关键词: 关键词: 自动权重机制, 互补性, 锚点机制, 子空间聚类, 多视图聚类

Abstract: Abstract: In multi view clustering problems, how to fully explore the correlation information between views while reducing the impact of redundant information on clustering performance is an urgent problem that needs to be solved. Existing methods have some limitations, some ignore the complementary information across views, some ignore the differences among views, and some do not consider the interference caused by redundant information, resulting in poor clustering performance. To address these issues, a Self-adaptive Multi-view clustering algorithm with Complementarity based on Weighted Anchors (SMCWA) was proposed. When dealing with the challenges of high-dimensional multi view data, firstly, feature concatenation was transferred to be applied in anchor mechanism to fuse the anchor graphs to utilize the complementary information between views, Secondly, to weaken the expression of redundant information, the weights of each anchor point were dynamically determined through a weighted matrix during the iteration process; Finally, to utilize the differences between views, an auto-weighted mechanism was used to adaptively assign appropriate weights to each view. The complementarity across views, the redundant information in each view, and differences among views promoted and learned from each other in multi-step iterations in a integrated framework to obtain better clustering performance. Experimental results show that the proposed algorithm improves Matthews Correlation Coefficient evaluation index (MCC) by 42% on dataset BDGP(a RGB-D Scene Understanding benchmark suite) compared to feature-Concatenated based Spectral Clustering(SC-Concat), improves MCC by 12% on dataset CCV(Columbia Consumer Video) compared to Large-scale Multi-View Subspace Clustering in linear time (LMVSC), and improves MCC by 20% on dataset Caltech101-all compared to the best result of Spectral Clustering(SC-Best) on each view, which means that it makes full use of the complementary information, the differences among the views and the redundant information to obtain better clustering performance.

Key words: Keywords: auto-weighted mechanism, complementarity, anchor mechanism, subspace clustering, multi-view clustering