The goal of multi-view learning is to learn latent patterns from various data sources. Most of previous research focused on fitting feature embedding in target tasks. There is very limited research on ...