Spectral Integrated Gradients for Coarse-to-Fine Feature Attribution

2026-05-19

AI Summary

This work addresses a key limitation of the standard Integrated Gradients method, whose straight-line integration path simultaneously introduces all input features and is prone to accumulating noisy gradients, thereby degrading attribution quality. To overcome this, the authors propose a novel integration path derived from the singular value decomposition (SVD) of the difference between the input and baseline. This path activates components sequentially in descending order of singular values, enabling a coarse-to-fine attribution process that progresses from global structure to local details. By incorporating SVD into path design for the first time, the method naturally encodes a hierarchy of feature importance and effectively suppresses attribution noise. Experiments across multiple image classification benchmarks demonstrate that the resulting attribution maps are visually sharper and achieve superior quantitative performance compared to existing path-based attribution approaches.

Abstract

Integrated Gradients (IG) is a widely adopted feature attribution method that satisfies desirable axiomatic properties. However, the choice of integration path significantly affects the quality of attributions, and the standard straight-line path introduces all input features simultaneously, often accumulating noisy gradients along the way. To address this limitation, we propose Spectral Integrated Gradients, which constructs integration paths based on singular value decomposition (SVD) of the baseline-to-input difference. By progressively activating singular components from largest to smallest, SIG introduces global structure before fine-grained details, naturally following a coarse-to-fine progression. Through extensive evaluation across diverse image classification datasets, we demonstrate that SIG produces cleaner attribution maps with reduced noise and achieves improved quantitative performance compared to existing path-based attribution methods. Our code is available at https://github.com/leekwoon/sig/.

Problem

• feature attribution
• Integrated Gradients
• integration path
• gradient noise
• coarse-to-fine

Innovation

• Spectral Integrated Gradients
• feature attribution
• singular value decomposition
• coarse-to-fine
• Integrated Gradients