Struct SGD 

pub struct SGD<T: Float> { /* private fields */ }

Stochastic Gradient Descent(SGD)

Formula:

W^{t+1}_{i} = W^t_{i+1} - \eta \cdot \nabla L(W_{t})

where:

- W^{t+1}_{i} — updated weights,
- W^t_{i} — current weights,
- \eta — learning rate,
- \nabla L(W_{t}) — gradient of the loss function with respect to the weights.

Implementations

impl<T: Float> SGD<T>

pub fn new(learning_rate: T) -> Self

Trait Implementations

impl<T: Float> Optimizer<T> for SGD<T>

fn step(&mut self, _id: usize, weights: &mut Matrix<T>, gradients: &Matrix<T>)

fn change_learning_rate(&mut self, new_learning_rate: T)