Maximum Entropy Baseline for Integrated Gradients

TL;DR

This paper introduces the Maximum Entropy Baseline for Integrated Gradients, aiming to improve explanation credibility by addressing baseline selection ambiguity and evaluating explanation reliability.

Contribution

It proposes a new uniform baseline for IG, an improved evaluation method, and analyzes the invariance of IG baselines from an information perspective.

Findings

The Maximum Entropy Baseline aligns with the uninformative property of IG baselines.

The new baseline enhances the reliability of explanations.

Extensive experiments validate the effectiveness of the proposed approach.

Abstract

Integrated Gradients (IG), one of the most popular explainability methods available, still remains ambiguous in the selection of baseline, which may seriously impair the credibility of the explanations. This study proposes a new uniform baseline, i.e., the Maximum Entropy Baseline, which is consistent with the "uninformative" property of baselines defined in IG. In addition, we propose an improved ablating evaluation approach incorporating the new baseline, where the information conservativeness is maintained. We explain the linear transformation invariance of IG baselines from an information perspective. Finally, we assess the reliability of the explanations generated by different explainability methods and different IG baselines through extensive evaluation experiments.