Multi-view Low-rank Preserving Embedding: A novel method for multi-view representation

In recent years, we have witnessed a surge of interest in multi-view representation learning. When facing multiple views that are highly related but sightly different from each other, most existing methods might fail to fully explore information. Additionally, pairwise correlations among often vary drastically, which makes challenging. Therefore, how learn appropriate information is still an open challenging problem. To handle this issue, paper proposes novel learning method, named Multi-view Low-rank Preserving Embedding (MvLPE). It integrates all into common latent space, termed as centroid view, by minimizing the disagreement between view and encourages mutually other. Unlike with explicit weight definition, proposed method could automatically allocate ideal for according its contribution. Besides, MvLPE maintain low-rank reconstruction structure while integrating view. Since there no closed-form solution MvLPE, effective algorithm based on iterative alternating strategy provided obtain solution. The experiments six benchmark datasets validate effectiveness achieves superior performance over counterparts.