Skip to content



A fast Plus Plus approximator that replaces the brute force method with a substantially faster Markov Chain Monte Carlo (MCMC) sampling procedure with comparable results.


# Name Default Type Description
1 m 50 int The number of candidate nodes in the Markov Chain.
2 kernel Euclidean Distance The distance kernel used to compute the distance between samples.


use Rubix\ML\Clusterers\Seeders\KMC2;
use Rubix\ML\Kernels\Distance\Euclidean;

$seeder = new KMC2(200, new Euclidean());


  1. O. Bachem et al. (2016). Approximate K-Means++ in Sublinear Time.