Ensemble-Based Dirichlet Modeling for Predictive Uncertainty and Selective Classification

TL;DR

This paper proposes an ensemble-based Dirichlet modeling approach to improve predictive uncertainty estimates and stability in neural network classifiers, enhancing their performance in uncertainty-guided tasks.

Contribution

It introduces a novel Dirichlet parameter estimation method using ensemble outputs, decoupling uncertainty estimation from fragile evidential training.

Findings

Ensemble-based Dirichlet estimates improve uncertainty stability.

The method enhances performance in uncertainty-guided applications.

It reduces variability across training runs compared to traditional methods.

Abstract

Neural network classifiers trained with cross-entropy loss achieve strong predictive accuracy but lack the capability to provide inherent predictive uncertainty estimates, thus requiring external techniques to obtain these estimates. In addition, softmax scores for the true class can vary substantially across independent training runs, which limits the reliability of uncertainty-based decisions in downstream tasks. Evidential Deep Learning aims to address these limitations by producing uncertainty estimates in a single pass, but evidential training is highly sensitive to design choices including loss formulation, prior regularization, and activation functions. Therefore, this work introduces an alternative Dirichlet parameter estimation strategy by applying a method of moments estimator to ensembles of softmax outputs, with an optional maximum-likelihood refinement step. This ensemble-based construction decouples uncertainty estimation from the fragile evidential loss design while also mitigating the variability of single-run cross-entropy training, producing explicit Dirichlet predictive distributions. Across multiple datasets, we show that the improved stability and predictive uncertainty behavior of these ensemble-derived Dirichlet estimates translate into stronger performance in downstream uncertainty-guided applications such as prediction confidence scoring and selective classification.