美文网首页
Clustering

Clustering

作者: shawn233 | 来源:发表于2018-06-10 10:22 被阅读0次

We perform clustering because we believe the underlying cluster labels are meaningful, will lead to a more efficient description of our data, and will help us choose better actions.

K-means Clustering

The K-means algorithm is an algorithm for putting N data points in an I-dimensional space into K clusters. Each cluster is parameterized by a vector m(k) called its mean.

Each data point is denoted by x(n), which is consisted of I components.

Distance between data points are defined, such as:

Simply two steps

  • assignment step: Each data point x(n) is assigned to the nearest mean.
  • update step: The means are adjusted to match the sample means of the data points that they are responsible for.

After iterations of the two steps, this algorithm will definitely converge. (This can be proved.) The convergence is indicated by the means remaining unmoved when updated.

Withdraw about K-means: It is a 'hard' algorithm. 'Hard' means that it assigns each data point to exactly one cluster, and all data points in a cluster are equal in updating the mean. Maybe points on the borderline of two or more clusters should have less vote in updating step.

Soft K-means clustering

The bad things about 'hard' K-means algorithm gives rise to the soft K-means algorithm.

This algorithm still has some flaws. It is hopefully improved by using maximum-likelihood.

相关文章

  • 2019-01-12[Stay Sharp]hierarchic

    what is hierarchical clustering ? hierarchical clustering...

  • ML - hw4

    1. Spectral Clustering (a) Spectral Clustering on synthes...

  • 100天持续行动—Day13

    10.26看了K-Means clustering, Hierarchical clustering 和 DBSC...

  • 2019-01-10[Stay Sharp]k-means cl

    what is k-means clustering? K-means clustering is a metho...

  • 漫谈 Clustering

    漫谈 Clustering (1): k-means 漫谈 Clustering (2): k-medoids 漫...

  • Clustering

    We perform clustering because we believe the underlying c...

  • Clustering

    本文结构安排 经典聚类算法:线性聚类 Kmeans 经典聚类算法:非线性聚类 DBSCAN、谱聚类 新兴聚类算法:...

  • Clustering

    Single linkage clustering 类似层次聚类的概念,时间复杂度是O(n^3)。 SLC聚类效果...

  • Clustering

    Clustering 算法概念 算法类型 K-means算法是非监督学习聚类(clustering)中的经典算法,...

  • clustering

    运行结果:运行结果

网友评论

      本文标题:Clustering

      本文链接:https://www.haomeiwen.com/subject/cqlfeftx.html