mlsanityData to 7 Oct 2026

Building Reliable Explanations of Unreliable Neural Networks: Locally Smoothing Perspective of Model Interpretation

arXiv 2103.14332 ↗First posted 26 Mar 2021cs.CV cs.LGcited by 12

Not on the map. Why?

Abstract (arXiv)

We present a novel method for reliably explaining the predictions of neural networks. We consider an explanation reliable if it identifies input features relevant to the model output by considering the input and the neighboring data points. Our method is built on top of the assumption of smooth landscape in a loss function of the model prediction: locally consistent loss and gradient profile. A theoretical analysis established in this study suggests that those locally smooth model explanations are learned using a batch of noisy copies of the input with the L1 regularization for a saliency map. Extensive experiments support the analysis results, revealing that the proposed saliency maps retrieve the original classes of adversarial examples crafted against both naturally and adversarially trained models, significantly outperforming previous methods. We further demonstrated that such good performance results from the learning capability of this method to identify input features that are truly relevant to the model output of the input and the neighboring data points, fulfilling the requirements of a reliable explanation.

Key facts. Cited by 12 papers in the map (data through 7 Oct 2026). Builds on 33 of its 41 references that are in the map. First posted on arXiv 26 Mar 2021.

Builds on

33 of 41 references.

References · 33

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  3. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,690
  4. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  5. 2013Intriguing properties of neural networkscited by 5,114
  6. 2016Image-to-Image Translation with Conditional Adversarial Networkscited by 4,510
  7. 2017Axiomatic Attribution for Deep Networkscited by 3,350
  8. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
  9. 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,510
  10. 2017Learning Important Features Through Propagating Activation Differencescited by 1,454
  11. 2017High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANscited by 1,176
  12. 2017SmoothGrad: removing noise by adding noisecited by 1,097
21 more
  1. 2014Striving for Simplicity: The All Convolutional Netcited by 983
  2. 2018Robustness May Be at Odds with Accuracycited by 873
  3. 2018RISE: Randomized Input Sampling for Explanation of Black-box Modelscited by 825
  4. 2017Interpretable Explanations of Black Boxes by Meaningful Perturbationcited by 611
  5. 2018Towards Robust Interpretability with Self-Explaining Neural Networkscited by 505
  6. 2017Interpretation of Neural Networks is Fragilecited by 443
  7. 2019Fooling LIME and SHAP: Adversarial Attacks on Post hoc Explanation Methodscited by 440
  8. 2018Adversarially Robust Generalization Requires More Datacited by 318
  9. 2017Real Time Image Saliency for Black Box Classifierscited by 246
  10. 2017Improving the Adversarial Robustness and Interpretability of Deep Neural Networks by Regularizing their Input Gradientscited by 236
  11. 2019Explanations can be manipulated and geometry is to blamecited by 215
  12. 2018Robustness via curvature regularization, and vice versacited by 174
  13. 2019Adversarial Robustness through Local Linearizationcited by 146
  14. 2018Explaining Image Classifiers by Counterfactual Generationcited by 144
  15. 2016Technical Report on the CleverHans v2.1.0 Adversarial Examples Librarycited by 123
  16. 2019XRAI: Better Attributions Through Regionscited by 116
  17. 2019Fooling Neural Network Interpretations via Adversarial Model Manipulationcited by 112
  18. 2018Adversarial vulnerability for any classifiercited by 78
  19. 2020There and Back Again: Revisiting Backpropagation Saliency Methodscited by 60
  20. 2019Interpretable and Fine-Grained Visual Explanations for Convolutional Neural Networkscited by 49
  21. 2018Towards Explanation of DNN-based Prediction with Guided Feature Inversioncited by 31

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 1; 2022: 3; 2023: 3; 2024: 3; 2025: 0; 2026: 2.

12 papers.

  1. 2021Robust Explainability: A Tutorial on Gradient-Based Attribution Methods for Deep Neural Networkscited by 48
  2. 2022What Makes a Good Explanation?: A Harmonized View of Properties of Explanationscited by 21
  3. 2022The Manifold Hypothesis for Gradient-Based Explanationscited by 15
  4. 2023MoreauGrad: Sparse and Robust Interpretation of Neural Networks via Moreau Envelopecited by 5
  5. 2023PAMI: partition input and aggregate outputs for model interpretationcited by 2
  6. 2022Illuminating Salient Contributions in Neuron Activation with Attribution Equilibriumcited by 1
  7. 2023On Model Explanations with Transferable Neural Pathwayscited by 1
  8. 2024Explainable Transformer Prototypes for Medical Diagnosescited by 1
  9. 2024HOLMES: HOLonym-MEronym based Semantic inspection for Convolutional Image Classifierscited by 0
  10. 2024Robust Explainable Recommendationcited by 0
  11. 2026Training for Trustworthy Saliency Maps: Adversarial Training Meets Feature-Map Smoothingcited by 0
  12. 2026Improving Adversarial Robustness of Attribution via Implicit Regularizationcited by 0