Struct DropoutLayer 

pub struct DropoutLayer {
    pub rate: f64,
    pub training: bool,
    pub last_mask: Vec<f64>,
    /* private fields */
}

Dropout regularisation layer.

During training, each neuron is set to zero with probability rate. During inference (training=false) the layer passes inputs through unchanged but scales by 1 - rate to maintain expected magnitude.

Fields

rate: f64

Probability of dropping a unit (0.0 = no dropout, 1.0 = all dropped).

training: bool

Whether the layer is in training mode.

last_mask: Vec<f64>

Mask used in the last forward pass (1.0 = kept, 0.0 = dropped).

Implementations

impl DropoutLayer

pub fn new(rate: f64, training: bool) -> Self

Create a new dropout layer.

pub fn set_seed(&mut self, seed: u64)

Set the seed for reproducible mask generation.

pub fn forward(&mut self, input: &[f64]) -> Vec<f64>

Apply dropout to input.

In training mode, randomly zeros elements with probability rate and scales the rest by 1 / (1 - rate) (inverted dropout). In eval mode, passes through unchanged.

pub fn backward(&self, delta_out: &[f64]) -> Vec<f64>

Backward pass: applies the stored mask to the upstream gradient.

Trait Implementations

impl Clone for DropoutLayer

fn clone(&self) -> DropoutLayer

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DropoutLayer

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

Formats the value using the given formatter.