On the Practicality of Deterministic Epistemic Uncertainty

TL;DR

This paper evaluates deterministic epistemic uncertainty methods in deep neural networks, showing they scale well to real-world vision tasks but suffer from poor calibration under distributional shifts, affecting their practical deployment.

Contribution

The paper provides a taxonomy of DUMs, evaluates their calibration under shifts, and extends them to semantic segmentation, highlighting their strengths and limitations.

Findings

DUMs perform well on out-of-distribution detection.

DUMs scale to realistic vision tasks.

Calibration issues under distributional shifts limit practicality.

Abstract

A set of novel approaches for estimating epistemic uncertainty in deep neural networks with a single forward pass has recently emerged as a valid alternative to Bayesian Neural Networks. On the premise of informative representations, these deterministic uncertainty methods (DUMs) achieve strong performance on detecting out-of-distribution (OOD) data while adding negligible computational costs at inference time. However, it remains unclear whether DUMs are well calibrated and can seamlessly scale to real-world applications - both prerequisites for their practical deployment. To this end, we first provide a taxonomy of DUMs, and evaluate their calibration under continuous distributional shifts. Then, we extend them to semantic segmentation. We find that, while DUMs scale to realistic vision tasks and perform well on OOD detection, the practicality of current methods is undermined by poor calibration under distributional shifts.