Label-Guided Representation Learning for Incomplete Multi-View Multi-Label Classification

Incomplete multi-view multi-label classification addresses scenarios where views and labels are partially missing. While existing methods treat labels solely as supervision signals, they overlook the semantic structure inherent in partial annotations. To tackle this problem, we propose Label-Guided Representation Learning (LGRL) to systematically exploit label semantics as structural prior information. Specifically, we construct a semantic-aware mixture prior via learnable category prototypes to explicitly guide representation extraction across views. Furthermore, categoryspecific conditional posteriors are introduced to leverage these prototypes as Bayesian experts to govern multi-view fusion. Besides, we further derive a principled label-driven information bottleneck objective balancing reconstruction sufficiency with cross-view consistency, enabling category-conditional reasoning. Extensive experimental results demonstrate the effectiveness of LGRL across benchmark datasets as well as applications in sports analytics and medical imaging.