On the Over-Memorization During Natural, Robust and Catastrophic Overfitting

• URL: http://arxiv.org/abs/2310.08847v4
• Date: Sat, 14 Sep 2024 00:23:05 GMT
• Title: On the Over-Memorization During Natural, Robust and Catastrophic Overfitting
• Authors: Runqi Lin, Chaojian Yu, Bo Han, Tongliang Liu,
• Abstract summary: Overfitting negatively impacts the generalization ability of deep neural networks (DNNs) in both natural and adversarial training. We propose a general framework, Distraction Over-Memorization (DOM), which explicitly prevents over-memorization.
• Score: 58.613079045392446
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Overfitting negatively impacts the generalization ability of deep neural networks (DNNs) in both natural and adversarial training. Existing methods struggle to consistently address different types of overfitting, typically designing strategies that focus separately on either natural or adversarial patterns. In this work, we adopt a unified perspective by solely focusing on natural patterns to explore different types of overfitting. Specifically, we examine the memorization effect in DNNs and reveal a shared behaviour termed over-memorization, which impairs their generalization capacity. This behaviour manifests as DNNs suddenly becoming high-confidence in predicting certain training patterns and retaining a persistent memory for them. Furthermore, when DNNs over-memorize an adversarial pattern, they tend to simultaneously exhibit high-confidence prediction for the corresponding natural pattern. These findings motivate us to holistically mitigate different types of overfitting by hindering the DNNs from over-memorization training patterns. To this end, we propose a general framework, Distraction Over-Memorization (DOM), which explicitly prevents over-memorization by either removing or augmenting the high-confidence natural patterns. Extensive experiments demonstrate the effectiveness of our proposed method in mitigating overfitting across various training paradigms.

Related papers

• Towards Adversarial Robustness via Debiased High-Confidence Logit Alignment [24.577363665112706]
  Recent adversarial training techniques have utilized inverse adversarial attacks to generate high-confidence examples. Our investigation reveals that high-confidence outputs under inverse adversarial attacks are correlated with biased feature activation. We propose Debiased High-Confidence Adversarial Training (DHAT) to address this bias. DHAT achieves state-of-the-art performance and exhibits robust generalization capabilities across various vision datasets.
  arXiv  Detail & Related papers  (2024-08-12T11:56:06Z)
• Robust Overfitting Does Matter: Test-Time Adversarial Purification With FGSM [5.592360872268223]
  Defense strategies usually train deep neural networks (DNNs) for a specific adversarial attack method and can achieve good robustness in defense against this type of adversarial attack. However, when subjected to evaluations involving unfamiliar attack modalities, empirical evidence reveals a pronounced deterioration in the robustness of DNNs. Most defense methods often sacrifice the accuracy of clean examples in order to improve the adversarial robustness of DNNs.
  arXiv  Detail & Related papers  (2024-03-18T03:54:01Z)
• F$^2$AT: Feature-Focusing Adversarial Training via Disentanglement of Natural and Perturbed Patterns [74.03108122774098]
  Deep neural networks (DNNs) are vulnerable to adversarial examples crafted by well-designed perturbations. This could lead to disastrous results on critical applications such as self-driving cars, surveillance security, and medical diagnosis. We propose a Feature-Focusing Adversarial Training (F$2$AT) which enforces the model to focus on the core features from natural patterns.
  arXiv  Detail & Related papers  (2023-10-23T04:31:42Z)
• Improving Adversarial Robustness via Mutual Information Estimation [144.33170440878519]
  Deep neural networks (DNNs) are found to be vulnerable to adversarial noise. In this paper, we investigate the dependence between outputs of the target model and input adversarial samples from the perspective of information theory. We propose to enhance the adversarial robustness by maximizing the natural MI and minimizing the adversarial MI during the training process.
  arXiv  Detail & Related papers  (2022-07-25T13:45:11Z)
• Latent Boundary-guided Adversarial Training [61.43040235982727]
  Adrial training is proved to be the most effective strategy that injects adversarial examples into model training. We propose a novel adversarial training framework called LAtent bounDary-guided aDvErsarial tRaining.
  arXiv  Detail & Related papers  (2022-06-08T07:40:55Z)
• Improving White-box Robustness of Pre-processing Defenses via Joint Adversarial Training [106.34722726264522]
  A range of adversarial defense techniques have been proposed to mitigate the interference of adversarial noise. Pre-processing methods may suffer from the robustness degradation effect. A potential cause of this negative effect is that adversarial training examples are static and independent to the pre-processing model. We propose a method called Joint Adversarial Training based Pre-processing (JATP) defense.
  arXiv  Detail & Related papers  (2021-06-10T01:45:32Z)
• Attribute-Guided Adversarial Training for Robustness to Natural Perturbations [64.35805267250682]
  We propose an adversarial training approach which learns to generate new samples so as to maximize exposure of the classifier to the attributes-space. Our approach enables deep neural networks to be robust against a wide range of naturally occurring perturbations.
  arXiv  Detail & Related papers  (2020-12-03T10:17:30Z)
• Semantics-Preserving Adversarial Training [12.242659601882147]
  Adversarial training is a technique that improves adversarial robustness of a deep neural network (DNN) by including adversarial examples in the training data. We propose semantics-preserving adversarial training (SPAT) which encourages perturbation on the pixels that are shared among all classes. Experiment results show that SPAT improves adversarial robustness and achieves state-of-the-art results in CIFAR-10 and CIFAR-100.
  arXiv  Detail & Related papers  (2020-09-23T07:42:14Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.