Density-Informed Pseudo-Counts for Calibrated Evidential Deep Learning
This work addresses uncertainty calibration in deep learning for applications requiring reliable predictions under distributional shift, though it is incremental as it builds on the existing EDL framework.
The paper tackled the problem of Evidential Deep Learning (EDL) conflating epistemic and aleatoric uncertainty, leading to overconfidence on out-of-distribution inputs, and introduced Density-Informed Pseudo-count EDL (DIP-EDL) to decouple these uncertainties, achieving improved robustness and uncertainty calibration under distributional shift.
Evidential Deep Learning (EDL) is a popular framework for uncertainty-aware classification that models predictive uncertainty via Dirichlet distributions parameterized by neural networks. Despite its popularity, its theoretical foundations and behavior under distributional shift remain poorly understood. In this work, we provide a principled statistical interpretation by proving that EDL training corresponds to amortized variational inference in a hierarchical Bayesian model with a tempered pseudo-likelihood. This perspective reveals a major drawback: standard EDL conflates epistemic and aleatoric uncertainty, leading to systematic overconfidence on out-of-distribution (OOD) inputs. To address this, we introduce Density-Informed Pseudo-count EDL (DIP-EDL), a new parametrization that decouples class prediction from the magnitude of uncertainty by separately estimating the conditional label distribution and the marginal covariate density. This separation preserves evidence in high-density regions while shrinking predictions toward a uniform prior for OOD data. Theoretically, we prove that DIP-EDL achieves asymptotic concentration. Empirically, we show that our method enhances interpretability and improves robustness and uncertainty calibration under distributional shift.