Natural Posterior Network: Deep Bayesian Uncertainty for Exponential Family Distributions
This work addresses the need for reliable uncertainty awareness in ML models, offering a method applicable to any task with exponential family target distributions, but it is incremental as it builds on existing Bayesian and flow-based approaches.
The paper tackles the problem of uncertainty estimation in machine learning by proposing the Natural Posterior Network (NatPN), which leverages Normalizing Flows and Bayesian updates to achieve competitive performance in calibration and out-of-distribution detection across classification, regression, and count prediction tasks.
Uncertainty awareness is crucial to develop reliable machine learning models. In this work, we propose the Natural Posterior Network (NatPN) for fast and high-quality uncertainty estimation for any task where the target distribution belongs to the exponential family. Thus, NatPN finds application for both classification and general regression settings. Unlike many previous approaches, NatPN does not require out-of-distribution (OOD) data at training time. Instead, it leverages Normalizing Flows to fit a single density on a learned low-dimensional and task-dependent latent space. For any input sample, NatPN uses the predicted likelihood to perform a Bayesian update over the target distribution. Theoretically, NatPN assigns high uncertainty far away from training data. Empirically, our extensive experiments on calibration and OOD detection show that NatPN delivers highly competitive performance for classification, regression and count prediction tasks.