Trait forester::data::TrainingData

pub trait TrainingData<Sample>: DataSet<Sample> where
    Sample: SampleDescription, {
    type Criterion: SplitCriterion<Sample::Target>;
    fn n_samples(&self) -> usize;
    fn gen_split_feature(&self) -> Sample::ThetaSplit;
    fn train_leaf_predictor(&self) -> Sample::ThetaLeaf;
    fn feature_bounds(
        &self, 
        theta: &Sample::ThetaSplit
    ) -> (Sample::Feature, Sample::Feature);

    fn all_split_features(
        &self
    ) -> Option<Box<Iterator<Item = Sample::ThetaSplit>>> { ... }
}

Data set that can be used for training decision trees

Associated Types

type Criterion: SplitCriterion<Sample::Target>

Required Methods

fn n_samples(&self) -> usize

Return number of samples in the data set

fn gen_split_feature(&self) -> Sample::ThetaSplit

Generate a new split feature (typically, this will be randomized)

fn train_leaf_predictor(&self) -> Sample::ThetaLeaf

Train a new leaf predictor

fn feature_bounds(
    &self,
    theta: &Sample::ThetaSplit
) -> (Sample::Feature, Sample::Feature)

Return minimum and maximum value of a feature

Provided Methods

fn all_split_features(&self) -> Option<Box<Iterator<Item = Sample::ThetaSplit>>>

Return an iterator over all features.

Data sets where this is not possible (infinite feature space, anyone?) should return None instead, which is also the default impl.

Implementations on Foreign Types

impl<'a, X> TrainingData<Sample<'a, X, f64>> for [Sample<'a, X, f64>] where     X: Clone + PartialOrd + SampleRange + Bounded + SplitBetween,

type Criterion = VarianceCriterion

fn n_samples(&self) -> usize

fn gen_split_feature(&self) -> usize

fn train_leaf_predictor(&self) -> f64

fn feature_bounds(&self, theta: &usize) -> (X, X)

fn all_split_features(&self) -> Option<Box<Iterator<Item = Sample::ThetaSplit>>>

impl<'a, X> TrainingData<Sample<'a, X, Classes>> for [Sample<'a, X, Classes>] where     X: Clone + PartialOrd + SampleRange + Bounded + SplitBetween,

type Criterion = GiniCriterion

fn n_samples(&self) -> usize

fn gen_split_feature(&self) -> usize

fn train_leaf_predictor(&self) -> ClassCounts

fn feature_bounds(&self, theta: &usize) -> (X, X)

fn all_split_features(&self) -> Option<Box<Iterator<Item = Sample::ThetaSplit>>>

Implementors