Struct ExpectedModelChange 

pub struct ExpectedModelChange { /* private fields */ }

Expected Model Change strategy.

Selects samples that would cause the largest gradient update if labeled.

Reference

• Settles, B., et al. (2008). An Analysis of Active Learning Strategies for Sequence Labeling Tasks. EMNLP.

Implementations

impl ExpectedModelChange

pub fn new() -> Self

Create a new expected model change selector.

pub fn with_min_grad(min_grad_norm: f32) -> Self

Create with minimum gradient norm threshold.

pub fn score( &self, predictions: &[Vector<f32>], gradient_norms: Option<&[f32]>, ) -> Vec<f32>

Score samples by expected gradient magnitude.

Uses prediction uncertainty as proxy for gradient magnitude.

pub fn select(&self, predictions: &[Vector<f32>], k: usize) -> Vec<usize>

Select top-k samples by expected model change.

Trait Implementations

impl Clone for ExpectedModelChange

fn clone(&self) -> ExpectedModelChange

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ExpectedModelChange

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

Formats the value using the given formatter.

impl Default for ExpectedModelChange

fn default() -> Self

Returns the “default value” for a type.