Robust Attribution Regularization

TL;DR

This paper introduces a new training approach for neural networks that enhances the robustness of their attribution explanations, grounded in axiomatic attribution theory and integrated gradients, with promising experimental results.

Contribution

It proposes novel training objectives for robust attribution based on integrated gradients, extending existing robust optimization methods and providing theoretical analysis.

Findings

Effective robust attribution achieved in experiments

Identifies challenges requiring better optimization or architectures

Generalizes previous theories for attribution robustness

Abstract

An emerging problem in trustworthy machine learning is to train models that produce robust interpretations for their predictions. We take a step towards solving this problem through the lens of axiomatic attribution of neural networks. Our theory is grounded in the recent work, Integrated Gradients (IG), in axiomatically attributing a neural network's output change to its input change. We propose training objectives in classic robust optimization models to achieve robust IG attributions. Our objectives give principled generalizations of previous objectives designed for robust predictions, and they naturally degenerate to classic soft-margin training for one-layer neural networks. We also generalize previous theory and prove that the objectives for different robust optimization models are closely related. Experiments demonstrate the effectiveness of our method, and also point to intriguing problems which hint at the need for better optimization techniques or better neural network architectures for robust attribution training.