Posterior Network: Uncertainty Estimation without OOD Samples via Density-Based Pseudo-Counts
This addresses the need for reliable uncertainty estimation in safe AI systems, offering a novel method that eliminates the unrealistic assumption of known OOD data at training time.
The paper tackles the problem of estimating aleatoric and epistemic uncertainty without requiring out-of-distribution (OOD) data during training, proposing the Posterior Network (PostNet) that uses Normalizing Flows to predict closed-form posterior distributions, achieving state-of-the-art results in OOD detection and uncertainty calibration under dataset shifts.
Accurate estimation of aleatoric and epistemic uncertainty is crucial to build safe and reliable systems. Traditional approaches, such as dropout and ensemble methods, estimate uncertainty by sampling probability predictions from different submodels, which leads to slow uncertainty estimation at inference time. Recent works address this drawback by directly predicting parameters of prior distributions over the probability predictions with a neural network. While this approach has demonstrated accurate uncertainty estimation, it requires defining arbitrary target parameters for in-distribution data and makes the unrealistic assumption that out-of-distribution (OOD) data is known at training time. In this work we propose the Posterior Network (PostNet), which uses Normalizing Flows to predict an individual closed-form posterior distribution over predicted probabilites for any input sample. The posterior distributions learned by PostNet accurately reflect uncertainty for in- and out-of-distribution data -- without requiring access to OOD data at training time. PostNet achieves state-of-the art results in OOD detection and in uncertainty calibration under dataset shifts.