Struct BaggedSGBT 

pub struct BaggedSGBT<L: Loss = SquaredLoss> { /* private fields */ }

Available on crate feature alloc only.

Bagged (Oza) SGBT ensemble for variance reduction.

Each of the M bags is an independent SGBT<L> trained on a Poisson(1)- weighted stream. Predictions are averaged across bags, reducing the variance of the ensemble without increasing bias.

This implements SGB(Oza) from Gunasekara et al. (2025), the streaming analogue of Breiman’s bootstrap aggregation adapted for gradient boosted trees with Hoeffding-bound splits.

Implementations

impl BaggedSGBT<SquaredLoss>

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

Create a new bagged SGBT with squared loss (regression).

Errors

Returns IrithyllError::InvalidConfig if n_bags < 1.

impl<L: Loss + Clone> BaggedSGBT<L>

pub fn with_loss(config: SGBTConfig, loss: L, n_bags: usize) -> Result<Self>

Create a new bagged SGBT with a custom loss function.

Each bag receives a unique seed derived from the config seed, ensuring diverse tree structures across bags.

Errors

Returns IrithyllError::InvalidConfig if n_bags < 1.

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

Train all bags on a single observation with Poisson(1) weighting.

For each bag, draws k ~ Poisson(1) and calls bag.train_one(sample) k times. On average, each bag sees the sample once, but the randomness creates diverse training sets across bags.

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

Train on a batch of observations.

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

Predict the raw output as the mean across all bags.

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

Predict with loss transform applied (e.g., sigmoid for logistic loss), averaged across bags.

pub fn predict_batch(&self, feature_matrix: &[Vec<f64>]) -> Vec<f64>

Batch prediction.

pub fn n_bags(&self) -> usize

Number of bags in the ensemble.

pub fn n_samples_seen(&self) -> u64

Total samples seen.

pub fn bags(&self) -> &[SGBT<L>]

Immutable access to all bags.

pub fn bag(&self, idx: usize) -> &SGBT<L>

Immutable access to a specific bag.

Panics

Panics if idx >= n_bags.

pub fn is_initialized(&self) -> bool

Whether the base prediction has been initialized for all bags.

pub fn reset(&mut self)

Reset all bags to initial state.

Trait Implementations

impl<L: Loss + Clone> Clone for BaggedSGBT<L>

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<L: Loss> Debug for BaggedSGBT<L>

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

Formats the value using the given formatter.

impl<L: Loss + Clone> StreamingLearner for BaggedSGBT<L>

fn train_one(&mut self, features: &[f64], target: f64, weight: f64)

Train on a single observation with explicit sample weight.

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

Predict the target for the given feature vector.

fn n_samples_seen(&self) -> u64

Total number of observations trained on since creation or last reset.

fn reset(&mut self)

Reset the model to its initial (untrained) state.

fn train(&mut self, features: &[f64], target: f64)

Train on a single observation with unit weight.

fn predict_batch(&self, feature_matrix: &[&[f64]]) -> Vec<f64>

Predict for each row in a feature matrix.