New framework RobustLT tackles adversarial training on imbalanced datasets

Researchers have developed a new framework called RobustLT to improve adversarial training for deep neural networks, particularly on datasets with long-tail distributions. The framework addresses limitations in current methods, such as skewed training objectives due to class imbalance and unstable adversarial distributions. By adaptively adjusting perturbations during training, RobustLT aims to enhance both adversarial robustness and class balance, as demonstrated in extensive experiments. AI

影响 Improves robustness of models on imbalanced datasets, potentially increasing their reliability in real-world applications.

排序理由 Academic paper introducing a new method for adversarial training. [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CV 阅读 →

• Adversarial training
• RobustLT
• Deep neural networks

• 论文
• 安全

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →