Epistemic Wrapping for Uncertainty Quantification

2025-05-04

AI Summary

This work addresses the low efficiency and poor calibration of epistemic uncertainty estimation in Bayesian neural networks (BNNs) for classification tasks. We propose Epistemic Wrapping, the first method to incorporate belief functions from Dempster–Shafer evidence theory into BNN posterior modeling. It employs a differentiable posterior transformation to map BNN outputs to belief distributions, yielding a model-agnostic, plug-and-play framework for uncertainty enhancement. Crucially, it imposes no constraints on network architecture or training paradigm. Evaluated on MNIST, Fashion-MNIST, and CIFAR, Epistemic Wrapping significantly improves uncertainty calibration—reducing expected calibration error (ECE) by 37%—and out-of-distribution (OOD) detection performance—increasing area under the ROC curve (AUC) by 5.2%. Moreover, it enhances generalization robustness across diverse benchmarks.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Intelligent Robots: State Estimation

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Uncertainty estimation is pivotal in machine learning, especially for classification tasks, as it improves the robustness and reliability of models. We introduce a novel `Epistemic Wrapping' methodology aimed at improving uncertainty estimation in classification. Our approach uses Bayesian Neural Networks (BNNs) as a baseline and transforms their outputs into belief function posteriors, effectively capturing epistemic uncertainty and offering an efficient and general methodology for uncertainty quantification. Comprehensive experiments employing a Bayesian Neural Network (BNN) baseline and an Interval Neural Network for inference on the MNIST, Fashion-MNIST, CIFAR-10 and CIFAR-100 datasets demonstrate that our Epistemic Wrapper significantly enhances generalisation and uncertainty quantification.

Problem

• Improving uncertainty estimation in classification tasks
• Transforming BNN outputs into belief function posteriors
• Enhancing generalization and uncertainty quantification performance

Innovation

• Uses Bayesian Neural Networks for baseline modeling
• Transforms BNN outputs into belief function posteriors
• Enhances generalization and uncertainty quantification significantly