Python Forum

I read some online materials on how K-means++ works in selecting initial centroids in KMeans. But I still don't understand.

Eg, if I specify 3 clusters,

1) K-Means++ will select the 1st centroid point (C1) randomly
2) Compute the Euclidean distance of all points from C1.
3) Make the point with the largest distance C2? Then how is C3 found?

Thank you

Wikipedia gives a slightly different description of the kmeans++ algorithm with a way to select successive centroids.