Struct WeightedKMeans 

pub struct WeightedKMeans { /* private fields */ }

A weighted clustering-problem where each center can be any point in the metric space.

Use KMeans instead if all your points have the same weight.

The metric space is the same space the weighted points live in.

The cost of a cluster, given a center, is the sum of the squared Euclidean distances between the center and each point in the cluster, multiplied by the point’s weight. The center is automatically calculated to minimise the cost, which turns out to simply be the weighted average of all point-positions in the cluster.

See the wikipedia-article on k-means-clustering for more information.

Implementations

impl WeightedKMeans

pub fn new(weighted_points: &[WeightedPoint]) -> Result<Self, Error>

Construct a new k-means clustering instance.

The algorithm runs significantly faster if you sort the points by weight first, merge points that have the same positions into one (adding their weights), and break symmetries by applying a small amount of noise.

TODO: Do so internally instead of at callsite. This probably requires better return-values in the API. Afterwards, rework the get_high_kmeans_price_of_greedy_instance function in integration-tests to not sort the points on its own.

Examples

use ndarray::array;
use exact_clustering::WeightedKMeans;

WeightedKMeans::new(&[(1.0, array![0.0, 0.0]), (2.0, array![1.0, 2.0])]).unwrap();

Trait Implementations

impl Clone for WeightedKMeans

fn clone(&self) -> WeightedKMeans

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Cost for WeightedKMeans

fn num_points(&self) -> usize

Get the number of points that must be clustered.

fn cost(&mut self, cluster: Cluster) -> f64

Get the cost of a cluster.

fn total_cost(&mut self, clustering: &Clustering) -> f64

Get the total cost of a clustering.

fn approximate_clusterings(&mut self) -> Vec<(f64, Clustering)>

Quickly calculate a not-necessarily-optimal clustering.

fn optimal_clusterings(&mut self) -> Vec<(f64, Clustering)>

Calculate an optimal k-clustering for every 0 ≤ k ≤ self.num_points().

fn price_of_hierarchy(&mut self) -> (f64, Vec<Clustering>)

Calculate the price-of-hierarchy of the clustering-problem, together with an optimal hierarchy.

fn greedy_hierarchy(&mut self) -> Vec<(f64, Clustering)>

Calculate a greedy hierarchical clustering.

fn price_of_greedy(&mut self) -> (f64, Vec<Clustering>)

Calculate the cost-ratio of a greedy hierarchical clustering.

impl Debug for WeightedKMeans

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

Formats the value using the given formatter.