Predictive Uncertainty Estimation via Prior Networks

Estimating uncertainty is important to improving the safety of AI systems. Recently baseline tasks and metrics have been defined and several practical methods for estimating uncertainty developed. However, these approaches attempt to model distributional uncertainty either implicitly through model uncertainty or as data uncertainty. This work proposes a new framework for modeling predictive uncertainty called Prior Networks (PNs) which explicitly models distributional uncertainty. PNs do this by parameterizing a prior distribution over predictive distributions. This work focuses on uncertainty for classification and evaluates PNs on the tasks of identifying out-of-distribution (OOD) samples and detecting misclassification on the MNIST dataset, where they are found to outperform previous methods. Experiments on synthetic and MNIST data show that unlike previous methods PNs are able to distinguish between data and distributional uncertainty.