mlsanityData to 7 Oct 2026

Manifold Integrated Gradients: Riemannian Geometry for Feature Attribution

arXiv 2405.09800 ↗First posted 16 May 2024cs.LG cs.HC math.DGcited by 21

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. Cited by 21 papers in the map (data through 7 Oct 2026). Builds on 33 of its 60 references that are in the map. First posted on arXiv 16 May 2024.

Builds on

33 of 60 references.

References · 33

  1. 2017A Unified Approach to Interpreting Model Predictionscited by 6,488
  2. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  3. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
  4. 2017Axiomatic Attribution for Deep Networkscited by 3,350
  5. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
  6. 2017Learning Important Features Through Propagating Activation Differencescited by 1,454
  7. 2017SmoothGrad: removing noise by adding noisecited by 1,097
  8. 2014Striving for Simplicity: The All Convolutional Netcited by 983
  9. 2018Robustness May Be at Odds with Accuracycited by 873
  10. 2017Interpretable Explanations of Black Boxes by Meaningful Perturbationcited by 611
  11. 2017Interpretation of Neural Networks is Fragilecited by 443
  12. 2019Fooling LIME and SHAP: Adversarial Attacks on Post hoc Explanation Methodscited by 440
21 more
  1. 2016Generative Visual Manipulation on the Natural Image Manifoldcited by 348
  2. 2019Explanations can be manipulated and geometry is to blamecited by 215
  3. 2018Hyperspherical Variational Auto-Encoderscited by 188
  4. 2017Latent Space Oddity: on the Curvature of Deep Generative Modelscited by 169
  5. 2019Fooling Neural Network Interpretations via Adversarial Model Manipulationcited by 112
  6. 2019Continuous Hierarchical Representations with Poincar'e Variational Auto-Encoderscited by 111
  7. 2021Guided Integrated Gradients: An Adaptive Path Method for Removing Noisecited by 102
  8. 2019Mixed-curvature Variational Autoencoderscited by 85
  9. 2017The Riemannian Geometry of Deep Generative Modelscited by 81
  10. 2022A Rigorous Study of Integrated Gradients Method and Extensions to Internal Neuron Attributionscited by 64
  11. 2020Attribution in Scale and Spacecited by 61
  12. 2019Fast and Robust Shortest Paths on Manifolds Learned from Datacited by 25
  13. 2020Detecting Out-of-distribution Samples via Variational Auto-encoder with Reliable Uncertainty Estimationcited by 24
  14. 2020Investigating Saturation Effects in Integrated Gradientscited by 22
  15. 2023Which Models have Perceptually-Aligned Gradients? An Explanation via Off-Manifold Robustnesscited by 20
  16. 2021Fooling Partial Dependence via Data Poisoningcited by 16
  17. 2023IDGI: A Framework to Eliminate Explanation Noise from Integrated Gradientscited by 16
  18. 2022The Manifold Hypothesis for Gradient-Based Explanationscited by 15
  19. 2019Learning Weighted Submanifolds with Variational Autoencoders and Riemannian Variational Autoencoderscited by 14
  20. 2022Hyperbolic VAE via Latent Gaussian Distributionscited by 14
  21. 2020FAR: A General Framework for Attributional Robustnesscited by 13

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 2; 2025: 9; 2026: 10.

21 papers.

Sort:

By area:

  1. 2025Using the Path of Least Resistance to Explain Deep Networkscited by 3
  2. 2025Tangentially Aligned Integrated Gradients for User-Friendly Explanationscited by 3
  3. 2025SMA: Who Said That? Auditing Membership Leakage in Semi-Black-box RAG Controllingcited by 3
  4. 2024AdaptGrad: Adaptive Sampling to Reduce Noisecited by 2
  5. 2025Riemannian Integrated Gradients: A Geometric View of Explainable AIcited by 2
  6. 2026Counterfactual Explanations on Robust Perceptual Geodesicscited by 2
  7. 2026Auditing Sybil: Explaining Deep Lung Cancer Risk Prediction Through Generative Interventional Attributionscited by 2
  8. 2026Manifold-Aligned Guided Integrated Gradients for Reliable Feature Attributioncited by 2
  9. 2026Spectral Integrated Gradients for Coarse-to-Fine Feature Attributioncited by 2
  10. 2025Fast Fourier Correlation is a Highly Efficient and Accurate Feature Attribution Algorithm from the Perspective of Control Theory and Game Theorycited by 1
  11. 2025Measuring the (Un)Faithfulness of Concept-Based Explanationscited by 1
  12. 2025Path-Weighted Integrated Gradients for Interpretable Dementia Classificationcited by 1
9 more
  1. 2026Diffusion Integrated Gradients: Controllable Path Generation for Flexible Feature Attributioncited by 1
  2. 2024Comprehensive and Reliable Feature Attribution for Diverse Modalities and Models via Frequency-Domain Insightscited by 0
  3. 2025On the notion of missingness for path attribution explainability methods in medical settings: Guiding the selection of medically meaningful baselinescited by 0
  4. 2025Graph-based Integrated Gradients for Explaining Graph Neural Networkscited by 0
  5. 2026From Baselines to Transport Geodesics: Axiomatic Attribution via Optimal Generative Flowscited by 0
  6. 2026FRInGe: Distribution-Space Integrated Gradients with Fisher--Rao Geometrycited by 0
  7. 2026COGENT: Counterfactual Gaussian Explanations for Volumetric Medical Imagescited by 0
  8. 2026Pushing the (Decision) Boundaries: Dynamically Calibrating Differentially Private Noise to Explainability in Federated Learningcited by 0
  9. 2026Explaining Hyperbolic Neural Networks via Geometry-Aware Relevance Propagationcited by 0