Struct wyrm::SGD

pub struct SGD { /* fields omitted */ }

Standard stochtic gradient descent optimizer with a fixed learning rate.

Methods

impl SGD

fn new(learning_rate: f32, parameters: Vec<Variable<ParameterNode>>) -> Self

Create a new optimizer instance with a given set of parameters.

fn zero_gradients(&self)

Zero the gradients on each of the parameters being optimized.

fn step(&mut self)

Perform a single SGD step.