kmedoids
kmedoids<
T>(n,k,ops,options?):KmeansCluster<T>[]
Defined in: kmedoids.ts:28
K-medoids clustering.
Like k-means, but the cluster representative is always an actual data point
(the medoid): the point that minimises the sum of centroidDistance to all
other cluster members. This guarantees the representative is a real observed
value, not a computed average — useful when the mean is not meaningful (e.g.
discrete labels, non-Euclidean spaces).
Uses the same k-means++ seeding and assignment loop as kmeans.
Type Parameters
Section titled “Type Parameters”T
Parameters
Section titled “Parameters”number
Number of data points (indexed 0 … n-1).
number
Desired number of clusters; clamped to min(k, n).
KmedoidsOps<T>
Problem-specific operations. computeCentroid is handled
internally.
options?
Section titled “options?”Optional algorithm tuning.
Returns
Section titled “Returns”KmeansCluster<T>[]
One entry per non-empty cluster, with its medoid and assigned indices.