Simple and sharp analysis of k-means||

03/05/2020

∙

by   Václav Rozhoň, et al.

∙

0

∙

share

We present a truly simple analysis of k-means|| (Bahmani et al., PVLDB 2012) – a distributed variant of the k-means++ algorithm (Arthur and Vassilvitskii, SODA 2007) – and improve it from O(logVar X), where Var X is the variance of the input data set, to O(logVar X / loglogVar X), which we show to be tight.