mlsanityData to 7 Oct 2026

Class-Level Logit Perturbation

arXiv 2209.05668 ↗First posted 13 Sep 2022cs.LG cs.AIcited by 4

Not on the map. Why?

Abstract (arXiv)

Features, logits, and labels are the three primary data when a sample passes through a deep neural network. Feature perturbation and label perturbation receive increasing attention in recent years. They have been proven to be useful in various deep learning approaches. For example, (adversarial) feature perturbation can improve the robustness or even generalization capability of learned models. However, limited studies have explicitly explored for the perturbation of logit vectors. This work discusses several existing methods related to class-level logit perturbation. A unified viewpoint between positive/negative data augmentation and loss variations incurred by logit perturbation is established. A theoretical analysis is provided to illuminate why class-level logit perturbation is useful. Accordingly, new methodologies are proposed to explicitly learn to perturb logits for both single-label and multi-label classification tasks. Extensive experiments on benchmark image classification data sets and their long-tail versions indicated the competitive performance of our learning method. As it only perturbs on logit, it can be used as a plug-in to fuse with any existing classification algorithms. All the codes are available at https://github.com/limengyang1992/lpl.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Builds on 42 of its 74 references that are in the map. First posted on arXiv 13 Sep 2022.

Builds on

42 of 74 references.

References · 42

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  3. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  4. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  5. 2015Rethinking the Inception Architecture for Computer Visioncited by 5,451
  6. 2017mixup: Beyond Empirical Risk Minimizationcited by 5,155
  7. 2017Squeeze-and-Excitation Networkscited by 4,763
  8. 2016Wide Residual Networkscited by 3,057
  9. 2014Conditional Generative Adversarial Netscited by 2,152
  10. 2017Improved Regularization of Convolutional Neural Networks with Cutoutcited by 1,623
  11. 2019Class-Balanced Loss Based on Effective Number of Samplescited by 1,239
  12. 2018Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labelscited by 994
30 more
  1. 2019Learning Imbalanced Datasets with Label-Distribution-Aware Margin Losscited by 987
  2. 2016InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Netscited by 858
  3. 2019Large-Scale Long-Tailed Recognition in an Open Worldcited by 704
  4. 2019Square Attack: a query-efficient black-box adversarial attack via random searchcited by 682
  5. 2016Making Deep Neural Networks Robust to Label Noise: a Loss Correction Approachcited by 663
  6. 2019Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and Entailmentcited by 663
  7. 2020Long-tail learning via logit adjustmentcited by 605
  8. 2016Conditional Image Synthesis With Auxiliary Classifier GANscited by 562
  9. 2019BBN: Bilateral-Branch Network with Cumulative Learning for Long-Tailed Visual Recognitioncited by 403
  10. 2014Training Deep Neural Networks on Noisy Labels with Bootstrappingcited by 322
  11. 2019Multi-Label Image Recognition with Graph Convolutional Networkscited by 320
  12. 2016Large-Margin Softmax Loss for Convolutional Neural Networkscited by 296
  13. 2019Adversarial Examples Improve Image Recognitioncited by 283
  14. 2018Large Scale Fine-Grained Categorization and Domain-Specific Transfer Learningcited by 174
  15. 2020To be Robust or to be Fair: Towards Fairness in Adversarial Trainingcited by 137
  16. 2020Distribution-Balanced Loss for Multi-Label Classification in Long-Tailed Datasetscited by 122
  17. 2015Relay Backpropagation for Effective Learning of Deep Convolutional Neural Networkscited by 115
  18. 2019Implicit Semantic Data Augmentation for Deep Networkscited by 106
  19. 2021MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognitioncited by 99
  20. 2020Learnable Boundary Guided Adversarial Trainingcited by 93
  21. 2017Learning with Average Top-k Losscited by 52
  22. 2021Adversarial Example Detection for DNN Models: A Review and Experimental Comparisoncited by 51
  23. 2021Adversarial Robustness under Long-Tailed Distributioncited by 50
  24. 2019Learning Data Manipulation for Augmentation and Weightingcited by 44
  25. 2016DisturbLabel: Regularizing CNN on the Loss Layercited by 41
  26. 2020Learning to Purify Noisy Labels via Meta Soft Label Correctorcited by 39
  27. 2021Adversarial Training Helps Transfer Learning via Better Representationscited by 36
  28. 2020ProSelfLC: Progressive Self Label Correction for Training Robust Deep Neural Networkscited by 33
  29. 2019Label Smoothing and Logit Squeezing: A Replacement for Adversarial Training?cited by 18
  30. 2020Reconstruction Regularized Deep Metric Learning for Multi-label Image Classificationcited by 7

Built on by

Cited by year (2026 to 7 Oct)

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

4 papers.

  1. 2023Data Optimization in Deep Learning: A Surveycited by 6
  2. 2025Mitigating Spurious Correlations with Causal Logit Perturbationcited by 0
  3. 2026Gradient Perturbation: Learning to Perturb Gradients for Adaptive Trainingcited by 0
  4. 2026Learning to Perturb Hidden Representations for Generalizable Deep Learningcited by 0