Enum Optimizer 

pub enum Optimizer {
    Sgd(Sgd),
    AdaGrad(AdaGrad),
}

Concrete optimizer enum wrapping all supported optimizers.

Variants

Sgd(Sgd)

Stochastic gradient descent with optional L2 regularization.

AdaGrad(AdaGrad)

AdaGrad with per-parameter squared gradient accumulation.

Implementations

impl Optimizer

pub fn sgd(feature_count: usize, config: SgdConfig) -> Result<Self, RillError>

Create an SGD optimizer for feature_count features (plus intercept).

pub fn adagrad( feature_count: usize, config: AdaGradConfig, ) -> Result<Self, RillError>

Create an AdaGrad optimizer for feature_count features (plus intercept).

pub fn param_count(&self) -> usize

The number of parameters this optimizer manages (features + intercept).

pub fn samples_seen(&self) -> u64

Number of samples the optimizer has processed.

pub fn step( &mut self, weights: &mut [f64], intercept: &mut f64, grad_weights: &[f64], grad_intercept: f64, ) -> Result<(), RillError>

Apply a gradient step to weights (length feature_count) and intercept (single value). The gradient vector passed in must have the same length as weights; the intercept gradient is passed separately.

pub fn reset(&mut self)

Reset the optimizer to its initial state.

Trait Implementations

impl Clone for Optimizer

fn clone(&self) -> Optimizer

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Optimizer

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

Formats the value using the given formatter.