Pattern-Guided Integrated Gradients

TL;DR

This paper introduces Pattern-Guided Integrated Gradients (PGIG), a novel explainability method for neural networks that combines two established techniques, inheriting their strengths and passing stress tests, outperforming other methods in image degradation benchmarks.

Contribution

The paper proposes PGIG, a new explainability method that integrates Integrated Gradients and PatternAttribution, improving robustness and performance over existing approaches.

Findings

PGIG passes stress tests that original methods fail.

PGIG outperforms nine alternative explainability methods in large-scale image degradation experiments.

PGIG inherits properties from both parent methods, enhancing explainability.

Abstract

Integrated Gradients (IG) and PatternAttribution (PA) are two established explainability methods for neural networks. Both methods are theoretically well-founded. However, they were designed to overcome different challenges. In this work, we combine the two methods into a new method, Pattern-Guided Integrated Gradients (PGIG). PGIG inherits important properties from both parent methods and passes stress tests that the originals fail. In addition, we benchmark PGIG against nine alternative explainability approaches (including its parent methods) in a large-scale image degradation experiment and find that it outperforms all of them.