Struct ModelDiagnostics 

pub struct ModelDiagnostics {
    pub trees: Vec<TreeDiagnostic>,
    pub location_trees: Vec<TreeDiagnostic>,
    pub scale_trees: Vec<TreeDiagnostic>,
    pub feature_split_counts: Vec<usize>,
    pub location_base: f64,
    pub scale_base: f64,
    pub empirical_sigma: f64,
    pub scale_mode: ScaleMode,
    pub scale_trees_active: usize,
    pub auto_bandwidths: Vec<f64>,
    pub ensemble_grad_mean: f64,
    pub ensemble_grad_std: f64,
}

Available on crate feature alloc only.

Full model diagnostics for DistributionalSGBT.

Contains per-tree summaries, feature usage, base predictions, and empirical σ state.

Fields

trees: Vec<TreeDiagnostic>

Per-tree diagnostic summaries (location trees first, then scale trees).

location_trees: Vec<TreeDiagnostic>

Location trees only (view into trees).

scale_trees: Vec<TreeDiagnostic>

Scale trees only (view into trees).

feature_split_counts: Vec<usize>

How many trees each feature is used in (split count per feature).

location_base: f64

Base prediction for location (mean).

scale_base: f64

Base prediction for scale (log-sigma).

empirical_sigma: f64

Current empirical σ (sqrt(ewma_sq_err)), always available.

scale_mode: ScaleMode

Scale mode in use.

scale_trees_active: usize

Number of scale trees that actually split (>1 leaf). 0 = frozen chain.

auto_bandwidths: Vec<f64>

Per-feature auto-calibrated bandwidths for smooth prediction. f64::INFINITY means that feature uses hard routing.

ensemble_grad_mean: f64

Ensemble-level gradient running mean.

ensemble_grad_std: f64

Ensemble-level gradient standard deviation.

Trait Implementations

impl Clone for ModelDiagnostics

fn clone(&self) -> ModelDiagnostics

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelDiagnostics

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

Formats the value using the given formatter.