Andrew Ng chapter14 无监督学习------K Means

K均值算法,是一个迭代算法,迭代过程有两个:簇分配,以及聚类中心的移动。
Andrew Ng chapter14 无监督学习------K Means
Andrew Ng chapter14 无监督学习------K Means
输入:K的值以及数据
Andrew Ng chapter14 无监督学习------K Means
Andrew Ng chapter14 无监督学习------K Means

优化目标

Andrew Ng chapter14 无监督学习------K Means
失真代价值

随机初始化

Andrew Ng chapter14 无监督学习------K Means
Andrew Ng chapter14 无监督学习------K Means
Andrew Ng chapter14 无监督学习------K Means
为了避免结果是局部最小值,可以多次随机初始化K,但是只是当m的数量不太多时效果会比较好。

K值的选取

Andrew Ng chapter14 无监督学习------K Means
Andrew Ng chapter14 无监督学习------K Means
肘部法则或者是根据需求来选择