mlsanityData to 7 Oct 2026

Learning Sample Reweighting for Accuracy and Adversarial Robustness

arXiv 2210.11513 ↗First posted 20 Oct 2022cs.LGcited by 6

Not on the map. Why?

Abstract (arXiv)

There has been great interest in enhancing the robustness of neural network classifiers to defend against adversarial perturbations through adversarial training, while balancing the trade-off between robust accuracy and standard accuracy. We propose a novel adversarial training framework that learns to reweight the loss associated with individual training samples based on a notion of class-conditioned margin, with the goal of improving robust generalization. We formulate weighted adversarial training as a bilevel optimization problem with the upper-level problem corresponding to learning a robust classifier, and the lower-level problem corresponding to learning a parametric function that maps from a sample's textitmulti-class margin to an importance weight. Extensive experiments demonstrate that our approach consistently improves both clean and robust accuracy compared to related methods and state-of-the-art baselines.

Key facts. Cited by 6 papers in the map (data through 7 Oct 2026). Builds on 34 of its 42 references that are in the map. First posted on arXiv 20 Oct 2022.

Builds on

34 of 42 references.

References · 34

  1. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  2. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  3. 2013Intriguing properties of neural networkscited by 5,114
  4. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,861
  5. 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,283
  6. 2016Wide Residual Networkscited by 3,057
  7. 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,510
  8. 2020Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attackscited by 1,442
  9. 2020Fast is better than free: Revisiting adversarial trainingcited by 723
  10. 2018Learning to Reweight Examples for Robust Deep Learningcited by 670
  11. 2017One pixel attack for fooling deep neural networkscited by 579
  12. 2017Provable defenses against adversarial examples via the convex outer adversarial polytopecited by 564
22 more
  1. 2019Meta-Learning with Implicit Gradientscited by 523
  2. 2020Overfitting in adversarially robust deep learningcited by 504
  3. 2018Meta-Learning with Latent Embedding Optimizationcited by 476
  4. 2019Provably Robust Deep Learning via Adversarially Trained Smoothed Classifierscited by 328
  5. 2020Attacks Which Do Not Kill Training Make Adversarial Learning Strongercited by 242
  6. 2021Improving Robustness using Generated Datacited by 233
  7. 2020Geometry-aware Instance-reweighted Adversarial Trainingcited by 198
  8. 2021Data Augmentation Can Improve Robustnesscited by 193
  9. 2018MMA Training: Direct Input Space Margin Maximization through Adversarial Trainingcited by 171
  10. 2021Probabilistic Margins for Instance Reweighting in Adversarial Trainingcited by 57
  11. 2019Instance adaptive adversarial training: Improved accuracy tradeoffs in neural netscited by 55
  12. 2018Deep Bilevel Learningcited by 48
  13. 2020How benign is benign overfitting?cited by 40
  14. 2019Provable robustness against all adversarial lp-perturbations for p≥ 1cited by 39
  15. 2021Understanding the role of importance weighting for deep learningcited by 38
  16. 2020Calibrated Surrogate Losses for Adversarially Robust Classificationcited by 36
  17. 2019Identifying Mislabeled Instances in Classification Datasetscited by 26
  18. 2017Robustness to Adversarial Examples through an Ensemble of Specialistscited by 21
  19. 2021Reweighting Augmented Samples by Minimizing the Maximal Expected Losscited by 21
  20. 2020A Closer Look at Accuracy vs. Robustnesscited by 14
  21. 2020Are Adversarial Examples Created Equal? A Learnable Weighted Minimax Risk for Robustness under Non-uniform Attackscited by 14
  22. 2018Gradient Agreement as an Optimization Objective for Meta-Learningcited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 2; 2024: 2; 2025: 1; 2026: 1.

6 papers.

  1. 2023An Introduction to Bi-level Optimization: Foundations and Applications in Signal Processing and Machine Learningcited by 39
  2. 2024A Short Survey on Importance Weighting for Machine Learningcited by 15
  3. 2023Identifying Adversarially Attackable and Robust Samplescited by 5
  4. 2024Improving Generalization via Meta-Learning on Hard Samplescited by 5
  5. 2025Revisiting Meta-Learning with Noisy Labels: Reweighting Dynamics and Theoretical Guaranteescited by 0
  6. 2026Rethinking Loss Reweighting for Imbalance Learning as an Inverse Problem: A Neural Collapse Point of Viewcited by 0