Struct Dropout 

pub struct Dropout { /* private fields */ }

Dropout regularization layer.

During training, randomly zeroes elements with probability p. During evaluation, returns input unchanged.

The output is scaled by 1/(1-p) during training to maintain expected values (inverted dropout).

Example

use aprender::nn::{Module, Dropout};
use aprender::autograd::Tensor;

let mut dropout = Dropout::new(0.5);
let x = Tensor::ones(&[10, 10]);

dropout.train();
let y_train = dropout.forward(&x);  // ~50% zeros, rest scaled by 2

dropout.eval();
let y_eval = dropout.forward(&x);   // same as input

Implementations

impl Dropout

pub fn new(p: f32) -> Self

Create a new Dropout layer.

Arguments

• p - Probability of element being zeroed (must be in [0, 1))

Panics

Panics if p is not in [0, 1).

pub fn with_seed(p: f32, seed: u64) -> Self

Create a new Dropout layer with a specific seed for reproducibility.

pub fn probability(&self) -> f32

Get the dropout probability.

Trait Implementations

impl Debug for Dropout

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

Formats the value using the given formatter.

impl Module for Dropout

fn forward(&self, input: &Tensor) -> Tensor

Forward pass with dropout.

Panics

Panics if the RNG mutex is poisoned (indicates prior thread panic - unrecoverable). This is acceptable per Toyota Way: mutex poisoning means the system is already in a catastrophic state from a prior panic.

fn train(&mut self)

Set the module to training mode.

fn eval(&mut self)

Set the module to evaluation mode.

fn training(&self) -> bool

Check if the module is in training mode.

fn parameters(&self) -> Vec<&Tensor>

Get references to all learnable parameters.

fn parameters_mut(&mut self) -> Vec<&mut Tensor>

Get mutable references to all learnable parameters.

fn refresh_caches(&mut self)

Refresh any cached computations after parameters have been modified.

fn zero_grad(&mut self)

Zero out gradients for all parameters.

fn num_parameters(&self) -> usize

Get the number of learnable parameters.