The Unreasonable Effectiveness of Deep Evidential Regression

Abstract

There is a significant need for principled uncertainty reasoning in machine learning systems as they are increasingly deployed in safety-critical domains. A new approach with uncertainty-aware regression-based neural networks (NNs), based on learning evidential distributions for aleatoric and epistemic uncertainties, shows promise over traditional deterministic methods and typical Bayesian NNs, notably with the capabilities to disentangle aleatoric and epistemic uncertainties. Despite some empirical success of Deep Evidential Regression (DER), there are important gaps in the mathematical foundation that raise the question of why the proposed technique seemingly works. We detail the theoretical shortcomings and analyze the performance on synthetic and real-world data sets, showing that Deep Evidential Regression is a heuristic rather than an exact uncertainty quantification. We go on to discuss corrections and redefinitions of how aleatoric and epistemic uncertainties should be extracted from NNs.

11 pages, 25 figures

References in corpus (6)

• On Calibration of Modern Neural Networks
• Towards calibrated and scalable uncertainty representations for neural networks
• Quantifying Aleatoric and Epistemic Uncertainty in Machine Learning: Are Conditional Entropy and Mutual Information Appropriate Measures?
• Performance of a GPU- and Time-Efficient Pseudo 3D Network for Magnetic Resonance Image Super-Resolution and Motion Artifact Reduction
• Vision-Based Autonomous Car Racing Using Deep Imitative Reinforcement Learning
• Pitfalls of Epistemic Uncertainty Quantification through Loss Minimisation

Cited by in corpus (1)

• An Epistemic Position-Based Click Model: From Interactions to Epistemic Distributions of Relevance and Bias