Density Uncertainty Layers for Reliable Uncertainty Estimation
This addresses the need for reliable uncertainty estimation in safety-critical deep learning applications, representing a novel method for a known bottleneck.
The paper tackles the problem of unreliable predictive uncertainty estimates in deep neural networks by proposing a novel criterion linking uncertainty to input data density, and introduces density uncertainty layers that achieve more reliable uncertainty estimates and robust out-of-distribution detection on benchmarks like UCI and CIFAR-10/100.
Assessing the predictive uncertainty of deep neural networks is crucial for safety-related applications of deep learning. Although Bayesian deep learning offers a principled framework for estimating model uncertainty, the common approaches that approximate the parameter posterior often fail to deliver reliable estimates of predictive uncertainty. In this paper, we propose a novel criterion for reliable predictive uncertainty: a model's predictive variance should be grounded in the empirical density of the input. That is, the model should produce higher uncertainty for inputs that are improbable in the training data and lower uncertainty for inputs that are more probable. To operationalize this criterion, we develop the density uncertainty layer, a stochastic neural network architecture that satisfies the density uncertain criterion by design. We study density uncertainty layers on the UCI and CIFAR-10/100 uncertainty benchmarks. Compared to existing approaches, density uncertainty layers provide more reliable uncertainty estimates and robust out-of-distribution detection performance.