Struct SelfPacedCurriculum 

pub struct SelfPacedCurriculum { /* private fields */ }

Self-paced curriculum learning

Reference: [Kumar et al. 2010] “Self-Paced Learning for Latent Variable Models”

• Automatically determines sample difficulty from loss
• Adapts pace based on model performance

Implementations

impl SelfPacedCurriculum

pub fn new(initial_threshold: f64, growth_rate: f64) -> SelfPacedCurriculum

Create a new self-paced curriculum

Arguments

• initial_threshold - Starting difficulty threshold (e.g., 0.1)
• growth_rate - How much to increase threshold per advance (e.g., 1.5)

pub fn with_max_threshold(self, max: f64) -> SelfPacedCurriculum

Set maximum threshold

pub fn add_samples(&mut self, samples: Vec<ScoredSample>)

Add samples with difficulty scores

pub fn update_difficulties<F>(&mut self, loss_fn: F)

where F: Fn(&[f64], f64) -> f64,

Update difficulties based on model predictions

Arguments

• loss_fn - Function that computes loss for a sample

pub fn next_batch(&mut self, batch_size: usize) -> Vec<&ScoredSample>

Get next batch of samples below current threshold

pub fn eligible_samples(&self) -> Vec<&ScoredSample>

Get all samples below current threshold

pub fn n_eligible(&self) -> usize

Get number of eligible samples

pub fn n_total(&self) -> usize

Get total number of samples

pub fn threshold(&self) -> f64

Get current threshold

Trait Implementations

impl Clone for SelfPacedCurriculum

fn clone(&self) -> SelfPacedCurriculum

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl CurriculumScheduler for SelfPacedCurriculum

fn stage(&self) -> f64

Get current stage (0.0 = easiest only, 1.0 = all data)

fn advance(&mut self)

Advance to next stage

fn is_complete(&self) -> bool

Check if curriculum is complete

fn current_threshold(&self) -> f64

Get the difficulty threshold for current stage

fn reset(&mut self)

Reset to beginning

impl Debug for SelfPacedCurriculum

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

Formats the value using the given formatter.