基于一致引导的不完全多视图聚类

INCOMPLETE MULTIPLE VIEW CLUSTERING BASED ON CONSISTENT GUIDANCE

为了解决传统聚类方法存在的效果差、泛化能力弱等问题,提出一种基于一致引导的不完全多视图聚类方法。将图学习和一致性表示学习集成到一个联合框架中,从而充分利用多视图数据信息。引入的自适应学习权值向量可以平衡不同视图的影响,联合正则化表示学习策略则为一致表示学习提供了更大的自由度。提出交替迭代优化算法对聚类进行优化。在七个数据集上的实验结果表明,提出的方法能够有效提升不完全多视图聚类的效果。

In order to solve the problems of poor effect and weak generalization ability of traditional clustering methods,an incomplete multiple view clustering method based on consistent guidance is proposed.Graph learning and consistent representation learning were integrated into a joint framework to make full use of multiple view data information.The adaptive learning weight vector was introduced to balance the influence of different views,and the joint regularization representation learning strategy provided more freedom for consistent representation learning.An alternative iterative optimization algorithm was proposed to optimize the clustering.Experimental results on seven data sets show that the proposed method can effectively improve the effect of incomplete multiple view clustering.