Struct SGBTConfig 

pub struct SGBTConfig {

    pub n_steps: usize,
    pub learning_rate: f64,
    pub feature_subsample_rate: f64,
    pub max_depth: usize,
    pub n_bins: usize,
    pub lambda: f64,
    pub gamma: f64,
    pub grace_period: usize,
    pub delta: f64,
    pub drift_detector: DriftDetectorType,
    pub variant: SGBTVariant,
    pub seed: u64,
    pub initial_target_count: usize,
    pub leaf_half_life: Option<usize>,
    pub max_tree_samples: Option<u64>,
    pub adaptive_mts: Option<(u64, f64)>,
    pub adaptive_mts_floor: f64,
    pub proactive_prune_interval: Option<u64>,
    pub split_reeval_interval: Option<usize>,
    pub feature_names: Option<Vec<String>>,
    pub feature_types: Option<Vec<FeatureType>>,
    pub gradient_clip_sigma: Option<f64>,
    pub monotone_constraints: Option<Vec<i8>>,
    pub quality_prune_alpha: Option<f64>,
    pub quality_prune_threshold: f64,
    pub quality_prune_patience: u64,
    pub error_weight_alpha: Option<f64>,
    pub uncertainty_modulated_lr: bool,
    pub scale_mode: ScaleMode,
    pub empirical_sigma_alpha: f64,
    pub max_leaf_output: Option<f64>,
    pub adaptive_leaf_bound: Option<f64>,
    pub adaptive_depth: Option<f64>,
    pub min_hessian_sum: Option<f64>,
    pub huber_k: Option<f64>,
    pub shadow_warmup: Option<usize>,
    pub leaf_model_type: LeafModelType,
    pub packed_refresh_interval: u64,
    pub hoeffding_r: Option<f64>,
}

Available on crate feature alloc only.

Configuration for the SGBT ensemble.

All numeric parameters are validated at build time via SGBTConfigBuilder.

Fields

n_steps: usize

Number of sequential boosting steps (trees) in the ensemble.

learning_rate: f64

Shrinkage applied to each tree’s contribution.

feature_subsample_rate: f64

Fraction of features sampled per split candidate evaluation.

max_depth: usize

Maximum tree depth (split decisions per root-to-leaf path).

n_bins: usize

Number of histogram bins per feature for split evaluation.

lambda: f64

L2 regularization on leaf values (lambda in XGBoost objective).

gamma: f64

Minimum gain required to accept a split (gamma in XGBoost).

grace_period: usize

Hoeffding bound grace period: minimum samples before any split is considered.

delta: f64

Hoeffding bound confidence parameter (smaller = more conservative splits).

drift_detector: DriftDetectorType

Drift detector configuration used for each boosting step.

variant: SGBTVariant

Which SGBT algorithm variant to use.

seed: u64

Random seed for feature subsampling and tie-breaking.

initial_target_count: usize

Number of target samples to buffer before fixing the base prediction.

leaf_half_life: Option<usize>

Leaf value exponential half-life (samples). None disables decay.

max_tree_samples: Option<u64>

Maximum training samples per tree before the tree is replaced. None = unlimited.

adaptive_mts: Option<(u64, f64)>

Adaptive max-tree-samples: (warmup_samples, percentile). Derives the threshold from the empirical sample distribution rather than a fixed value.

adaptive_mts_floor: f64

Floor on the adaptive MTS threshold (prevents degenerate collapses).

proactive_prune_interval: Option<u64>

Proactive pruning interval in samples. None disables proactive pruning.

split_reeval_interval: Option<usize>

Split re-evaluation interval (samples). None disables periodic re-evaluation.

feature_names: Option<Vec<String>>

Human-readable names for each feature column (used in diagnostics/explainability).

feature_types: Option<Vec<FeatureType>>

Per-feature type hints (continuous vs. categorical) for the binning strategy.

gradient_clip_sigma: Option<f64>

Gradient clipping: clip to sigma * gradient_clip_sigma. None disables.

monotone_constraints: Option<Vec<i8>>

Per-feature monotonicity constraints: 1 = increasing, -1 = decreasing, 0 = none.

quality_prune_alpha: Option<f64>

Quality pruning significance level (alpha). None disables quality pruning.

quality_prune_threshold: f64

Minimum EWMA contribution magnitude for a step to survive quality pruning.

quality_prune_patience: u64

Consecutive low-contribution rounds before a step is pruned.

error_weight_alpha: Option<f64>

Error-weighted sample importance EWMA alpha. None disables weighting.

uncertainty_modulated_lr: bool

Whether to modulate the learning rate by the model’s estimated uncertainty.

scale_mode: ScaleMode

Strategy for computing the scale (uncertainty) head in distributional mode.

empirical_sigma_alpha: f64

EWMA smoothing factor for the empirical sigma estimate.

max_leaf_output: Option<f64>

Maximum absolute leaf output value (clamp). None = no clamp.

adaptive_leaf_bound: Option<f64>

Adaptive leaf output bound derived from the rolling leaf magnitude. None disables.

adaptive_depth: Option<f64>

Adaptive depth fractional limit derived from leaf sample counts. None disables.

min_hessian_sum: Option<f64>

Minimum hessian sum required to accept a split. None = no minimum.

huber_k: Option<f64>

Huber loss delta override (used when loss is Huber). None uses the default.

shadow_warmup: Option<usize>

Shadow warmup: alternate trees train for this many samples before replacing. None = immediate.

leaf_model_type: LeafModelType

Leaf model type (constant, linear, MLP). Default is constant value.

packed_refresh_interval: u64

Interval (samples) between packed-node cache refreshes for fast inference.

hoeffding_r: Option<f64>

Override for the Hoeffding bound range R. None uses the default (1.0).

Implementations

impl SGBTConfig

pub fn builder() -> SGBTConfigBuilder

Create a new builder for SGBTConfig.

Trait Implementations

impl Clone for SGBTConfig

fn clone(&self) -> SGBTConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SGBTConfig

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

Formats the value using the given formatter.

impl Default for SGBTConfig

fn default() -> Self

Returns the “default value” for a type.

impl Display for SGBTConfig

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

Formats the value using the given formatter.

impl PartialEq for SGBTConfig

fn eq(&self, other: &SGBTConfig) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for SGBTConfig