Struct FeatureBinner 

#[non_exhaustive]
pub struct FeatureBinner { /* private fields */ }

Quantile-based feature binner.

Transforms each feature column into u8 bin indices using quantile boundaries computed during fit(). Missing values (NaN) are mapped to bin 0; valid values to bins 1–255.

The binning is designed for histogram-based gradient boosting where the O(256) histogram scan replaces the O(n log n) sorted-split search.

Implementations

impl FeatureBinner

pub fn new() -> Self

Create a new binner with the default 256 max bins.

Example

use scry_learn::tree::FeatureBinner;
let binner = FeatureBinner::new();

pub fn max_bins(self, bins: usize) -> Self

Set the maximum number of bins (2..=256, default 256).

pub fn fit(&mut self, data: &Dataset) -> Result<()>

Compute bin edges from training data.

For each feature column, sorts the non-NaN values and picks equally-spaced quantile boundaries to create up to max_bins - 1 valid bins (bin 0 is reserved for missing).

pub fn transform(&self, data: &Dataset) -> Result<Vec<Vec<u8>>>

Map features to u8 bin indices.

Returns binned[feature_idx][sample_idx]. NaN → 0, valid → 1..=255.

pub fn fit_transform(&mut self, data: &Dataset) -> Result<Vec<Vec<u8>>>

Combined fit + transform.

pub fn n_bins_per_feature(&self) -> &[usize]

Number of bins per feature (including the missing-value bin).

pub fn bin_edges(&self) -> &[Vec<f64>]

Bin edges per feature.

Trait Implementations

impl Clone for FeatureBinner

fn clone(&self) -> FeatureBinner

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for FeatureBinner

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

Formats the value using the given formatter.

impl Default for FeatureBinner

fn default() -> Self

Returns the “default value” for a type.