Struct MulticlassSGBT 

pub struct MulticlassSGBT { /* private fields */ }

Available on crate feature alloc only.

Multi-class SGBT using one-vs-rest committee of ensembles.

Each class gets its own SGBT<SoftmaxLoss> trained with softmax (logistic per-class) loss. The concrete loss type is monomorphized for each committee – no Box<dyn Loss> overhead.

Predictions are softmax-normalized across all class committees.

Implementations

impl MulticlassSGBT

pub fn new(config: SGBTConfig, n_classes: usize) -> Result<Self>

Create a new multi-class SGBT.

Errors

Returns IrithyllError::InvalidConfig if n_classes < 2.

pub fn train_one(&mut self, sample: &impl Observation)

Train on a single observation.

The observation’s target should be the class index as f64 (0.0, 1.0, 2.0, …).

Uses SampleRef internally to avoid cloning feature vectors for each committee (N classes = 0 clones instead of N clones).

pub fn train_batch<O: Observation>(&mut self, samples: &[O])

Train on a batch of observations.

pub fn predict_proba(&self, features: &[f64]) -> Vec<f64>

Predict class probabilities via softmax normalization.

Returns a vector of length n_classes summing to ~1.0.

pub fn predict(&self, features: &[f64]) -> usize

Predict the most likely class.

pub fn n_classes(&self) -> usize

Number of classes.

pub fn n_samples_seen(&self) -> u64

Total samples trained.

pub fn reset(&mut self)

Reset all committees.

Trait Implementations

impl Debug for MulticlassSGBT

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

Formats the value using the given formatter.