Struct SquaredLoss 

pub struct SquaredLoss;

Mean squared error loss (L2 loss, scaled by 0.5).

Optimal for regression when the noise is Gaussian. Sensitive to outliers – consider HuberLoss for heavy-tailed distributions.

Trait Implementations

impl Clone for SquaredLoss

fn clone(&self) -> SquaredLoss

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SquaredLoss

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

Formats the value using the given formatter.

impl Loss for SquaredLoss

fn n_outputs(&self) -> usize

Number of output dimensions (1 for regression/binary, C for multiclass).

fn gradient(&self, target: f64, prediction: f64) -> f64

First derivative of the loss with respect to prediction: dL/df.

fn hessian(&self, _target: f64, _prediction: f64) -> f64

Second derivative of the loss with respect to prediction: d^2L/df^2.

fn loss(&self, target: f64, prediction: f64) -> f64

Raw loss value L(target, prediction).

fn predict_transform(&self, raw: f64) -> f64

Transform raw model output to final prediction (identity for regression, sigmoid for binary).

fn initial_prediction(&self, targets: &[f64]) -> f64

Initial constant prediction (before any trees). Typically the optimal constant that minimizes sum of losses over the given targets.

fn loss_type(&self) -> Option<LossType>

Return the serialization tag for this loss function.

impl Copy for SquaredLoss