On Gradient-like Explanation under a Black-box Setting: When Black-box Explanations Become as Good as White-box

TL;DR

This paper introduces extmethodAbr, a gradient-estimation-based explanation method that provides high-quality feature attributions using only query access, bridging the gap between black-box and white-box explanations.

Contribution

It proposes a novel gradient estimation approach for explanations that requires no internal model access, with rigorous theoretical properties and strong empirical performance.

Findings

Outperforms existing black-box attribution methods

Achieves competitive results with white-box gradient-based methods

Provides theoretically guaranteed explanation quality

Abstract

Attribution methods shed light on the explainability of data-driven approaches such as deep learning models by uncovering the most influential features in a to-be-explained decision. While determining feature attributions via gradients delivers promising results, the internal access required for acquiring gradients can be impractical under safety concerns, thus limiting the applicability of gradient-based approaches. In response to such limited flexibility, this paper presents \methodAbr~(gradient-estimation-based explanation), an approach that produces gradient-like explanations through only query-level access. The proposed approach holds a set of fundamental properties for attribution methods, which are mathematically rigorously proved, ensuring the quality of its explanations. In addition to the theoretical analysis, with a focus on image data, the experimental results empirically demonstrate the superiority of the proposed method over state-of-the-art black-box methods and its competitive performance compared to methods with full access.