Struct LGDT 

pub struct LGDT<D>
where
    D: OptimalDepth2Tree + ?Sized,
{ /* private fields */ }

LGDT, a less greedy decision tree learner.

Like CART, it builds the tree top-down one test at a time, but it chooses each test by solving a depth-2 tree at the node with a OptimalDepth2Tree solver and keeping its root. This two-level lookahead is cheap thanks to the depth-2 solver and makes the tree much less myopic than a purely greedy one.

Kiossou, Schaus, Nijssen and Aglin, Efficient Lookahead Decision Trees (IDA 2024). Build one with LGDTBuilder or the functions in factories.

Implementations

impl<D> LGDT<D>

where D: OptimalDepth2Tree + ?Sized,

pub fn config(&self) -> &BaseSearchConfig

The settings of the search.

Trait Implementations

impl<D> TreeSearchAlgorithm for LGDT<D>

where D: OptimalDepth2Tree + ?Sized,

fn fit(&mut self, cover: &mut Cover) -> Result<(), FitError>

Learns a tree on the instances of cover.

fn tree(&self) -> &Tree

The tree learned by the last fit.

fn error(&self) -> f64

Training error of the learned tree.

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

Features that can split the current node. See find_valid_split_attributes.