Struct SGD 

pub struct SGD { /* private fields */ }

Stochastic Gradient Descent optimizer.

Implementations

impl SGD

pub fn new(learningrate: f64, momentum: Option<f64>) -> Self

Create a new SGD optimizer.

pub fn set_learningrate(&mut self, learningrate: f64)

Set the learning rate.

Trait Implementations

impl Optimizer for SGD

fn step(&mut self) -> CoreResult<()>

Step the optimizer to update variables.

fn zero_grad(&mut self)

Zero all gradients.

fn add_variable(&mut self, var: Variable)

Add a variable to the optimizer.

fn variables(&self) -> &[Variable]

Get all variables managed by the optimizer.

fn variables_mut(&mut self) -> &mut [Variable]

Get mutable reference to variables (for default implementation)

fn accumulate_gradients(&mut self, gradients: &GradientDict) -> CoreResult<()>

Accumulate gradients for momentum-based optimizers