Enum AdaptiveStrategy 

pub enum AdaptiveStrategy {
    HardSampling {
        intensity: f64,
    },
    EasySampling {
        intensity: f64,
    },
    Uniform,
    Uncertainty {
        temperature: f64,
    },
    InverseFrequency {
        power: f64,
    },
    GradientMagnitude {
        threshold: f64,
    },
}

Adaptive sampling strategies

These strategies define different approaches for selecting samples based on their characteristics such as difficulty, frequency, or uncertainty.

Variants

HardSampling

Focus on hard samples (high loss)

Prioritizes samples with high training loss, which are typically harder for the model to learn.

Arguments

• intensity - Controls how strongly to weight hard samples (typically 0.5-2.0)

Fields

intensity: f64

EasySampling

Focus on easy samples (low loss)

Prioritizes samples with low training loss, which can help with stable training or curriculum learning.

Arguments

• intensity - Controls how strongly to weight easy samples (typically 0.5-2.0)

Fields

intensity: f64

Uniform

Uniform sampling for exploration

Samples all data points with equal probability, providing a baseline for comparison and ensuring exploration.

Uncertainty

Uncertainty-based sampling

Focuses on samples where the model predictions are most uncertain.

Arguments

• temperature - Controls the sharpness of uncertainty weighting

Fields

temperature: f64

InverseFrequency

Frequency-based inverse sampling

Prioritizes samples that have been seen less frequently during training, helping to balance the training distribution.

Arguments

• power - Controls the strength of inverse frequency weighting

Fields

power: f64

GradientMagnitude

Gradient-based importance

Focuses on samples that produce gradients above a certain threshold, indicating they contribute significantly to learning.

Arguments

• threshold - Minimum gradient magnitude for inclusion

Fields

threshold: f64

Trait Implementations

impl Clone for AdaptiveStrategy

fn clone(&self) -> AdaptiveStrategy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for AdaptiveStrategy

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

Formats the value using the given formatter.

impl Default for AdaptiveStrategy

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for AdaptiveStrategy

fn eq(&self, other: &AdaptiveStrategy) -> 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 StructuralPartialEq for AdaptiveStrategy