Enhanced Latent Multi-view Subspace Clustering

• DOI: 10.48550/arxiv.2312.14763
• Publication Date: 2023-01-01
• Authors: Long Shi; Lei Cao; Jun Wang; Badong Chen
• Keywords: Block matrix; Representation; Biclustering

Latent multi-view subspace clustering has been demonstrated to have desirable performance. However, the original latent representation method vertically concatenates data matrices from multiple views into a single matrix along direction of dimensionality recover matrix, which may result in an incomplete information recovery. To fully space representation, we this paper propose Enhanced Multi-view Subspace Clustering (ELMSC) method. The ELMSC involves constructing augmented that enhances data. Specifically, stack various block-diagonal locations exploit complementary information. Meanwhile, non-block-diagonal entries are composed based on similarity between different capture consistent In addition, enforce sparse regularization for non-diagonal blocks self-representation avoid redundant calculations consistency Finally, novel iterative algorithm framework Alternating Direction Method Multipliers (ADMM) is developed solve optimization problem ELMSC. Extensive experiments real-world datasets demonstrate our proposed able achieve higher performance than some state-of-art methods.