Higher Order Correlation Analysis for Multi-View Learning
This addresses the limitation of pairwise correlation methods in multi-view learning for data science applications, though it appears incremental as it builds on existing correlation frameworks.
The paper tackles the problem of capturing intrinsic interconnections among multiple views in multi-view learning by proposing to maximize higher order correlations instead of pairwise ones, formulated as a low rank approximation problem solved using the generating polynomial method, and numerical results show it consistently outperforms prior methods.
Multi-view learning is frequently used in data science. The pairwise correlation maximization is a classical approach for exploring the consensus of multiple views. Since the pairwise correlation is inherent for two views, the extensions to more views can be diversified and the intrinsic interconnections among views are generally lost. To address this issue, we propose to maximize higher order correlations. This can be formulated as a low rank approximation problem with the higher order correlation tensor of multi-view data. We use the generating polynomial method to solve the low rank approximation problem. Numerical results on real multi-view data demonstrate that this method consistently outperforms prior existing methods.