摘要
针对现实社会中由多种表示或视图组成的多视图数据广泛存在的问题,深度矩阵分解模型因其能够挖掘数据的层次信息而备受关注,但该模型忽略了数据的几何结构信息。为解决以上问题,本文提出基于深度图正则化矩阵分解的多视图聚类算法,通过获取每个视图的局部结构信息和全局结构信息在逐层分解中加入两个图正则化限制,保护多视图数据的几何结构信息,同时将视图的权重与特征表示矩阵进行结合获得共识表示矩阵,最大化视角间的互补性,保证数据的一致性和差异性。除此之外,本文使用迭代更新变量的方法最小化目标函数,不断优化模型并进行收敛性分析。将本文算法和多个算法在三个人脸数据集和两个图像数据集上运行,通过多项指标的对比可以看出本文提出的算法具备良好的性能表现。
In view of the extensive multi-view data composed of multiple representations or views in real world,the deep matrix factorization(DMF)model has attracted much attention because of its ability to explore the hierarchical information of data.However,it ignores geometric structure of data.In order to solve the above problem,this paper proposes a multi-view clustering algorithm based on deep matrix factorization with graph regularization,which can protect geometric structure information of data by acquiring the local and global structure information of each view and adding two graph regularization limits in the layer-by-layer decomposition.It combines the weight of views with feature representation matrix to acquire consensus representation matrix to maximize complementarity of data and ensure consistency and difference among data.In addition,this paper uses the iterative updating variables method to minimize the objective function,continuously optimize model and conduct convergence analysis.This algorithm and other multiple algorithms are run on three face benchmark datasets and two image data sets.Through the comparison of multiple indicators,it can be seen that the algorithm proposed in this paper has good performance.
作者
刘相男
丁世飞
王丽娟
LIU Xiangnan;DING Shifei;WANG Lijuan(School of Computer Science and Technology,China University of Mining and Technology,Xuzhou 221116,China;School of Information and Electrical Engineering,Xuzhou College of Industrial Technology,Xuzhou 221400,China)
出处
《智能系统学报》
CSCD
北大核心
2022年第1期158-169,共12页
CAAI Transactions on Intelligent Systems
基金
国家自然科学基金项目(61976216,61672522).
关键词
多视图聚类
深度矩阵分解
几何结构
图正则化
矩阵分解
多视图表示学习
层次结构信息
深度学习
multi-view clustering
deep matrix factorization
geometric structure
graph regularization
matrix factorization
multi-view representation learning
hierarchical structure information
deep learning
