基于一致低维嵌入的快速多视图聚类算法OA

Anchor-based multi-view clustering methods have attracted significant attention due to their ability to effectively reduce time complexity.However,as the data scale expands,the growing number of anchor points significantly increases the dimensionality of the anchor graph,resulting in substantial computational costs for subsequent clustering strategies.Moreover,most current methods treat multi-view learning and clustering as separate tasks,which prevents the joint optimization of multi-view information and clustering results,thereby limiting clustering performance.To address these challenges,a fast multi-view clustering algorithm based on consistent low-dimensional embedding(CLE-FMC)is proposed,which integrates multi-view learning and clustering into a unified framework.Specifically,the anchor graph is first decomposed into a transition matrix and a low-dimensional embedding matrix.With the introduction of the transition matrix,subsequent clustering is primarily performed on the low-dimensional representation,which alleviates,to some extent,the additional computational overhead incurred by the increase in the number of anchor points.Subsequently,a one-step clustering approach is proposed,which directly leverages the low-dimensional embedding matrix to obtain cluster labels.Finally,this unified framework enables the joint optimization of multi-view learning and clustering,while enhancing the algorithm's computational efficiency.Extensive experiments demonstrate the superior clustering performance of the proposed algorithm.