Spectral Integrated Gradients for Coarse-to-Fine Feature Attribution

2026/05/19 by Soyeon Kim, Seongwoo Lim, Kyowoon Lee +1
Computer Science · Engineering · #Attribution #Code (set theory) #Face and Expression Recognition #Feature (linguistics) #Noise (video) #Path (computing) #Pattern recognition (psychology) #Remote-Sensing Image Classification #Singular value decomposition #Stochastic Gradient Optimization Techniques #acm:68T05 #acm:68T07 #acm:68T45 #cs.AI #cs.CV #cs.LG #msc:68T05 #msc:68T07 #msc:68T45

paper · pdf · open access · doi:10.48550/arxiv.2605.19607

published in arXiv (Cornell University) (Cornell University) · 21 pages, 13 figures, 9 tables. Accepted to ACM KDD 2026; includes appendices

openalex publication_date 2026/05/19 · openalex created_date 2026/05/21 · openalex updated_date 2026/07/28 · arxiv created 2026/08/05 · arxiv updated 2026/08/06

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/.

Citations

• State-Covering Trajectory Stitching for Diffusion Planners
• Local Manifold Approximation and Projection for Manifold-Aware Diffusion Planning
• One Wave To Explain Them All: A Unifying Perspective On Feature Attribution
• Manifold Integrated Gradients: Riemannian Geometry for Feature Attribution
• Refining Diffusion Planner for Reliable Behavior Synthesis by Automatic Detection of Infeasible Plans
• A Consistent and Efficient Evaluation Strategy for Attribution Methods
• Look at the Variance! Efficient Black-box Explanations with Sobol-based Sensitivity Analysis
• Guided Integrated Gradients: An Adaptive Path Method for Removing Noise
• Do Input Gradients Highlight Discriminative Features?
• Debugging Tests for Model Explanations
• Shortcut learning in deep neural networks
• Attribution in Scale and Space
• Unmasking Clever Hans Predictors and Assessing What Machines Really Learn
• A Benchmark for Interpretability Methods in Deep Neural Networks
• The (Un)reliability of saliency methods
• SmoothGrad: removing noise by adding noise
• Learning Important Features Through Propagating Activation Differences
• Axiomatic Attribution for Deep Networks
• Top-down Neural Attention by Excitation Backprop
• "Why Should I Trust You?": Explaining the Predictions of Any Classifier
• Deep Residual Learning for Image Recognition
• Rethinking the Inception Architecture for Computer Vision
• Striving for Simplicity: The All Convolutional Net
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
• The Approximation of One Matrix by Another of Lower Rank