Struct LossTracker 

pub struct LossTracker {
    pub window_size: usize,
    /* private fields */
}

Rolling-window tracker for scalar loss values recorded during training.

Provides moving average, min/max, and a simple improvement check useful for early stopping decisions.

Fields

window_size: usize

Implementations

impl LossTracker

pub fn new(window_size: usize) -> Self

Create a new tracker with the given sliding window capacity.

pub fn push(&mut self, loss: f64)

Record a new loss value, evicting the oldest if the window is full.

pub fn moving_average(&self) -> Option<f64>

Arithmetic mean over the current window; None if empty.

pub fn min(&self) -> Option<f64>

Minimum value in the current window; None if empty.

pub fn max(&self) -> Option<f64>

Maximum value in the current window; None if empty.

pub fn is_improving(&self, patience: usize) -> bool

Returns true when the minimum loss seen in the most-recent patience values is strictly less than the minimum over the full window excluding those recent values. This captures “the model has improved recently”.

Returns false when there are not enough data points to compare.

pub fn len(&self) -> usize

Number of values currently held.

pub fn is_empty(&self) -> bool

true when no values have been recorded yet.

Trait Implementations

impl Clone for LossTracker

fn clone(&self) -> LossTracker

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LossTracker

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

Formats the value using the given formatter.