Fuzzy Multi-View Learning: Principles, Methodologies, and Future Directions

Abstract

Fuzzy multi-view learning (FMVL) integrates fuzzy modeling with multi-view learning to address uncertainty, incompleteness, heterogeneity, and unequal contributions across multiple views. Despite growing interest, existing studies remain dispersed across different fuzzy mechanisms, learning paradigms, and data settings, leaving the methodological structure of the field insufficiently clarified. This review organizes FMVL along three complementary dimensions: fundamental principles, methodological frameworks, and complex multi-view data conditions. We examine how fuzzy mechanisms support cross-view interaction, uncertainty representation, adaptive view contribution, reliability modeling, and information fusion, and organize existing methods into clustering-based, supervised, representation-based, structure-guided, and rule-based frameworks. We further discuss how FMVL adapts to challenging settings involving domain shifts, incomplete or corrupted views, limited supervision, heterogeneous modalities, evolving data, and decentralized learning. Representative applications demonstrate its potential across biomedical, visual, multimedia, transportation, and security tasks. Finally, we discuss the main challenges that still limit the robustness, scalability, interpretability, and reproducibility of FMVL.