This module implements a K-medoids algorithm using as input a list of points
and a distance function, like distance(p, q) = ||q - p||:
>>> points = [1, 2, 3, 4, 5, 6, 7]
>>> def distance(a, b):
... return abs(b - a)Just import the main k_medoids function:
>>> from medoids import k_medoids
>>> diameter, medoids = k_medoids(points, k=2, distance, spawn=2) #doctest: +SKIP
* New chosen kernels: [2, 3]
* Iteration over after 3 steps, max diameter 3
* New chosen kernels: [1, 2]
* Iteration over after 4 steps, max diameter 3
~~ Spawn end: min of max diameters 3.000 for medoids: [Medoid(2, [1, 2, 3]), Medoid(5, [4, 5, 6, 7])]There is also a k_medoids_auto_k which increases automatically the number of clusters
until we have homogeneous clusters:
>>> from medoids import k_medoids_auto_k
>>> diameter, medoids = k_medoids_auto_k(points, diam_max=3, distance, spawn=3) #doctest: +SKIP
* New chosen kernels: [2]
* Iteration over after 2 steps, max diameter 6
* New chosen kernels: [6]
* Iteration over after 2 steps, max diameter 6
* New chosen kernels: [2]
* Iteration over after 2 steps, max diameter 6
~~ Spawn end: min of max diameters 6.000 for medoids: [Medoid(4, [1, 2, 3, 4, 5, 6, 7])]
*** Diameter too big 6.000 > 3.000
*** Now trying 2 clusters
* New chosen kernels: [6, 2]
* Iteration over after 2 steps, max diameter 3
* New chosen kernels: [2, 6]
* Iteration over after 1 steps, max diameter 3
* New chosen kernels: [2, 1]
* Iteration over after 3 steps, max diameter 4
~~ Spawn end: min of max diameters 3.000 for medoids: [Medoid(5, [4, 5, 6, 7]), Medoid(2, [1, 2, 3])]
*** Diameter ok 3.000 <= 3.000
*** Stopping, 2 clusters enough (7 points initially)