kmeans

k-means clustering algorithm implementation written in Go

What It Does

k-means clustering partitions an n-dimensional data set into k clusters, where each data point belongs to the cluster with the nearest mean, serving as a prototype of the cluster.

Example

import "github.com/muesli/kmeans"

km := kmeans.New()

// set up "random" data points (float64 values between 0.0 and 1.0)
var d kmeans.Points
for x := 0; x < 255; x += 4 {
	for y := 0; y < 255; y += 4 {
		d = append(d, kmeans.Point{
			float64(x) / 255.0,
			float64(y) / 255.0,
		})
	}
}

// Partition the data points into 16 clusters
clusters, err := km.Partition(d, 16)

for _, c := range clusters {
	fmt.Printf("Centered at x: %.2f y: %.2f\n", c.Center[0]*255.0, c.Center[1]*255.0)
	fmt.Printf("Points: %+v\n\n", c.Points)
}

Complexity

If k (the amount of clusters) and d (the dimensions) are fixed, the problem can be exactly solved in time O(n^dk+1), where n is the number of entities to be clustered.

The running time of the algorithm is O(nkdi), where n is the number of d-dimensional vectors, k the number of clusters and i the number of iterations needed until convergence. On data that does have a clustering structure, the number of iterations until convergence is often small, and results only improve slightly after the first dozen iterations. The algorithm is therefore often considered to be of "linear" complexity in practice, although it is in the worst case superpolynomial when performed until convergence.

You can greatly reduce the running time by adjusting the required delta threshold. The following code executes the algorithm until less than 5% of the data points shifted their cluster assignment in the last iteration:

km, _ := kmeans.NewWithOptions(0.05, false)

Development