minimal_gain_to_split: Option<f64>

Only keep split point if the gain exceeds minimal_gain_to_split. Relevant for change in mean. Note that this is relative to the number of observations. Use value motivated by BIC minimal_gain_to_split = log(n_samples) * n_features / n_samples if None.

model_selection_alpha: f64

Type two error in model selection to be approximated. Relevant for classifier- based changepoint detection.

model_selection_n_permutations: usize

Number of permutations for model selection in classifier-based change point detection.

number_of_wild_segments: usize

Number of randomly drawn segments. Corresponds to parameter M in https://arxiv.org/pdf/1411.0858.pdf.

seeded_segments_alpha: f64

Decay parameter in seeded binary segmentation. Should be in [1/2, 1), with a value close to 1 resulting in many segments. Corresponds to \alpha in https://arxiv.org/pdf/2002.06633.pdf.

seed: u64

Seed used for segmentation.

random_forest_parameters: RandomForestParameters

Hyperparameters for random forests.

forbidden_segments: Option<Vec<(usize, usize)>>

Segments of indexes were no segmentation is allowed.

Implementations

impl Control

pub fn default() -> Control

pub fn with_minimal_relative_segment_length( self, minimal_relative_segment_length: f64 ) -> Self

pub fn with_minimal_gain_to_split( self, minimal_gain_to_split: Option<f64> ) -> Self

pub fn with_model_selection_alpha(self, model_selection_alpha: f64) -> Self

pub fn with_model_selection_n_permutations( self, model_selection_n_permutations: usize ) -> Self

pub fn with_number_of_wild_segments( self, number_of_wild_segments: usize ) -> Self

pub fn with_seeded_segments_alpha(self, seeded_segments_alpha: f64) -> Self

pub fn with_seed(self, seed: u64) -> Self

pub fn with_random_forest_parameters( self, random_forest_parameters: RandomForestParameters ) -> Self

pub fn with_forbidden_segments( self, forbidden_segments: Option<Vec<(usize, usize)>> ) -> Self

Trait Implementations

impl Clone for Control

fn clone(&self) -> Control

Returns a copy of the value.

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

Performs copy-assignment from source.