中文说明：k-均值是非常敏感的初始化。一个错误的初始化可以延迟收敛或甚至错误的聚类结果。均值漂移是相当强大的初始化。通常情况下，均值漂移是运行的每一点或点的特征空间中的[ 2 ]的一致选择。同样，k-均值是敏感的异常值，但均值漂移是不是很敏感。
English Description：
K-means is very sensitiveto initializations. A wrong initialization can delay convergence or sometimeseven result in wrong clusters. Mean shift is fairly robust to initializations.Typically, mean shift is run for each point or sometimes points are selected uniformly from the feature space [2] . Similarly, K-means is sensitive to outliers but Mean Shift is not very sensitive.