Are Independently Estimated View Uncertainties Comparable? Unified Routing for Trusted Multi-View Classification
2026-04-10 • Machine Learning
Machine Learning
AI summaryⓘ
The authors explain that in multi-view classification, different views usually give separate confidence scores, but these scores aren't always directly comparable because views vary in quality and detail. To fix this, they propose TMUR, a method that separates how each view gathers evidence from how that evidence is combined, using a global router that looks at all views together. This helps the system weigh the views more fairly based on the specific sample, improving reliability. They also show theoretically why combining views independently can cause problems and why their unified approach works better.
multi-view classificationevidential fusionuncertainty estimationfeature spacesemantic granularityevidence extractionroutingload-balancingregularizationcross-view consistency
Authors
Yilin Zhang, Cai Xu, Haishun Chen, Ziyu Guan, Wei Zhao
Abstract
Trusted multi-view classification typically relies on a view-wise evidential fusion process: each view independently produces class evidence and uncertainty, and the final prediction is obtained by aggregating these independent opinions. While this design is modular and uncertainty-aware, it implicitly assumes that evidence from different views is numerically comparable. In practice, however, this assumption is fragile. Different views often differ in feature space, noise level, and semantic granularity, while independently trained branches are optimized only for prediction correctness, without any constraint enforcing cross-view consistency in evidence strength. As a result, the uncertainty used for fusion can be dominated by branch-specific scale bias rather than true sample-level reliability. To address this issue, we propose Trusted Multi-view learning with Unified Routing (TMUR), which decouples view-specific evidence extraction from fusion arbitration. TMUR uses view-private experts and one collaborative expert, and employs a unified router that observes the global multi-view context to generate sample-level expert weights. Soft load-balancing and diversity regularization further encourage balanced expert utilization and more discriminative expert specialization. We also provide theoretical analysis showing why independent evidential supervision does not identify a common cross-view evidence scale, and why unified global routing is preferable to branch-local arbitration when reliability is sample-dependent.