Struct InfoGainMaximizer 

pub struct InfoGainMaximizer<E>
where
    E: ErrorWrapper,
{ /* private fields */ }

Depth-2 solver that maximises the information gain of the tree rather than minimising its error. Used by LGDT for a less greedy lookahead.

Trait Implementations

impl Default for InfoGainMaximizer<NativeError>

fn default() -> Self

Returns the “default value” for a type.

impl<E> OptimalDepth2Tree for InfoGainMaximizer<E>

where E: ErrorWrapper,

fn find_optimal_depth_one_tree( &self, min_sup: usize, cover: &mut Cover, provided_candidates: Option<&[usize]>, ) -> Result<Tree, FitError>

The best tree with a single test.

fn find_optimal_depth_two_tree( &self, min_sup: usize, cover: &mut Cover, provided_candidates: Option<&[usize]>, ) -> Result<Tree, FitError>

The best tree with at most two levels of tests.

fn error(&self, distribution: &[usize]) -> (f64, f64)

(error, prediction) of a leaf with the given class counts.

fn fit( &self, min_sup: usize, depth: usize, cover: &mut Cover, provided_candidates: Option<&[usize]>, ) -> Result<Tree, FitError>

Finds the best tree of the given depth (1 or 2) on the instances of cover, using only provided_candidates when given.

fn get_candidates( &self, cover: &mut Cover, min_sup: usize, provided_candidates: Option<&[usize]>, ) -> Vec<usize>

Features that split the instances of cover with at least min_sup instances on each side.