mlsanityData to 14 Oct 2026

Provable Robust Saliency-based Explanations

arXiv 2212.14106 ↗First posted 28 Dec 2022cs.LG cs.AIcited by 1

Not on the map. Why?

Abstract (arXiv)

To foster trust in machine learning models, explanations must be faithful and stable for consistent insights. Existing relevant works rely on the ℓp distance for stability assessment, which diverges from human perception. Besides, existing adversarial training (AT) associated with intensive computations may lead to an arms race. To address these challenges, we introduce a novel metric to assess the stability of top-k salient features. We introduce R2ET which trains for stable explanation by efficient and effective regularizer, and analyze R2ET by multi-objective optimization to prove numerical and statistical stability of explanations. Moreover, theoretical connections between R2ET and certified robustness justify R2ET's stability in all attacks. Extensive experiments across various data modalities and model architectures show that R2ET achieves superior stability against stealthy attacks, and generalizes effectively across different explanation methods.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 71 of its 102 references that are in the map. First posted on arXiv 28 Dec 2022.

Builds on

71 of 102 references.

References · 71

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  3. 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
  4. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  5. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,070
  6. 2017Axiomatic Attribution for Deep Networkscited by 3,384
  7. 2015Learning Deep Features for Discriminative Localizationcited by 2,436
  8. 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,517
  9. 2017Learning Important Features Through Propagating Activation Differencescited by 1,458
  10. 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,268
  11. 2018Sanity Checks for Saliency Mapscited by 1,156
  12. 2017SmoothGrad: removing noise by adding noisecited by 1,109
38 more
  1. 2018Robustness May Be at Odds with Accuracycited by 880
  2. 2020Fast is better than free: Revisiting adversarial trainingcited by 727
  3. 2019Explaining Machine Learning Classifiers through Diverse Counterfactual Explanationscited by 614
  4. 2019ERASER: A Benchmark to Evaluate Rationalized NLP Modelscited by 530
  5. 2017Interpretation of Neural Networks is Fragilecited by 444
  6. 2019Is Attention Interpretable?cited by 389
  7. 2016European Union regulations on algorithmic decision-making and a "right to explanation"cited by 344
  8. 2021On Explainability of Graph Neural Networks via Subgraph Explorationscited by 281
  9. 2019On Gradient Descent Ascent for Nonconvex-Concave Minimax Problemscited by 270
  10. 2019Adversarial Attacks and Defenses in Images, Graphs and Text: A Reviewcited by 237
  11. 2019Explanations can be manipulated and geometry is to blamecited by 216
  12. 2020PGM-Explainer: Probabilistic Graphical Model Explanations for Graph Neural Networkscited by 211
  13. 2020GraphLIME: Local Interpretable Model Explanations for Graph Neural Networkscited by 209
  14. 2017Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulationcited by 185
  15. 2018Robustness via curvature regularization, and vice versacited by 175
  16. 2019Explainability Techniques for Graph Convolutional Networkscited by 161
  17. 2020Understanding and Mitigating the Tradeoff Between Robustness and Accuracycited by 149
  18. 2019Interpretations are useful: penalizing explanations to align neural networks with prior knowledgecited by 145
  19. 2021On the Convergence and Robustness of Adversarial Trainingcited by 130
  20. 2019Explaining Deep Neural Networks with a Polynomial Time Algorithm for Shapley Values Approximationcited by 115
  21. 2019Fooling Neural Network Interpretations via Adversarial Model Manipulationcited by 112
  22. 2021Towards Robust and Reliable Algorithmic Recoursecited by 102
  23. 2018L-Shapley and C-Shapley: Efficient Model Interpretation for Structured Datacited by 100
  24. 2019Neural Network Attributions: A Causal Perspectivecited by 77
  25. 2015Learning with a Strong Adversarycited by 68
  26. 2018Theoretical Impediments to Machine Learning With Seven Sparks from the Causal Revolutioncited by 66
  27. 2021On the Adversarial Robustness of Causal Algorithmic Recoursecited by 56
  28. 2021Robust Explainability: A Tutorial on Gradient-Based Attribution Methods for Deep Neural Networkscited by 48
  29. 2019Robust Attribution Regularizationcited by 47
  30. 2018Smooth Loss Functions for Deep Top-k Classificationcited by 46
  31. 2020Towards Robust Explanations for Deep Neural Networkscited by 43
  32. 2020Transferable, Controllable, and Inconspicuous Adversarial Attacks on Person Re-identification With Deep Mis-Rankingcited by 42
  33. 2020Proper Network Interpretability Helps Adversarial Robustness in Classificationcited by 41
  34. 2015Loss Functions for Top-k Error: Analysis and Insightscited by 38
  35. 2020Smoothed Geometry for Robust Attributioncited by 38
  36. 2019On the Consistency of Top-k Surrogate Lossescited by 35
  37. 2020Towards Understanding Fast Adversarial Trainingcited by 33
  38. 2018Theoretical Analysis of Adversarial Learning: A Minimax Approachcited by 31

50 most cited of 71 All 71 (JSON)

Built on by

Cited by year

Citing papers per year: 2025: 1.

1 paper.

  1. 2025Pixel-level Certified Explanations via Randomized Smoothingcited by 4