Manifold Integrated Gradients: Riemannian Geometry for Feature Attribution

TL;DR

This paper enhances Integrated Gradients by incorporating Riemannian geometry to produce more reliable, perceptually intuitive feature attributions that are robust against adversarial attacks in deep learning models.

Contribution

It introduces a manifold-aware adaptation of Integrated Gradients that aligns attribution paths with the data's intrinsic geometry, improving explanation quality and robustness.

Findings

Geodesic-based IG produces more intuitive visualizations.

Method increases robustness to attributional attacks.

Experiments on real-world datasets validate effectiveness.

Abstract

In this paper, we dive into the reliability concerns of Integrated Gradients (IG), a prevalent feature attribution method for black-box deep learning models. We particularly address two predominant challenges associated with IG: the generation of noisy feature visualizations for vision models and the vulnerability to adversarial attributional attacks. Our approach involves an adaptation of path-based feature attribution, aligning the path of attribution more closely to the intrinsic geometry of the data manifold. Our experiments utilise deep generative models applied to several real-world image datasets. They demonstrate that IG along the geodesics conforms to the curved geometry of the Riemannian data manifold, generating more perceptually intuitive explanations and, subsequently, substantially increasing robustness to targeted attributional attacks.