Multi-View Projection Learning via Adaptive Graph Embedding for Dimensionality Reduction

In order to explore complex structures and relationships hidden in data, plenty of graph-based dimensionality reduction methods have been widely investigated extended the multi-view learning field. For reduction, key point is extracting complementary compatible information analyze underlying structure samples, which still a challenging task. We propose novel algorithm that integrates for each view into one framework. Because prespecified graph derived from original noisy high-dimensional data usually low-quality, subspace constructed based on such also low-quality. To obtain optimal we framework learns affinity low-dimensional representation all views performs it jointly. Although noisy, local them valuable. Therefore, process, introduce predefined graphs feature graph. Moreover, assigning weight its importance essential learning, proposed GoMPL automatically allocates an appropriate process. The obtained then adopted learn projection matrix individual by embedding. provide effective alternate update method jointly view. conduct many experiments various benchmark datasets evaluate effectiveness method.