Struct MiniBatchKmeans 

pub struct MiniBatchKmeans<D = SquaredEuclidean>
where
    D: DistanceMetric,
{ /* private fields */ }

Mini-Batch K-means clustering (Sculley, 2010).

Maintains k centroids and updates them incrementally as new data arrives. The first batch initializes centroids via k-means++; subsequent batches use a decaying learning rate per centroid.

use clump::MiniBatchKmeans;

let mut mbk = MiniBatchKmeans::new(2).with_seed(42);

// First batch seeds centroids via k-means++
let batch1 = vec![
    vec![0.0f32, 0.0], vec![0.1, 0.1],
    vec![10.0, 10.0], vec![10.1, 10.1],
];
let labels = mbk.update_batch(&batch1).unwrap();
assert_eq!(labels.len(), 4);

// Subsequent updates refine centroids
let label = mbk.update(&[0.05, 0.05]).unwrap();
assert_eq!(mbk.n_clusters(), 2);

Implementations

impl MiniBatchKmeans

pub fn new(k: usize) -> MiniBatchKmeans

Create a new Mini-Batch K-means clusterer with default squared Euclidean distance.

impl<D> MiniBatchKmeans<D>

where D: DistanceMetric,

pub fn with_metric(k: usize, metric: D) -> MiniBatchKmeans<D>

Create a new Mini-Batch K-means clusterer with a custom distance metric.

pub fn with_seed(self, seed: u64) -> MiniBatchKmeans<D>

Set random seed for reproducibility.

impl<D> MiniBatchKmeans<D>

where D: DistanceMetric,

pub fn update(&mut self, point: &[f32]) -> Result<usize, Error>

Update the model with a single new point, returning its cluster assignment.

pub fn update_batch( &mut self, points: &(impl DataRef + ?Sized), ) -> Result<Vec<usize>, Error>

Update the model with a mini-batch of points.

pub fn predict( &self, data: &(impl DataRef + ?Sized), ) -> Result<Vec<usize>, Error>

Predict cluster labels for new points without modifying centroids.

Read-only inference: assigns each point to its nearest centroid but does not update centroid positions or counts.

pub fn centroids(&self) -> &[Vec<f32>]

Get current cluster centroids.

pub fn counts(&self) -> &[usize]

Get the per-centroid assignment count.

pub fn n_clusters(&self) -> usize

Get the current number of clusters.

Trait Implementations

impl<D> Clone for MiniBatchKmeans<D>

where D: Clone + DistanceMetric,

fn clone(&self) -> MiniBatchKmeans<D>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<D> Debug for MiniBatchKmeans<D>

where D: Debug + DistanceMetric,

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

Formats the value using the given formatter.