Rethinking Epistemic Uncertainty in Node Classification through Information Growth

2026-10-02

AI Summary

This study addresses the lack of theoretical guarantees in existing graph evidential learning regarding the reduction of epistemic uncertainty as information increases. We construct a statistical framework for information growth and establish consistency criteria to reveal the limitations of current methods. Furthermore, we propose a graph Bootstrap ensemble approach that captures both data and process uncertainties through resampling and stochastic training. Experimental validation, conducted under a projected graph data-generating process and an information growth protocol, demonstrates that the proposed method achieves superior uncertainty reduction compared to standard ensembles as information accumulates. This work provides a reliable theoretical and methodological foundation for uncertainty quantification in graph-structured data.

Abstract

Epistemic uncertainty should decrease as additional information about the data-generating process (DGP) becomes available to the predictor. Yet, existing graph evidential deep learning (EDL) methods for node classification typically construct epistemic uncertainty from graph-specific properties and evaluate it on downstream tasks such as out-of-distribution detection, which do not test its reducibility as information about the DGP increases. To make reducibility directly testable, we introduce a statistical framework for studying epistemic uncertainty under information growth. Our framework specifies an information-growth experimental protocol and a consistency criterion for epistemic predictors, while using projective graph DGPs to ensure that growing graphs, which in general need not provide increasing information about the same DGP, constitute coherent observations of the same underlying process. We show that EDL methods do not explicitly estimate data uncertainty arising from a single finite graph observation and instead regulate epistemic uncertainty through model hyperparameters, precluding consistency, as corroborated by controlled information-growth experiments. As an alternative, we propose graph bootstrap ensembles, capturing both data and procedural uncertainty through graph resampling and randomized training. Under the same experimental protocol, these ensembles exhibit epistemic uncertainty reduction beyond standard deep ensembles. These findings support bootstrap ensembles as candidate consistent epistemic predictors under information growth.

Problem

• Epistemic Uncertainty
• Node Classification
• Information Growth
• Evidential Deep Learning
• Data-generating Process

Innovation

• Epistemic Uncertainty
• Information Growth
• Graph Bootstrap Ensembles
• Evidential Deep Learning
• Projective Graph DGPs