17
4

Exact Acceleration of K-Means++ and K-Means\|

Abstract

K-Means++ and its distributed variant K-Means\| have become de facto tools for selecting the initial seeds of K-means. While alternatives have been developed, the effectiveness, ease of implementation, and theoretical grounding of the K-means++ and \| methods have made them difficult to "best" from a holistic perspective. By considering the limited opportunities within seed selection to perform pruning, we develop specialized triangle inequality pruning strategies and a dynamic priority queue to show the first acceleration of K-Means++ and K-Means\| that is faster in run-time while being algorithmicly equivalent. For both algorithms we are able to reduce distance computations by over 500×500\times. For K-means++ this results in up to a 17×\times speedup in run-time and a 551×551\times speedup for K-means\|. We achieve this with simple, but carefully chosen, modifications to known techniques which makes it easy to integrate our approach into existing implementations of these algorithms.

View on arXiv
Comments on this paper