Struct ImportanceSampler 

pub struct ImportanceSampler {
    pub num_samples: usize,
    pub self_normalize: bool,
}

Importance sampling for approximate inference.

Importance sampling draws samples from a proposal distribution q(x) and weights them by p(x)/q(x) to estimate expectations under p(x).

Example

use tensorlogic_quantrs_hooks::{FactorGraph, ImportanceSampler, ProposalDistribution};

let mut graph = FactorGraph::new();
graph.add_variable_with_card("x".to_string(), "Binary".to_string(), 2);

let sampler = ImportanceSampler::new(1000);
let result = sampler.run(&graph, ProposalDistribution::Uniform);

Fields

num_samples: usize

Number of samples to draw

self_normalize: bool

Whether to use self-normalized importance sampling

Implementations

impl ImportanceSampler

pub fn new(num_samples: usize) -> Self

Create a new importance sampler with specified number of samples.

pub fn with_self_normalize(self, self_normalize: bool) -> Self

Set whether to use self-normalized importance sampling.

pub fn run( &self, graph: &FactorGraph, proposal: ProposalDistribution, ) -> Result<HashMap<String, ArrayD<f64>>>

Run importance sampling to approximate marginals.

pub fn draw_weighted_samples( &self, graph: &FactorGraph, proposal: &ProposalDistribution, ) -> Result<Vec<WeightedSample>>

Draw weighted samples from the proposal distribution.

pub fn get_weighted_samples( &self, graph: &FactorGraph, proposal: &ProposalDistribution, ) -> Result<Vec<WeightedSample>>

Get all weighted samples for analysis.

pub fn effective_sample_size(samples: &[WeightedSample]) -> f64

Compute the effective sample size (ESS).

ESS measures the efficiency of importance sampling. Higher ESS indicates better proposal distribution.

pub fn weight_coefficient_of_variation(samples: &[WeightedSample]) -> f64

Compute the coefficient of variation of weights.

pub fn resample(samples: &[WeightedSample]) -> Vec<WeightedSample>

Resample particles based on their weights (for particle filtering).

Trait Implementations

impl Default for ImportanceSampler

fn default() -> Self

Returns the “default value” for a type.