arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2209.05668cs.LGcs.AI
收藏

类级Logit扰动

Class-Level Logit Perturbation

National Center for Applied Mathematics, Tianjin University(天津大学国家应用数学中心);University of Southern Queensland(南昆士兰大学)

Mengyang Li, Fengguang Su, Ou Wu, Ji Zhang

摘要:

特征、logits和标签是样本通过深度神经网络时的三个主要数据。特征扰动和标签扰动是近年来研究的热点。它们已被证明在各种深度学习方法中非常有用。例如,(对抗性)特征扰动可以提高学习模型的鲁棒性甚至泛化能力。然而,有限的研究已经明确探讨了罗吉特向量的扰动。这项工作讨论了现有的几种方法有关的类级logit扰动。建立了正、负数据增强与logit扰动引起的损失变化之间的统一观点。一个理论分析,以阐明为什么类水平的logit扰动是有用的。因此,提出了新的方法来明确学习扰动logits单标签和多标签分类任务。在基准图像分类数据集及其长尾版本上的大量实验表明,我们的学习方法具有竞争力的性能。由于它只对logit进行扰动,因此它可以作为一个插件与任何现有的分类算法融合。所有代码都可以在https://github.com/limengyang1992/lpl上找到。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑