Aggregating explanation methods for stable and robust explainability

TL;DR

This paper proposes aggregating multiple explanation methods to improve the stability, robustness, and accuracy of neural network explanations, addressing the lack of consensus and vulnerability to attacks.

Contribution

It introduces aggregation schemes for explanations, demonstrating improved feature importance identification and robustness against adversarial attacks.

Findings

Aggregated explanations outperform individual methods in identifying important features.

Aggregated explanations are significantly more robust to adversarial attacks.

Aggregation reduces model uncertainty in explanations.

Abstract

Despite a growing literature on explaining neural networks, no consensus has been reached on how to explain a neural network decision or how to evaluate an explanation. Our contributions in this paper are twofold. First, we investigate schemes to combine explanation methods and reduce model uncertainty to obtain a single aggregated explanation. We provide evidence that the aggregation is better at identifying important features, than on individual methods. Adversarial attacks on explanations is a recent active research topic. As our second contribution, we present evidence that aggregate explanations are much more robust to attacks than individual explanation methods.