基于成对约束的半监督聚类方法

Semi-supervised clustering based on paired constraints

成对约束作为半监督学习的一个重要分支,它以无监督的聚类算法为基础,通过利用少量的监督信息来提高聚类的性能。提出了一种基于成对约束的半监督聚类方法,在FCM-NMF聚类算法框架下,采用非负矩阵分解提取样本的本质特征,并且加入成对约束条件指导聚类过程进行模糊聚类。构造出的新的目标函数采用梯度下降法和交替迭代公式来找局部最优解,并在UCI数据集上讨论了成对约束的数量对聚类的影响、价值系数的设置对聚类的影响,并与常见的半监督聚类性能进行了比较,证明了其正确性、有效性、稳定性。

As an important branch of semi-supervised learning,paired constraint is based on unsupervised clustering algorithm,which improves the performance of clustering by using a small amount of supervised information.In this paper,a new semi-supervised clustering method based on paired constraints is proposed.Under the framework of FCM-NMF algorithm,the essential features of samples are extracted by non-negative matrix decomposition,and paired constraints are added to guide the clustering process for fuzzy clustering.It constructs the new objective function using the gradient descent method and the alternate iteration formula to find local optimal solution,and discusses on UCI data sets the number of constraints in pairs of clustering coefficient,the influence of value setting influence on clustering.And compared with common semi-supervised clustering performance,the correctness,validity and stability of this methed are proved.