mlsanityData to 7 Oct 2026

Are Classification Robustness and Explanation Robustness Really Strongly Correlated? An Analysis Through Input Loss Landscape

arXiv 2403.06013 ↗First posted 9 Mar 2024cs.LG cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

This paper delves into the critical area of deep learning robustness, challenging the conventional belief that classification robustness and explanation robustness in image classification systems are inherently correlated. Through a novel evaluation approach leveraging clustering for efficient assessment of explanation robustness, we demonstrate that enhancing explanation robustness does not necessarily flatten the input loss landscape with respect to explanation loss - contrary to flattened loss landscapes indicating better classification robustness. To deeply investigate this contradiction, a groundbreaking training method designed to adjust the loss landscape with respect to explanation loss is proposed. Through the new training method, we uncover that although such adjustments can impact the robustness of explanations, they do not have an influence on the robustness of classification. These findings not only challenge the prevailing assumption of a strong correlation between the two forms of robustness but also pave new pathways for understanding relationship between loss landscape and explanation loss.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 49 of its 65 references that are in the map. First posted on arXiv 9 Mar 2024.

Builds on

49 of 65 references.

References · 49

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  3. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  4. 2013Intriguing properties of neural networkscited by 5,114
  5. 2018MobileNetV2: Inverted Residuals and Linear Bottleneckscited by 4,773
  6. 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,283
  7. 2017MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applicationscited by 3,827
  8. 2016Wide Residual Networkscited by 3,057
  9. 2016Towards Evaluating the Robustness of Neural Networkscited by 2,976
  10. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
  11. 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,510
  12. 2017Learning Important Features Through Propagating Activation Differencescited by 1,454
37 more
  1. 2020Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attackscited by 1,442
  2. 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,259
  3. 2014Striving for Simplicity: The All Convolutional Netcited by 983
  4. 2017Ensemble Adversarial Training: Attacks and Defensescited by 833
  5. 2015Distillation as a Defense to Adversarial Perturbations against Deep Neural Networkscited by 675
  6. 2019Adversarial Training for Free!cited by 638
  7. 2020Captum: A unified and generic model interpretability library for PyTorchcited by 542
  8. 2017Feature Squeezing: Detecting Adversarial Examples in Deep Neural Networkscited by 484
  9. 2017Interpretation of Neural Networks is Fragilecited by 443
  10. 2019Fooling LIME and SHAP: Adversarial Attacks on Post hoc Explanation Methodscited by 440
  11. 2019Unlabeled Data Improves Adversarial Robustnesscited by 415
  12. 2018Dynamic Few-Shot Visual Learning without Forgettingcited by 385
  13. 2019Using Pre-Training Can Improve Model Robustness and Uncertaintycited by 373
  14. 2018On the Robustness of Interpretability Methodscited by 310
  15. 2017PixelDefend: Leveraging Generative Models to Understand and Defend against Adversarial Examplescited by 243
  16. 2019Explanations can be manipulated and geometry is to blamecited by 215
  17. 2023Better Diffusion Models Further Improve Adversarial Trainingcited by 207
  18. 2020Bag of Tricks for Adversarial Trainingcited by 199
  19. 2016Grad-CAM: Why did you say that?cited by 159
  20. 2022(Certified!!) Adversarial Robustness for Free!cited by 130
  21. 2019Fooling Neural Network Interpretations via Adversarial Model Manipulationcited by 112
  22. 2019Defensive Quantization: When Efficiency Meets Robustnesscited by 95
  23. 2020Smooth Adversarial Trainingcited by 89
  24. 2018Interpretable Deep Learning under Firecited by 74
  25. 2020On the Loss Landscape of Adversarial Training: Identifying Challenges and How to Overcome Themcited by 59
  26. 2019Robust Attribution Regularizationcited by 46
  27. 2020Proper Network Interpretability Helps Adversarial Robustness in Classificationcited by 41
  28. 2022SAFARI: Versatile and Efficient Evaluations for Robustness of Interpretabilitycited by 28
  29. 2023Randomized Adversarial Training via Taylor Expansioncited by 28
  30. 2022Robust Explanation Constraints for Neural Networkscited by 23
  31. 2022Understanding and Combating Robust Overfitting via Input Loss Landscape Analysis and Regularizationcited by 22
  32. 2021Robust Models Are More Interpretable Because Attributions Look Normalcited by 21
  33. 2022Foiling Explanations in Deep Neural Networkscited by 14
  34. 2019Self-supervised Adversarial Trainingcited by 12
  35. 2018Adversarial Vision Challengecited by 9
  36. 2020Using Single-Step Adversarial Training to Defend Iterative Adversarial Examplescited by 9
  37. 2021Defense Against Explanation Manipulationcited by 6

Built on by

Not cited yet.