Struct Sgd 

pub struct Sgd { /* private fields */ }

SGD optimizer with optional L2 regularization.

The update rule for each weight w_i is:

w_i -= lr * (grad_i + l2 * w_i)

The intercept is not regularized.

Implementations

impl Sgd

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

Create a new SGD optimizer.

pub const fn learning_rate(&self) -> f64

The configured learning rate.

pub const fn l2(&self) -> f64

The configured L2 regularization.

pub const fn param_count(&self) -> usize

Number of parameters (features + intercept).

pub const fn samples_seen(&self) -> u64

Number of samples processed.

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

Apply one gradient step.

pub fn reset(&mut self)

Reset the sample counter.

Trait Implementations

impl Clone for Sgd

fn clone(&self) -> Sgd

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Sgd

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

Formats the value using the given formatter.