Struct ImportanceSampling 

pub struct ImportanceSampling {
    pub n_samples: usize,
    pub self_normalized: bool,
}

Configuration for importance sampling.

Fields

n_samples: usize

Number of samples.

self_normalized: bool

Whether to use self-normalized estimator.

Implementations

impl ImportanceSampling

pub fn new(n_samples: usize) -> Self

Create new importance sampling configuration.

pub fn self_normalized(self) -> Self

Use self-normalized importance sampling.

The self-normalized estimator divides by sum of weights: sum(w_i * f_i) / sum(w_i)

This is biased but can have lower variance and doesn’t require knowing the normalizing constant of p.

pub fn estimate<R: GpuRng, S, F, LP, LQ>( &self, state: &mut R::State, sample_q: S, f: F, log_p: LP, log_q: LQ, ) -> (f32, f32)

where S: Fn(&mut R::State) -> f32, F: Fn(f32) -> f32, LP: Fn(f32) -> f32, LQ: Fn(f32) -> f32,

Estimate E_p[f(X)] using importance sampling.

Arguments

• state - RNG state
• sample_q - Function to sample from proposal distribution q
• f - Function to estimate expectation of
• log_p - Log of target density (unnormalized OK if self_normalized)
• log_q - Log of proposal density

Returns

(estimate, effective_sample_size)

Trait Implementations

impl Clone for ImportanceSampling

fn clone(&self) -> ImportanceSampling

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ImportanceSampling

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

Formats the value using the given formatter.

impl Default for ImportanceSampling

fn default() -> Self

Returns the “default value” for a type.