Adaptive Disentangled Representation Learning for Incomplete Multi-View Multi-Label Classification

TL;DR

ADRL addresses the challenging iMvMLC problem by integrating adaptive missing-view completion, disentangled shared/private representations, and graph-based label semantics into a unified framework. It employs neighborhood-aware feature propagation for view recovery, mutual-information based objectives to balance cross-view consistency with view-specific independence, and prototype-guided label embeddings to drive discriminative view fusion. The method yields strong performance gains across six public datasets and a real-world NBA dataset under varying missingness, demonstrating robust feature representation learning and effective label correlation modeling. The approach offers practical impact for real-world multimodal, sparsely annotated data and suggests future integration with large-language-model priors to further boost representation quality and label understanding.

Abstract

Multi-view multi-label learning frequently suffers from simultaneous feature absence and incomplete annotations, due to challenges in data acquisition and cost-intensive supervision. To tackle the complex yet highly practical problem while overcoming the existing limitations of feature recovery, representation disentanglement, and label semantics modeling, we propose an Adaptive Disentangled Representation Learning method (ADRL). ADRL achieves robust view completion by propagating feature-level affinity across modalities with neighborhood awareness, and reinforces reconstruction effectiveness by leveraging a stochastic masking strategy. Through disseminating category-level association across label distributions, ADRL refines distribution parameters for capturing interdependent label prototypes. Besides, we formulate a mutual-information-based objective to promote consistency among shared representations and suppress information overlap between view-specific representation and other modalities. Theoretically, we derive the tractable bounds to train the dual-channel network. Moreover, ADRL performs prototype-specific feature selection by enabling independent interactions between label embeddings and view representations, accompanied by the generation of pseudo-labels for each category. The structural characteristics of the pseudo-label space are then exploited to guide a discriminative trade-off during view fusion. Finally, extensive experiments on public datasets and real-world applications demonstrate the superior performance of ADRL.

Figures (9)

• Figure 1: The main framework of our proposed ADRL. Different shapes signify different samples.
• Figure 2: Performance under different LMR with FMR fixed at 0.9.
• Figure 3: Performance under different FMR with LMR fixed at 0.9.
• Figure 4: Metric distributions under 70% LMR and 90% FMR.
• Figure 5: Parameter study of $\alpha$.