Enum LeafModelType 

pub enum LeafModelType {
    ClosedForm,
    Linear {
        learning_rate: f64,
        decay: Option<f64>,
        use_adagrad: bool,
    },
    MLP {
        hidden_size: usize,
        learning_rate: f64,
        decay: Option<f64>,
    },
    Adaptive {
        promote_to: Box<LeafModelType>,
    },
}

Describes which leaf model architecture to use.

Used by tree builders to construct fresh leaf models when creating new leaves.

Variants

• ClosedForm – Standard constant leaf weight. Zero overhead (default).
• Linear – Per-leaf online ridge regression with AdaGrad optimization. Each leaf learns a local linear surface w . x + b. Recommended for low-depth trees (depth 2–4). Optional decay for concept drift.
• MLP – Per-leaf single-hidden-layer neural network with ReLU. Optional decay for concept drift.
• Adaptive – Starts as closed-form, auto-promotes to promote_to when the Hoeffding bound confirms it is statistically superior. Uses the tree’s existing delta parameter – no arbitrary thresholds.

Variants

ClosedForm

Standard closed-form leaf weight.

Linear

Online ridge regression, optionally with AdaGrad optimization.

decay: optional exponential weight decay for non-stationary streams. Typical values: 0.999 (slow drift) to 0.99 (fast drift).

use_adagrad: when true, per-weight squared gradient accumulators give each feature its own adaptive learning rate. When false (default), all weights share a single Newton-scaled learning rate.

Fields

learning_rate: f64

decay: Option<f64>

use_adagrad: bool

MLP

Single hidden layer MLP with the given hidden size and learning rate.

decay: optional exponential weight decay for non-stationary streams.

Fields

hidden_size: usize

learning_rate: f64

decay: Option<f64>

Adaptive

Adaptive leaf that starts as closed-form and auto-promotes when the Hoeffding bound confirms the promoted model is better.

The promote_to field specifies the shadow model type to evaluate against the default closed-form baseline.

Fields

promote_to: Box<LeafModelType>

Implementations

impl LeafModelType

pub fn create(&self, seed: u64, delta: f64) -> Box<dyn LeafModel>

Create a fresh boxed leaf model of this type.

The seed parameter controls deterministic initialization (MLP weights, adaptive model seeding). The delta parameter is the Hoeffding bound confidence level, used by Adaptive leaves for promotion decisions. For non-adaptive types, delta is unused.

Trait Implementations

impl Clone for LeafModelType

fn clone(&self) -> LeafModelType

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LeafModelType

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

Formats the value using the given formatter.

impl Default for LeafModelType

fn default() -> LeafModelType

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for LeafModelType

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for LeafModelType

fn eq(&self, other: &LeafModelType) -> 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 Serialize for LeafModelType

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for LeafModelType