clustering/

Unsupervised algorithms for grouping similar data instances.

Clusterers find structure without labels. They differ mainly in what they assume a cluster is — a centroid, a density region, or a node in a hierarchy — so the right choice depends on the shape you expect.

Algorithms

KMeansClusterer: Partition into k spherical clusters by centroid.
GaussianMixtureClusterer: Soft assignment via expectation-maximisation.
DBSCANClusterer: Density-based; finds arbitrary shapes and labels

sparse points as noise, without being told how many clusters to expect.

DensityBasedClusterer: Density estimation wrapped around a clusterer.
AgglomerativeClusterer: Bottom-up hierarchy of merges.

Notes

Every distance-based clusterer here takes its metric from distance, which is re-exported at this level. Scale features first when using Euclidean distance, or the widest-ranging column silently dominates the metric.

python

>>> from tuiml.algorithms.clustering import KMeansClusterer
>>> from tuiml.datasets import load_iris
>>> data = load_iris()
>>> labels = KMeansClusterer(n_clusters=3, random_state=0).fit_predict(data.X)
>>> len(set(labels.tolist()))
3

Packages

distance/

Modules

dbscan

Density-Based Spatial Clustering of Applications with Noise.

1 class · 0 functions

gaussian_mixture

Expectation-Maximization clustering for Gaussian Mixture Models (GMM).

1 class · 0 functions

hierarchical

Hierarchical Agglomerative Clustering (HAC) algorithm.

2 classes · 0 functions

kmeans

K-Means clustering with multiple initialization methods.

1 class · 0 functions