Non-Aligned Multi-View Multi-Label Classification Via Learning View-Specific Labels

In the multi-view multi-label (MVML) classification problem, multiple views are simultaneously associated with semantic representations. Multi-view learning inevitably has problems of consistency, diversity, and non-alignment among correlation labels. Most existing methods for non-aligned assume that each view a common or shared label set, but because single cannot contain entire information, they often learn suboptimal results. Based on this, this paper proposes method learns view-specific labels (LVSL), aiming to explicitly mine information low-rank structures in unified model framework. Furthermore, alleviate insufficient available we thoroughly explored global local structural Specifically, first, there is consistency between space then construct turn. Second, enrich original consistent hidden Finally, contribution weight combined complementary decision-making stage, extend handle nonlinear data. The results proposed compared state-of-the-art algorithms several datasets validate its effectiveness.