【摘要】 模糊C均值(FCM)算法是模式识别领域应用最广的聚类算法之一.但是FCM算法存在很多缺点,其中以对噪声数据敏感,鲁棒性较差最为突出.针对这种情况,Lee于1994年提出了一种所谓的改进模糊C均值算法＿Lee’s算法.但是本文证明了Lee’s算法并不是一种真正意义上的模糊C均值改进算法,而是Krishnapuram和Keller于1993年所提出的PCM算法的一种特殊情况.数值实验进一步证明了我们的结论.这对合理地使用模糊聚类算法提供了一定的理论依据.更多还原

【Abstract】 Fuzzy C-means clustering algorithm is one of the most widely used algorithms in pattern recognition. However, FCM algorithm has a lot of drawbacks. Among these drawbacks, being sensitive to the noise is the most outstanding. In order to overcome this drawback, Lee proposed a modified FCM algorithm-Lee’s algorithm. In this paper we will prove that Lee’s model isn’t a kind of FCM (algorithm,) but a special case of PCM which was proposed by Krishnapuram & Keller in 1993. Moreover, the numerical experiments demonstrate our conclusion.更多还原