Few-class Fidelity: Evaluating Explanations of Real-conditions CNN classifiers with Optimized Perturbations

2026-06-23

AI Summary

This study addresses the lack of a unified and reliable fidelity evaluation framework for explainable artificial intelligence (XAI) methods in real-world few-class classification scenarios. The authors propose a fidelity assessment framework tailored to practical deployment conditions, which quantifies the faithfulness of CNN explanation methods by generating in-distribution perturbations optimized to elicit model uncertainty. The framework aligns these quantitative measures with human-perception-based object localization and segmentation metrics for validation. Innovatively integrating perturbation generation with few-class application contexts, it uncovers complex interactions among domain characteristics, data distributions, and explanation methods. Experiments on medical and natural image tasks demonstrate that the choice of XAI method must be closely aligned with specific domain and data properties, thereby validating the effectiveness and practical utility of the proposed evaluation mechanism.

Abstract

The wide use of Convolutional Neural Networks (CNN) in numerous domains and real-world classification applications is justified by their high precision and automation speed, helping users concentrate on higher-expertise tasks. To better understand the models and avoid bias during deployment, eXplainable Artificial Intelligence (XAI) techniques can be used after training. But as the list of XAI solutions expand, comparisons between them diverge, and consensus over their evaluation cannot be reached. This paper proposes a variation of Fidelity-based XAI metrics, with a focus on real-conditions applications, where the number of classes is often low. The approach generates in-distribution, uncertainty-provoking perturbations, to ensure proper measurement of the XAI methods faithfulness. As demonstration of the evaluation framework usefulness, it is compared with human-centric object localization and segmentation metrics. Once applied to both medical and natural imaging applications, it highlights the intricate correlation between domain, data curation, and XAI solution choices in order to validate training of a new CNN model.

Problem

• Few-class
• Fidelity
• XAI evaluation
• CNN classifiers
• Real-conditions

Innovation

• Few-class Fidelity
• Optimized Perturbations
• XAI Evaluation
• In-distribution Uncertainty
• CNN Explainability