Enum Loss 

pub enum Loss {
    Mse,
    Mae,
    BinaryCrossEntropyWithLogits,
    SoftmaxCrossEntropy,
}

Supported loss functions.

Variants

Mse

Mean squared error.

Mae

Mean absolute error.

BinaryCrossEntropyWithLogits

Binary cross-entropy with logits.

This expects raw logits as predictions and targets in [0, 1]. In most cases you should use an Identity activation on the output layer.

SoftmaxCrossEntropy

Softmax cross-entropy.

This expects raw logits as predictions and a one-hot target vector. In most cases you should use an Identity activation on the output layer.

Implementations

impl Loss

pub fn validate(self) -> Result<()>

Validate a loss configuration.

pub fn forward(self, pred: &[f32], target: &[f32]) -> f32

Compute a loss value.

Shape contract: pred.len() == target.len().

pub fn backward(self, pred: &[f32], target: &[f32], d_pred: &mut [f32]) -> f32

Compute loss + gradient w.r.t pred.

Writes d_pred = dL/d(pred) into d_pred and returns the loss.

Shape contract:

• pred.len() == target.len()
• pred.len() == d_pred.len()

Trait Implementations

impl Clone for Loss

fn clone(&self) -> Loss

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Loss

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

Formats the value using the given formatter.

impl PartialEq for Loss

fn eq(&self, other: &Loss) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for Loss

impl Eq for Loss

impl StructuralPartialEq for Loss