Evidential Rule Learning for Interpretable Classification with Abstention

2026-08-06

AI Summary

This work addresses interpretable classification by proposing Fast Evidential Rule Learning (FERL), a novel approach that integrates Dempster-Shafer evidence theory directly into a fuzzy rule system. FERL uniquely enables a single-pass deterministic inference to simultaneously output belief and plausibility measures while supporting abstention under uncertainty—without requiring post-hoc calibration. The method exhibits Lipschitz stability and naturally produces set-valued predictions alongside explanations of anomalous feature contributions. Evaluated on 30 tabular datasets, FERL achieves an average accuracy 2.6% higher than state-of-the-art methods. Its set predictions demonstrate superior performance in both utility-adjusted accuracy and coverage, and it attains a near out-of-distribution detection AUROC of 77.7%, significantly outperforming strong baselines.

Abstract

Interpretable classification often requires more than accurate predictions for real-life deployment: models should be transparent about the evidence behind their decisions and abstain when they cannot decide reliably. We introduce Fast Evidential Rule Learning (FERL), a method that learns interpretable, accurate fuzzy rule models whose outputs are evidential. Unlike post-hoc calibration, FERL's belief, plausibility, and abstention capabilities arise directly from the fuzzy memberships in a single deterministic pass, with no auxiliary head, held-out set, or repeated inference. Our theoretical analysis further shows that FERL is Lipschitz stable, which means that its evidential outputs vary smoothly with the input. Against state-of-the-art rule learners, FERL is statistically significantly more accurate across a 30 tabular-dataset benchmark ($+2.6\%$ average accuracy over the second best). Its native set predictions attain the best utility-discounted accuracy among credal classifiers ($u_{65}/u_{80}=0.80/0.83$ vs.\ $0.79/0.80$ for the naive credal classifier), at higher set coverage ($0.92$ vs.\ $\le0.82$). FERL also matches dedicated out-of-distribution detectors on tabular near-OOD detection ($77.7$ vs.\ $77.4$ AUROC for the strongest baseline). Under detector-class-disjoint concept-bottleneck evaluation, its it is within $2.3$ AUROC points of the strongest dedicated detector on both CUB and AwA2, while attaining the best AwA2 AUPR-Out ($68.3$) and novel-class rejection ($57.2$), while being able to name which attributes are anomalous.

Problem

• interpretable classification
• evidential reasoning
• abstention
• rule learning
• uncertainty quantification

Innovation

• Evidential Reasoning
• Interpretable Rule Learning
• Abstention Mechanism
• Fuzzy Rule Models
• Lipschitz Stability