Some Pairwise Constrained Semi-Supervised Fuzzy c-Means Clustering Algorithms
Some Pairwise Constrained Semi-Supervised Fuzzy c-Means Clustering Algorithms
复制标题
DOI：
10.1007/978-3-642-04820-3_25
复制
发表时间：
2009-11
期刊：
影响因子：
--
通讯作者：
Y. Kanzawa;Y. Endo;S. Miyamoto
中科院分区：
文献类型：
--
作者：
Y. Kanzawa;Y. Endo;S. Miyamoto
本文提出了两类约束条件下的半监督聚类方法：两个数据必须在同一个类中，两个数据必须在不同的类中。这两类约束条件分为两类：一类是基于标准模糊均值算法，另一类是基于熵正则化算法。首先介绍了标准模糊均值和熵正则化模糊均值。其次，引入了一种成对约束半监督模糊均值算法，该算法是由成对约束竞争聚集算法衍生而来的。第三，提出了一些新的优化问题，这些问题是在原优化问题的基础上增加新的隶属度损失函数而得到的。最后，通过求解优化问题，提出了一种迭代算法。
In this paper, some semi-supervised clustering methods are proposed with two types of pair constraints: two data have to be together in the same cluster, and two data have to be in different clusters, which are classified into two types: one is based on the standard fuzzyc-means algorithm and the other is on the entropy regularized one. First, the standard fuzzyc-means and the entropy regularized one are introduced. Second, a pairwise constrained semi-supervised fuzzycmeans are introduced, which is derived from pairwise constrained competitive agglomeration. Third, some new optimization problem are proposed, which are derived from adding new loss function of memberships to the original optimization problem, respectively. Last, an iterative algorithm is proposed by solving the optimization problem.