本文提出一种新的K Winners Take All神经网络 :High Speed K Winners Take All(HS K WTA) .HS K WTA是以竞争学习算法为基础 .HS K WTA能够从任何一个数集中 ,识别出K个较大的数 ,或较小的数 .本文给出HS K WTA算法及算法复杂性的分析结果 .结果显示HS K WTA要比Winstrons更好 ,更容易硬件实现 ,更稳定 ,尤其所取的数集中的数较大时 . In this paper, we propose a new K Winners Take All neural network: High Speed ​​K Winners Take All (HSK WTA). HSK WTA is based on the competitive learning algorithm. HSK WTA can identify from any number set K Larger number, or smaller number.This paper gives the HSK WTA algorithm and the analysis results of the algorithm complexity.The results show that the HS K WTA is better than Winstrons, easier to implement the hardware, more stable, especially the number of Concentration of the larger number.