Struct SemanticClusterer 

pub struct SemanticClusterer {
    pub algorithm: ClusterAlgorithm,
    pub dims: usize,
}

Multi-algorithm semantic vector clustering engine.

use ipfrs_semantic::semantic_clusterer::{
    SemanticClusterer, ClusterAlgorithm, ScClusterPoint,
};

let points: Vec<ScClusterPoint> = (0..20)
    .map(|i| ScClusterPoint::new(format!("p{i}"), vec![(i % 4) as f64, (i / 4) as f64]))
    .collect();

let clusterer = SemanticClusterer::new(
    ClusterAlgorithm::KMeans { k: 4, max_iter: 100, tolerance: 1e-6 },
    2,
);
let result = clusterer.fit(&points).expect("clustering failed");
assert_eq!(result.clusters.len(), 4);

Fields

algorithm: ClusterAlgorithm

Algorithm configuration.

dims: usize

Expected embedding dimensionality.

Implementations

impl SemanticClusterer

pub fn new(algorithm: ClusterAlgorithm, dims: usize) -> Self

Create a new clusterer with the given algorithm and embedding dimension.

pub fn fit( &self, points: &[ScClusterPoint], ) -> Result<ScClusteringResult, ClusterError>

Run clustering on the provided points.

Returns a ScClusteringResult on success.

pub fn predict( &self, point: &ScClusterPoint, result: &ScClusteringResult, ) -> Option<usize>

Assign a new point to the nearest cluster centroid in result.

Returns None when result has no clusters.

pub fn cosine_distance(a: &[f64], b: &[f64]) -> f64

Cosine distance: 1 - cosine_similarity(a, b).

Returns 1.0 for zero vectors.

pub fn euclidean_distance(a: &[f64], b: &[f64]) -> f64

Euclidean distance between two vectors of equal length.

pub fn silhouette_score( points: &[ScClusterPoint], result: &ScClusteringResult, ) -> f64

Mean silhouette coefficient for the clustering.

Uses a random subset of up to 100 points when the dataset is large.

pub fn compute_centroid(embeddings: &[&[f64]]) -> Vec<f64>

Compute the centroid (element-wise mean) of a set of embeddings.

Returns a zero vector when embeddings is empty or their slices are empty.

pub fn stats(result: &ScClusteringResult) -> ScClustererStats

Summary statistics for a completed clustering result.

Trait Implementations

impl Clone for SemanticClusterer

fn clone(&self) -> SemanticClusterer

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SemanticClusterer

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

Formats the value using the given formatter.