Struct SinkhornSolver 

pub struct SinkhornSolver { /* private fields */ }

Log-stabilized Sinkhorn solver for entropic optimal transport

Implementations

impl SinkhornSolver

pub fn new(regularization: f64, max_iterations: usize) -> Self

Create a new Sinkhorn solver

Arguments

• regularization - Entropy regularization ε (0.01-0.1 typical)
• max_iterations - Maximum Sinkhorn iterations (100-1000 typical)

pub fn with_threshold(self, threshold: f64) -> Self

Set convergence threshold

pub fn compute_cost_matrix( source: &[Vec<f64>], target: &[Vec<f64>], ) -> Vec<Vec<f64>>

Compute the cost matrix for squared Euclidean distance Uses SIMD-friendly 4-way unrolled accumulator for better performance

pub fn solve( &self, cost_matrix: &[Vec<f64>], source_weights: &[f64], target_weights: &[f64], ) -> Result<TransportPlan>

Solve optimal transport using log-stabilized Sinkhorn

Arguments

• cost_matrix - C[i,j] = cost to move from source[i] to target[j]
• source_weights - Marginal distribution a (sum to 1)
• target_weights - Marginal distribution b (sum to 1)

pub fn distance(&self, source: &[Vec<f64>], target: &[Vec<f64>]) -> Result<f64>

Compute Sinkhorn distance (optimal transport cost) between point clouds

pub fn barycenter( &self, distributions: &[&[Vec<f64>]], weights: Option<&[f64]>, support_size: usize, dim: usize, ) -> Result<Vec<Vec<f64>>>

Compute Wasserstein barycenter of multiple distributions

Returns the barycenter (mean distribution) in transport space

Trait Implementations

impl Clone for SinkhornSolver

fn clone(&self) -> SinkhornSolver

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SinkhornSolver

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

Formats the value using the given formatter.