$t$-$k$-means: A Robust and Stable $k$-means Variant

  • 2020-09-24 16:05:12
  • Yiming Li, Yang Zhang, Qingtao Tang, Weipeng Huang, Yong Jiang, Shu-Tao Xia
  • 0

Abstract

$k$-means algorithm is one of the most classical clustering methods, whichhas been widely and successfully used in signal processing. However, due to thethin-tailed property of the Gaussian distribution, $k$-means algorithm suffersfrom relatively poor performance on the dataset containing heavy-tailed data oroutliers. Besides, standard $k$-means algorithm also has relatively weakstability, $i.e.$ its results have a large variance, which reduces the modelcredibility. In this paper, we propose a robust and stable $k$-means variant,dubbed the $t$-$k$-means, as well as its fast version to alleviate thoseproblems. Theoretically, we derive the $t$-$k$-means and analyze its robustnessand stability from the aspect of the loss function and the expression of theclustering center, respectively. A large number of experiments are alsoconducted, which verify the effectiveness and efficiency of the proposedmethod.

 

Quick Read (beta)

loading the full paper ...