Scalable K-Means by ranked retrieval

   Abstract