Proximity-Based Evidence Retrieval for Uncertainty-Aware Neural Networks

2025-09-11

AI Summary

Uncertainty-aware neural networks often rely on global, fixed decision thresholds, leading to unreliable and uninterpretable predictions. To address this, we propose a proximity-based evidence retrieval framework that enables adaptive, instance-level decision-making. Our method retrieves nearest-neighbor instances of a test sample in the embedding space and fuses their predictive distributions using Dempster–Shafer evidence theory to dynamically compute an instance-specific confidence threshold. This yields auditable and traceable decisions grounded in local evidence rather than global heuristics. Crucially, we abandon fixed global cutoffs, instead constructing thresholds from a small set of semantically relevant neighbors—significantly reducing high-confidence misclassifications. Evaluated on CIFAR-10 and CIFAR-100 using BiT and ViT backbones, our approach matches or surpasses entropy-based thresholding baselines in accuracy and uncertainty calibration, while maintaining low auditing overhead and strong interpretability.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Security and Privacy: Data transparency and provenance

Abstract

This work proposes an evidence-retrieval mechanism for uncertainty-aware decision-making that replaces a single global cutoff with an evidence-conditioned, instance-adaptive criterion. For each test instance, proximal exemplars are retrieved in an embedding space; their predictive distributions are fused via Dempster-Shafer theory. The resulting fused belief acts as a per-instance thresholding mechanism. Because the supporting evidences are explicit, decisions are transparent and auditable. Experiments on CIFAR-10/100 with BiT and ViT backbones show higher or comparable uncertainty-aware performance with materially fewer confidently incorrect outcomes and a sustainable review load compared with applying threshold on prediction entropy. Notably, only a few evidences are sufficient to realize these gains; increasing the evidence set yields only modest changes. These results indicate that evidence-conditioned tagging provides a more reliable and interpretable alternative to fixed prediction entropy thresholds for operational uncertainty-aware decision-making.

Problem

• Proposes evidence-retrieval mechanism for uncertainty-aware decision-making
• Replaces global cutoff with instance-adaptive evidence-conditioned criterion
• Uses proximal exemplars and Dempster-Shafer fusion for transparent thresholds

Innovation

• Proximity-based evidence retrieval mechanism
• Dempster-Shafer theory fusion method
• Per-instance adaptive thresholding criterion