A Faster -means++ Algorithm

Abstract
-means++ is an important algorithm for choosing initial cluster centers for the -means clustering algorithm. In this work, we present a new algorithm that can solve the -means++ problem with nearly optimal running time. Given data points in , the current state-of-the-art algorithm runs in iterations, and each iteration takes time. The overall running time is thus . We propose a new algorithm \textsc{FastKmeans++} that only takes in time, in total.
View on arXivComments on this paper