High-Order Anchor Graph Tensor Representation Learning for Enhanced Multi-View Clustering

Multi-view clustering has become a prominent method in diverse fields due to its effectiveness in data analysis. Graph-based multi-view clustering is particularly advantageous as it captures complex interrelationships within data through graph structures. However, existing approaches in this area often face performance issues related to graph sparsity, do not effectively utilize spatial consistency information across different views, and are hindered by high computational complexity. To address these limitations, this paper introduces a structured multi-view graph learning method utilizing tensorized high-order anchor graphs. By incorporating high-order anchor graphs, the proposed approach provides a denser representation of proximity relationships, mitigating the sparsity issues of the initial input graph. Furthermore, the method employs a tensor representation to amalgamate multi-view high-order anchor graphs, leveraging the tensor nuclear norm constraint to fully capture and integrate complementary and consistent information across views. Additionally, the approach utilizes a Laplacian rank constraint to adaptively learn a structured anchor graph proximity matrix. Experimental results indicate that this method offers superior performance and holds significant potential for advancing multi-view clustering techniques.