David Arthur & Sergei Vassilvitskii (2007)
Proceedings of SODA 2007, 1027-1035.
URL: https://theory.stanford.edu/~sergei/papers/kMeansPP-soda.pdf
Abstract. Proposes the k-means++ initialisation, which selects initial centroids with probability proportional to squared distance from the nearest existing centroid, yielding an O(log k)-competitive approximation to the optimal clustering.
Tags: clustering k-means unsupervised url-only