Struct WeightedRandomSampler 

pub struct WeightedRandomSampler { /* private fields */ }

Weighted random sampler for probability-based sampling.

This sampler allows you to specify different probabilities for each sample in the dataset. Samples with higher weights are more likely to be selected. This is essential for handling imbalanced datasets or implementing custom sampling distributions.

Implementation Details

The sampler uses the alias method for efficient O(1) sampling after O(n) preprocessing. This makes it suitable for large datasets where you need to draw many samples.

Performance Characteristics

• Preprocessing: O(n) time and space to build alias table
• Sampling: O(1) per sample after preprocessing
• Memory: O(n) for alias table storage
• Numerical Stability: Handles unnormalized weights robustly

Implementations

impl WeightedRandomSampler

pub fn new(weights: Vec<f32>, replacement: bool) -> Self

Create a new weighted random sampler.

Arguments

• weights - Vector of weights for each sample (will be normalized)
• replacement - Whether to sample with replacement

Panics

Panics if weights vector is empty or contains only zeros.

Examples

use torsh_data::sampler::{Sampler, WeightedRandomSampler};

let weights = vec![1.0, 2.0, 3.0]; // Unnormalized weights
let sampler = WeightedRandomSampler::new(weights, true);

// Sample probabilities will be [1/6, 2/6, 3/6]
let indices: Vec<usize> = sampler.iter().take(100).collect();
// Index 2 should appear most frequently

pub fn with_generator(self, seed: u64) -> Self

Set random generator seed.

Arguments

• seed - Seed for deterministic sampling

Examples

use torsh_data::sampler::{Sampler, WeightedRandomSampler};

let weights = vec![1.0, 2.0, 3.0];
let sampler = WeightedRandomSampler::new(weights, true)
    .with_generator(42);

// Sampling will be deterministic

pub fn weights(&self) -> &[f32]

Get the weights used by this sampler.

pub fn uses_replacement(&self) -> bool

Check if sampling is done with replacement.

pub fn generator_seed(&self) -> Option<u64>

Get the generator seed if set.

Trait Implementations

impl Clone for WeightedRandomSampler

fn clone(&self) -> WeightedRandomSampler

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for WeightedRandomSampler

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

Formats the value using the given formatter.

impl Sampler for WeightedRandomSampler

type Iter = SamplerIterator

Iterator type returned by the sampler

fn iter(&self) -> Self::Iter

Create an iterator over indices

fn len(&self) -> usize

Total number of samples that will be yielded

fn is_empty(&self) -> bool

Check if sampler is empty

fn into_batch_sampler( self, batch_size: usize, drop_last: bool, ) -> BatchingSampler<Self>

where Self: Sized,

Convert this sampler into a batch sampler

fn into_distributed( self, num_replicas: usize, rank: usize, ) -> DistributedWrapper<Self>

where Self: Sized,

Create a distributed version of this sampler