Towards Robust Uncertainty-Aware Incomplete Multi-View Classification

TL;DR

This work proposes the Alternating Progressive Learning Network (APLN), specifically designed to enhance EDL-based methods in incomplete MVC scenarios by mitigating bias from corrupted observed data by first applying coarse imputation, followed by mapping the data to a latent space.

Abstract

Handling incomplete data in multi-view classification is challenging, especially when traditional imputation methods introduce biases that compromise uncertainty estimation. Existing Evidential Deep Learning (EDL) based approaches attempt to address these issues, but they often struggle with conflicting evidence due to the limitations of the Dempster-Shafer combination rule, leading to unreliable decisions. To address these challenges, we propose the Alternating Progressive Learning Network (APLN), specifically designed to enhance EDL-based methods in incomplete MVC scenarios. Our approach mitigates bias from corrupted observed data by first applying coarse imputation, followed by mapping the data to a latent space. In this latent space, we progressively learn an evidence distribution aligned with the target domain, incorporating uncertainty considerations through EDL. Additionally, we introduce a conflict-aware Dempster-Shafer combination rule (DSCR) to better handle conflicting evidence. By sampling from the learned distribution, we optimize the latent representations of missing views, reducing bias and enhancing decision-making robustness. Extensive experiments demonstrate that APLN, combined with DSCR, significantly outperforms traditional methods, particularly in environments characterized by high uncertainty and conflicting evidence, establishing it as a promising solution for incomplete multi-view classification.

Figures (10)

• Figure 1: Comparison between traditional EDL-based methods and our approach for handling IMVC. While traditional methods often result in high uncertainty due to inter-view conflicts after completing the missing views, our method leverages conflict-aware Dempster-Shafer (CA-DST) for progressive refinement, leading to more reliable inference and significantly reduced uncertainty.
• Figure 2: Workflow of ALPN.
• Figure 3: Conflictive degree visualization.
• Figure 4: T-SNE visualization of Evidence distributions across different stages of the APLN on the Handwritten dataset. UMAE-F represents the feature training stage with incomplete views, UMAE-V shows the distribution after the view-specific alignment, and UMAE-J illustrates the joint evidence distribution after final integration.
• Figure 6: Robustness to conflict: Performance comparison on conflict datasets with varying missing rates across different models.

Theorems & Definitions (1)

• Definition 1